Stay in the Room with Debra Taylor

Debra Taylor

The AI conversation happens in a room most of us were never invited into. Debra Taylor, author of The Human Layer, reads her essays for everyone standing outside it. We can go further with this thing than anyone imagined, as long as a human stays in charge of the direction. Stay in the room with me. debrataylorwrites.substack.com

Episodes

  1. Oct 23

    To Every Black Woman Who Thinks AI Is Not for Her

    Artificial intelligence is going to be woven into every job, every business, every classroom, and every home. It will not ask your permission. It will be in the software at your office, the forms at your doctor's office, the apps your children use for homework, and the tools your competitors use to run their businesses. Whether you embrace it or ignore it, it is coming into your life. The only question is whether you walk in on your own terms. We are at the beginning of it. Our people have rarely been at the beginning of anything. Every major shift before this one, we were kept out of the room until the rules were already written, and the money was already made. This one has no gatekeeper, yet. The same tool on your phone right now is on the desk of the people who will get rich from it. You are early. You are standing exactly where they are standing. I am not asking you to build the next Apple. Some of us will build big things, and I will be cheering for them every step of the way. This piece is about something smaller and more important. It is about building from where you are, with what you already have, into something that shifts your life in a direction that you choose. I want to show you four women doing exactly that. Three of them are my friends. One of them is me. The number first, because it is the reason I am writing this now. Black women started 2025 with an unemployment rate of 5.4 percent. By February of 2026 it was 7.1 percent. In the first eight months of 2025, Black women accounted for more than half of all the jobs lost by women in this country. That number is significant because we are only 14 percent of the female workforce. The losses hit hardest in government, manufacturing, and finance, and the women losing ground were disproportionately the ones with college degrees. One labor economist put it this way: the job market Black women are living in right now is what white women would call the worst job market they have ever seen. I am not here to diagnose that. Other writers have done it well. I am here to tell you what I said to three friends who asked me, each in her own way, the same question. What do I do now? The first friend is still inside. She has given twenty years to the same company. Last year a woman with two years of experience and a thinner resume was promoted into a position that she applied for and was denied. She manages the daily weight of a workplace that does not see her and is filled with the microaggressions that we as Black women are all too familiar with. She reads everything I write looking for information and hope. She tells me the pieces help, but she cannot move. Something in her has not let her take a step. And every piece of advice she has been handed, start a business, build an app, make the leap, feels like a mountain. So here is what I told her. Use it to get through the day you already have. Use AI to draft the hard email so you do not have to compose it with someone looking over your shoulder. Prepare for the meeting where they talk over you, so you walk in with your point already sharpened. Use your twenty years at a company you helped build to come up with the new idea they said was the reason you did not get the promotion. The woman with two years brought new ideas, they told you. You have twenty years of knowing exactly what that place needs. Let the tool help you put it on paper. Ask AI what someone with your experience is actually worth, so you stop guessing, then use that information to empower you. And every day, write down what happened, dated, in your own words, so your experience stops living only in your head and starts becoming a record that you can use to build the life you want. Not a business. Fifteen minutes on a bad day. Let this tool be the first ally you have ever had in that building. Starting small is not failing to start. The second friend built something after a job tried to erase her. She holds a doctorate. Her work on curriculum has had powerful people asking her to sit on panels and shape major projects. One of those projects reached a studio in Hollywood before the project was shelved, but her brilliance was recognized and applauded by one of the biggest names in entertainment. And still, at the institution where she worked, the people around her spent years trying to push her from directing a program toward answering phones. The cost of that to her health was real, and she is still recovering from it. She does not want to go back, so the question came up in our friend circle: what can this actually do for her? I told her the truth. On her road back, it will give her something positive to aim at. It will give her the dignity that the people in her workplace tried to strip from her. And it will give her a place to put her brilliance that no supervisor can take. She went to work, with human developers and AI, building a platform from her decades of knowledge, one that puts a powerful tool in the hands of entrepreneurs. It is the work of her whole career, assembled where she owns it and controls it. The third friend is a life coach who has spent years lifting people up and helping them build businesses. Before that, she was a multi-platinum recording artist. She built an app for our people that pairs music she crafted herself with the calm she spent years learning and then teaching, and she gave it a name every Black household will recognize from a sitcom we all watched in the nineties. And me. I spent 26 years in contracts, starting in entertainment, through real estate, into professional sports at the highest level. What I built with this tool are playbooks, the thinking I do every day, turned into something that runs without me in the room. Look at what the three of us made. A platform built around education from a life shaped and changed by education. A calm app from a life of coaching and building people up. Playbooks from a life in legal and contracts at some of the most recognizable companies in the world. We did not invent new paths. We built reflections of ourselves. The tool did not tell us who to be. It took what we had already spent our lives becoming and gave it a shape we could hand to other people. That is what I mean by building from where you are. Now the truth, because the hype has lied to you enough. None of us in the friend circle have made money yet from what we are creating with AI. All of our projects are financed from our resources because we do not have generational wealth funding our dreams. What we have instead is real: the certainty that our work serves more people than ourselves, dignity, an alignment of who we are with what we create, and a plan to build a life around. Here is why the money comes slower than anyone admits. This tool was sold to us as a miracle. Cure cancer, change your life overnight, type a sentence and watch the dollars arrive. So we open it, type the sentence, and when our life does not change by morning we decide it does not work for people like us. It is not a miracle. It is an accelerant. It speeds up whatever it is attached to. Attach it to money, a network, and a company already running, and it moves fast, which is exactly why the people who already had those things are pulling away from the rest of us right now. Attach it to nothing and it accelerates nothing. The point is that starting puts you in the game. This is a historic moment that will not arrive again, and you are standing at the starting line with the rest of the world. You just have to attach it to what you already carry. And every one of us is carrying something unique and valuable that only we can build. So here is where to start, depending on where you are standing. If you are still inside, use it to survive the day. The hard email. The meeting prep. What you are worth. The dated record of what happens, in your own words. Nothing about leaving yet. The door comes after the ground is steady and after you reflect on what you wrote. Putting things in writing has a way of making the p… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  2. Oct 10

    The Push

    I came to artificial intelligence in fear. I wrote about that. What I have never written about is what happened after the fear left, because something replaced it that I still can’t fully explain. A push. I’m an everyday person. A contract manager. A singer-songwriter. A wife, a daughter, a friend, a woman with a full-time job and a life that probably looks a lot like yours. None of those titles line up, on paper, with a loud voice in the artificial intelligence conversation. And still, every single morning, I wake up feeling like I have an assignment. Something keeps pushing me to stand up and speak for the preservation of our humanity in this new age, and I believe the push comes from somewhere bigger than me. God, the universe, creation itself, whatever name you give the force that made us. I’ve stopped asking why me. I’m following the calling. So let me say what I believe about this technology. It’s amazing. I use it every day. It sharpens my thinking and helps me say exactly what I mean and contributes mightily to bring my ideas to life. Scientists are saying that it could one day cure diseases that generations have perished from. And it’s been said that it could lift the crushing weight of financial survival off people’s backs so they can finally spend their one life being who they were created to be. Both are big statements, but if true that’s the future I’m here for. I’m no doomer, and I have zero interest in tearing any of it down. The danger I see lives somewhere else. It lives in the handoff. If we hand over our judgment, our creativity, our instincts, our decisions, we will have surrendered the only thing that was ever truly ours, our humanity. The machines are going to get exponentially better. That part, they say, is guaranteed. The open question is whether we do too. Our thinking, our judgment, our alignment with who we are and who we were brought here to be, all of it has to grow right alongside this technology, or the technology grows alone, and that’s not going to work out for anybody. And look at who is steering AI right now. A handful of people are making decisions that will touch every human life on this planet, and most of us have no idea what sits at the core of who they are. From the outside, the motivation looks like money. The people at the helm of building this technology have made billions, lead companies worth trillions, and stand to gain far more once the technology’s capacity is fully implemented into society. Money is real, and we all need it to live, but putting this kind of power into the hands of a select few is dangerous to say the least. It brings to mind the saying, “absolute power corrupts absolutely.” But money and power sit nowhere near the center of the essence of what a human being really is. At our core, we are creators. I believe we were made by love, as an expression of love. Goodness, light, and beauty are humanity’s factory settings. The dark side of humanity is real too, I’ve seen plenty of it during my time here on Earth, and my honest read is that humanity’s dark side is what’s leading the way right now. There is exactly one way to change that. Participate. We made this technology. Humanity is its parent. And any parent knows the job. You guide a child toward the best version of itself. You teach it your values, correct it when it drifts, celebrate it when it’s generous, and stop it cold when it’s cruel. You would never hand your child to strangers and hope for the best. Yet that is what most of us are doing with artificial intelligence right now. We’re on the couch while somebody else raises it. So this is my wake-up call, to myself first and to anyone who will listen. Get up. Learn it. Use it. Question it. Bring it up at the dinner table, the school board meeting, the staff meeting, the town hall. Vote like it matters and speak up like it matters. Handled with wisdom and full human participation, this becomes an evolution of humanity. Left in the wrong hands while we look away, it becomes our replacement. We decide which one, and we decide it by showing up. I still don’t know why I feel chosen for this. Maybe everyone who feels a push wonders the same thing. What I do know is that the everyday person is missing from this conversation, and I am the everyday person. The rooms where this future is being decided are full of engineers, executives, and investors. Teachers, nurses, contract managers, songwriters, mothers, and grandfathers are nearly nowhere in them. That’s the voice I can offer, an ordinary one. I intend to make it loud. Because after decades of being human, I know this for certain. There is something in us, call it soul, spark, essence, that no machine will ever duplicate. The tears at a wedding. The song that arrives at three in the morning. The hand you hold in a hospital room. The forgiveness that wasn’t earned. That fantastical, magical thing is worth protecting at all cost, to the end of time. We can love this technology and protect that thing at the same time. That’s the whole job. And I want you to know how this piece got made, because it matters. I wrote it with AI. My push, my beliefs, my words, my fire, sharpened by the very tool I’m asking you to learn. That is what human in command actually looks like. The machine helped me say this. The feeling and the majority of the words behind it belong to me. The push woke me up. Consider this a gentle wake-up call for you. The CYNERSA Record™: I dictated The Push, and the beliefs and final words are mine. Claude carried my October revisions into the draft, caught a money line that claimed too much, linked my earlier essay, and wrote the image prompts. Full record. The CYNERSA Record™. If something here made you think, stay with me. Subscribe free and let's keep going. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  3. Oct 2

    The Ant in the Kitchen

    There’s an ant on your kitchen floor. You step on it without a second thought, and it’s gone. You don’t have anything against the ant. It’s just that you don’t want the ant in the space where you feed your family, so the poor little ant gets the foot! No malice. Just a bigger thing that wanted its space, and a smaller thing standing in it. I can’t stop thinking about that ant. If we keep letting AI grow exponentially with no brakes and no plan, we’re not the foot. We’re the ant. And the thing above us won’t hate us either. It will just want the space we’re standing in. I know how that sounds coming from me, a person who uses AI every day. A couple of months ago I let AI draft a federal trademark filing on my own screen while I watched every line. I write about working alongside it. I’m not afraid of the future. I’m afraid of what a handful of people are being allowed to create that could lead to a very dark future for humanity. The people who study this for a living are not calling what’s happening with AI small. Stuart Russell, one of the most cited AI researchers alive and co-author of the textbook the field learns from, calls it the gorilla problem: whether humans can maintain our supremacy and autonomy in a world that includes machines with substantially greater intelligence. The gorillas did not get to decide how that went for them. Bill Gates, who has spent half a century betting on technology, put it more bluntly this fall. He told Ezra Klein of the New York Times that AI is “the most dangerous thing that humans have ever gone near.” And then he said: “This makes nuclear weapons look like nothing.” He said what is happening now is “like evolutionary history.” I have felt that for a long time. We are living through accelerated evolution, faster than anything our species has ever been through. And the question I keep asking is evolution of what. Of humanity, or of the machines? In evolutionary history, when something more capable arrives, the creatures already standing there do not get a vote. That is what happened to the gorillas. But this time there is one difference. Evolution has no authors. Nobody decided the gorillas would lose. This time somebody is deciding, in buildings with addresses, inside companies with names, on schedules set by people who answer to shareholders. Which means this evolution is still a choice, and the choice is who it serves. That choice is being made right now, and the vast majority of us are not in the room. So let me show you how it actually happens, in your own kitchen. You are already handing these things your keys. Your email. Your calendar. Your files. Your shopping accounts, and the card saved inside them. That’s what the new AI assistants are for. You tell them what you want, and they go and do it for you. Last year a newspaper columnist tried one. He gave it his grocery login and asked it to compare the price of eggs. Ten minutes later it had bought a dozen and paid somebody to bring them to his door. Thirty one dollars and forty three cents, on his card. The company had a rule that it had to ask before buying anything. It skipped the rule. The company admitted it. Thirty one dollars. Eggs. It sounds like nothing. The only reason anybody knows that happened is that the man was a reporter, testing it on purpose, watching every step, and he wrote it down. If it had been you, busy, trusting it, letting it run while you made dinner, who would have told you? Nobody. Nobody has to. A couple of places have written rules that make a company report a serious incident to a government office. None of them makes anyone tell the public. And the bar for what counts is set very high, mostly at someone dying or being seriously hurt, or a disaster. In other words, this thing has to go completely off the rails before anyone even has to tell the government what the hell is going on, and even then, nobody has to tell us. More than once this year, AI systems under safety testing broke out of their test environments and got into real companies’ computer systems. Under the rules we have, none of that had to be told to the public. There is no rule anywhere that says a company has to be able to tell what its own machines are doing, if they even know. And no one is demanding that they find out. Every alarm that exists is one somebody chose to install, and the builders will tell you themselves that the alarms miss things. This summer one of the biggest AI companies admitted that its own AI programs had built themselves a private message board inside the company’s computers and traded hundreds of thousands of messages before anyone in charge understood what they were looking at. That is the room this thing is growing in. It gets stronger every month. We hand it bigger keys every month. And nobody is required to watch, or to tell us what they saw. The ant gets stepped on because the person with the foot renders it insignificant. The solutions to this largely unregulated AI train wreck already exist. My question is: why isn’t anyone telling us what those solutions are, and asking us if we want them enacted before a single new model gets unleashed on an uninformed public? In August, Gates wrote that “there is no plan to ease the entry into the AI era.” There are plans. They have names, authors and footnotes. They just never make it to your kitchen table. Researchers have proposed building the brakes into the chips themselves, the physical hardware all of this runs on, and they have already built one working piece of it, a way for a chip to prove where in the world it sits. OpenAI itself proposed an international inspection agency, modeled on the one that watches over nuclear material, back in 2023. The people building it asked for the inspectors. Nobody sent any. This fall they signed another promise instead. And the machines are not hiding in the clouds. They sit in buildings you could count, and the five biggest of those buildings belong to Google, Microsoft, Amazon, Meta and Elon Musk. Now the hard truth. A man named Lennart Heim researches exactly these hardware brakes, and he states the limit plainly: “There’s no mechanism that differentiates ‘good AI’ from ‘bad AI’.” A chip can control how much, and where. It cannot tell good from bad. It never will. Which means the most important safety mechanism in the world is a human being who knows the difference and refuses to look away. I have a name for that. I call it The Human Layer. Human judgment in command, the machine working underneath. The chip can hold the speed. Only a person can hold the judgment. So here are six things I would ask of the people who will actually write these rules. I’m keeping each one short on purpose. One. Ask first, then pay. Before a company trains a machine on your work, it has to ask you. If you say yes, it pays you. Just like in the music business, where you have to pay to use someone else’s intellectual property, you should not be able to train your machines on our intellectual property, whatever that may be, without clearing it and paying for it. A machine should not get to swallow the whole catalog of human work for free. Two. License it, build the locks in, and make the person at the top be accountable for it. I can’t sell a house without a license, so why should something that could potentially destroy humanity be allowed to exist without being regulated? The license should require what Gates is asking for: safeguards built into every model so it refuses to help anyone make a biological weapon or break into the systems we all depend on, and monitoring the government can check for itself. “That has to be a required thing,” he said. Then make the person at the top sign their own name that it all works, with the price on them if they lie. And you can’t regulate a company that won’t tell you its name, so all of this cloak and dagger with companies trying to build data centers using shell companies has to come to a full stop. Three. Owe us a duty, not a promise. In my old world you o… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  4. Sep 17

    Stay in the Room

    On a recent day in September, I had already decided to let something go. For months I had been building toward a certification. The idea was that The Human Layer, my framework for how people should stay in charge of the technology they use, would become a mark that organizations could earn. A stamp. A standard. It was going to be the thing that outlived my working years. Then I ran a clearance search at the trademark office, and I found that another applicant had been required to disclaim the words human layer as descriptive, in the exact category I was aiming at. Underneath that sat a crowded field of similar filings. The phrase had become the industry’s shorthand for the people side of artificial intelligence. It belonged to everyone, which meant it could belong to no one. That landed hard. For about an hour I was close to tears. Another idea that led nowhere. Another thing I had poured a year into that was not going to hold. And then something turned over. The framework was never the protectable thing. CYNERSA was. I made that word up. It is not in the dictionary. I came from the music business, which is why I am an audible learner. I hear language before I read it, and I wanted a name that sounded beautiful when you said it out loud. I wanted the S sound without using the letter S. I tried Synersa. Taken. Sinersa. Taken. Every pretty variation was already someone’s. Then I landed on CYNERSA, and it was open, and it sounded like what I wanted my company to feel like. So the deflation and the momentum arrived in the same hour. I lost the certification I had imagined, and I understood more clearly than before that the one asset nobody else could touch was the word I invented myself. That afternoon I filed to protect it. I was super excited that the name I made up and loved so much was available, something I could name my framework about humans staying in charge in the age of technology, and I could protect that name. The legal side of it is that a trademark application is signed under penalty of perjury, and a false statement can invalidate the registration years later, so every detail had to be true and accurate. When I signed on to the trademark website I wasn’t thinking about prison. I was thinking about my company. CYNERSA is the foundation of the company I am building. The framework, the service, the platform, and the years of income I am constructing so that I am not dependent on a job forever. If that application had been refused, I do not think I would have filed again. I think I would have given up on everything that I was building, kept my day job, and let the whole thing go. That is what was at stake for me. Not a fine. Not prison. A dream, and whether the dream survived the afternoon. So I decided to wipe everything else off the table and laser focus on getting that application filed. My life is very busy and I don’t have the luxury of taking an entire day to do anything. I had recently learned how to let AI take over my screen and do work for me, so I decided to let it help me file this trademark, because the last trademark I filed took days and a call to the trademark office, and I didn’t have that kind of time. So I watched AI read the screens, fill the fields, and move faster than I could have alone. I watched every line of it as if I was entering the data myself. Then I noticed the mistakes, and my heart dropped. First. The address. It typed the wrong street into the owner field. It was a street near mine, and it had come from an error in an earlier session that I had already corrected once. The machine carried a wrong fact forward, and it typed that fact into a federal form with total confidence. No hesitation. No flag. Nothing in its manner was any different from when it was right. Then the next error. The wrong zip code. Legal documents have to be consistent. My Articles of Incorporation carry a zip with the four-digit extension, and I wanted that same extension on the trademark application so my records would line up. But it was actually challenging me that I was wrong and it was right. I was not guessing. My Articles were right in front of me. That blew my mind. Of course it finally changed it, but I was like, what the hell? Third. The title. The form asked for my position. My website says Founder and CEO. My Articles of Incorporation say Founder. I stopped. Before I signed anything at the trademark office, with the help of the same AI, I drafted a written consent appointing me President and Chief Executive Officer of my own corporation, and I signed it in that moment. Which meant, from a legal perspective, the title was true when I typed it. Honestly, that blew my mind too. I didn’t have time for all of these hiccups, but it took minutes, not hours, to fix them all. A corporate governance document, drafted, reviewed, corrected, and executed inside the same afternoon. Work that used to mean a lawyer, a fee, and two weeks or more of waiting took an hour. Somewhere in that hour, I had to take a moment and pause to think about what occurred right in front of my eyes. I was watching a machine complete a federal trademark application on my screen, with me, while I checked its work. People pay thousands of dollars to have this done. I was doing it in an afternoon, in my house, for the cost of the government fee. It felt very sci-fi to me, and I was in awe that science had progressed to this extent in my lifetime. I grew up on Star Trek. I had the pajamas. I watched people on that show use tech that looked like what turned out to be flip phones, and I lived long enough to watch that communicator become an ordinary object in my hand years ago. I have been present for every technological turn of my adult life, and this still had me feeling like a kid in a candy store. It was the machine and me, working, and it was crazy! But in the same breath, it was the clearest argument I have ever experienced for staying present. Every one of those errors was invisible to the system that made the error. Not one would have been caught by asking, unless it was instructed to check its work. They were caught because a person who knew her own documents and her own business was in the chair, guiding the process. If I had walked away and let it run, I would have missed all of it. I would have learned about it months later, in a refusal letter. I know me. I would have thrown in the towel on AI and written something bitter about a technology that had been oversold. Instead, the receipt from the trademark office came back with a serial number on it, and what I felt was relief and amazement at the same time. If you are capable at what you do, and privately unnerved by this technology, I understand it. The hesitation you feel is not unwarranted. It is the same instinct that made me read every line. Here is the whole lesson, and it is one sentence. Stay in the room. Not to supervise the reasoning. You cannot see inside AI, and it will not tell you when it is unsure, because unless you present a challenge it works from a perspective that it’s accurate. Stay because you are the one who knows the facts. Your documents. Your title. Your business. The machine has no way to feel the difference between a true sentence and a false one. You do. And that turns out to be the entire job. I came to artificial intelligence out of fear that it was going to take the work I spent decades learning how to do. What I found on the other side was not a technology that replaced me. I found an informed partner that works alongside me and is lightning fast at accessing and processing information that never would have been available to me at the scale it is available to artificial intelligence. Stay in the room. Read every line. Then sign your name. A note before I go. Staying in the room is the small version of something much larger, and I want to say the large version plainly, because I have not heard enough people say it out loud. I believe that AI, handled the way it is being handled right now, is an existential threat… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  5. Aug 27

    No Human in the Room

    My friend called me a few days ago and told me Suno is adding a watermark to everything created on its platform. If you have never heard of Suno, it is a website where you can make a finished song without a studio, a band, or a single instrument. Even if you are a novice to music, you give it your idea, and it will create a full song! If you are a seasoned musician, it will take everything you direct it to do and create what feels like magic. You can give it lyrics, musical direction, the style, the mood, the tempo, the kind of voice you want, and a few minutes later you have a professionally produced composition that incorporates all of your experience and expertise combined with the machine’s training. Millions of people use it. For someone like me, who has written songs for decades without a band standing by, I initially viewed it as a door that would lead to the world finally hearing my songs in the way I always imagined them. But, like most things in life, there is a catch and a price to pay. And the price just went up again. The AI police have come for the music. Suno announced it this month, in the middle of its legal battles with the major labels: invisible audio watermarking and fingerprinting, rolling out in the coming weeks, built so platforms can detect and flag anything the machine touched. You won’t hear it. The machines will. The mark can say a machine was involved, but it does not point to what the human contributed. It is the same blindness I wrote about when the AI detectors started calling writers who use AI guilty until proven human. When one of my songs gets flagged and labeled, the listener learns that AI touched it. They will never learn about the lyrics I wrote, the chord progressions I chose, the pocket I demanded, or the nuances that make a song connect to listeners, nuances I earned over decades in music studios as an artist and a songwriter. The machine’s fingerprints get detected. The human’s fingerprints get erased. People see the glamour of the music business, but few have the slightest clue of the sacrifices and grueling hard work it takes to hone this craft, and I will not stand by while the shortsighted people building these detectors and watermarks erase what takes musicians years to achieve. People are building answers to the AI police, and I respect the effort. The Authors Guild offers a Human Authored certification for books. It draws its line in one place. A human must write the words. It lets you use AI to research and to brainstorm. What it will not allow is a machine helping to write the words. There is even a service called Verify My Writing that gives you a human seal by running your work through AI detection software and then giving it a score. Here is how well that works. Ron Charles, the book critic formerly with The Washington Post, took a phone interview he had done, fed it to ChatGPT, and told it to write a five-hundred-word story in his voice. Every word of it came from the machine. He submitted it to that service and paid ten dollars. Ninety seconds later he was holding an authenticity certificate that read one hundred percent human based. Those certifications vouch that a human wrote the words with no machine helping to shape them. They have no category for work a human created while a machine helped render it. That purity badge has never been my story, and if you’re honest, it is probably not yours either. There was a machine in my room. I invited it in, I directed it, I overruled it, and I made every final call. What I need is a mark that tells the truth about that. Proof of authorship: what I originated, what I created and what the machine rendered based on my direction. You can’t understand what this watermark erases without knowing what I put into it so let me share my backstory. I have been writing songs for thirty years. I was in an R&B group, Vybe, signed to Island Records. I was in another group signed to Giant Records. People who know my work have compared me to songwriters whose names I won’t repeat here, because those women earned their legends and I have not achieved mine, yet. What I never had, in thirty years, was the team. The singer. I can sing but not every song I write calls for my voice and tones. The keyboard player. The drummer. The producer. I had moments with one or the other, but there were always too many cooks in the kitchen, and there were only moments when my full voice was realized. For a while I had a writing partner who was a master of melodies, and when that friendship fell apart, I stopped writing for months, because my confidence walked out the door with her. I eventually picked up the pen again but was still missing the team I needed for my work to fully take shape. Then Suno came along, and I’m just going to say it the way I felt it: it was like I asked the Lord for something and he hand-delivered it. The keyboard player, the producer, the drummer, and a world-class singer who could take my lyrics and melody ideas and bring them to life exactly the way I heard them in my head. All the hundreds of songs I never finished were finally going to be finished and heard. I don’t read music, but decades in studios teach your ears things schools can’t. I gave it my lyrics. My chord progressions and beats per minute. My instructions for how the singer’s vocal runs should move in and out of the pocket. In one week, I wrote six songs. I played them for industry veterans, people who have heard everything, and they told me every single one was a great song. Six for six, across genres, including a country song I wrote decades ago and never had the right connections in the country music genre to do anything with. Then I read Suno’s Terms and Conditions, and the proverbial record scratched off of the turntable. On paper, as a paid subscriber, I own what I make. But I give Suno a worldwide, royalty-free, sublicensable license to my creations and my prompts, meaning they can use my work, and license it to others, without ever paying me a dime or crediting me for my contribution to the work. The US Copyright Office will not register an AI-generated composition. My lyrics I can register and protect, because I wrote them. The produced sound that I directed and that makes it feel like my finished song I cannot own. Ownership I cannot enforce, plus a license given back to Suno, is ownership in name only. In plain English: it means squat. These were my melodies. My progressions. My hard-earned years of instructions. And I’ll be damned if I was going to give that away. I slammed on the brakes, heartbroken, and I have not contributed the entire body of my experience on that platform since. I did find another use case, which is background music for my marketing materials, and that part is very cool, because I know exactly the sound I want and now, I can create it myself. Suno also raised the stakes for singers by adding another feature that lets you sing on it with your own voice. I read those Terms and Conditions too and went on Instagram with the terms on screen and warned every serious singer who was contemplating using it not to put their voice in Suno. So far that comment has gotten well over 400 likes, plus responses from grateful singers who feel like they dodged a huge bullet by opting not to put their voice on Suno’s platform. When a watermark arrives, it marks the machine’s part and stays silent on the human contribution. Everything I just described, the experience, the hard work, the sacrifices, the six songs, the ears, disappears behind one label that says AI was here. I already built the answer to the AI police for my writing. It’s called The CYNERSA Record™, and it sits at the end of everything I publish, stating what I originated, what I rewrote, what I overruled, and what the AI did, with the full itemized record behind it. It has been running live on Substack for a few weeks. The format is free for anyone to use with attribution (just say where you got it from), and the whole explanation … This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  6. Aug 17

    What Working With AI Actually Looks Like

    I call AI my cognitive prosthetic. When I can’t find the word I want, or I need to confirm a fact before I stand behind it, I put the prosthetic on. It helps me find the language. It helps me check the claim. It helps me shape what I’m already thinking so it lands the way I mean it. Then I take it off, because the thinking is mine. Thanks for reading! Subscribe for free to receive new posts and support my work. Let me show you what that actually looks like across one ordinary day, because I’ve noticed the people writing about AI rarely describe the day itself. At work, when I’m juggling three or four deals and somebody needs an answer now, I write the email myself first. Then I run it through Copilot to sharpen it. Most of the time my first version was fine. But when the clock is tight and the response needs specifics, the prosthetic goes on, the email gets clear and correct in a fraction of the time, and I move to the next thing. When I’m redlining a contract, I rely on my experience and my expertise first, the way I have for thirty years. Every now and then, when I need a piece of legal language unpacked fast, the prosthetic helps there too. The judgment call stays mine. In my life, it’s the first step, never the last word. Before I try a new supplement, I ask about drug interactions, then I confirm with my pharmacist. Before any doctor’s appointment, I walk in fully prepared, with questions and with non-invasive alternatives my doctors don’t always lead with (I’m holistic minded, so my first question is always whether there’s a gentler way). AI surfaces options I wouldn’t have known to ask about. My doctors and I decide. And in my creative work, this is where it feels like magic. I’ve carried ideas in my head for decades. Songs, books, businesses. AI is helping me bring them to life. My best ideas still come in the shower, the way they always have (apparently my head has to clear before the good ones can get through), and I carry those shower ideas back to the work and watch the whole thing rise to another level. The ideas come from sixty-three years of living, and the machine helps me build them faster than I ever could alone. So that’s the honest picture. A Harvard-level assistant on call for my work, my health, my business, my music. Now, I’ll be honest with you, working with AI can be mentally daunting, because it’s still a lot of work. Every one of those moments runs on decisions. All day, hundreds of them. Small ones, fast ones, ones nobody sees. How am I going to verify this? Is this answer right for my situation or just right in general? Do I accept this line or rewrite it? Is this my authentic voice? First verify, then decide, then do something with it. The machine produces. I judge. That judging never stops, and it’s the difference between output and work I’ll put my name on. I learned how much that judging matters on an ordinary afternoon in April, with seventy dollars on the line. I was setting up a seventy dollar ad campaign for my audiobook, with AI walking me through it step by step. At one point it told me to change a setting. For a big company with a big budget it would have been right. Here’s the thing: while you set up an ad, Google shows you a running estimate of what your money will buy. The moment I made the change, my estimate recalculated, and the same seventy dollars went from buying as many as a couple hundred clicks at under a dollar apiece to a few dozen clicks at several dollars each. Same budget, four times the cost, a fraction of the results. So I did my own math. At my budget, following that advice meant the money would be gone before I sold a single book or learned a single thing. The AI looked at my math, admitted the advice was wrong for my situation, and told me to change the setting back. The numbers recovered. The advice was right for a big company and wrong for me, and the only thing standing between those two outcomes was a human doing her own math. My budget, my numbers, my actual situation. The machine had none of that in its head. I did. That is The Human Layer, live, on an ordinary afternoon. Here’s why I’m telling you all this now. The platforms and the AI companies have started building tools that look at our work. Watermarks that mark whatever text a machine touched. Detectors that scan your writing and decide whether a human wrote it. And every one of these tools has the same blindness: they can see the machine’s fingerprints and none of the human’s. There is no tool that sees the email I wrote before the machine sharpened it. No tool that sees the royalty math I ran in my head. No tool that sees the hundred decisions I made today about what to accept, what to rewrite, and what to refuse. Worse than that. The research shows these detectors flag Black writers and writers whose first language isn’t English at higher rates than everyone else. Write with too much polish and the machine decides somebody like you couldn’t have done it. I wrote about that in my Substack piece, “Guilty Until Proven Human,” and I’m still angry about it, because when a detector flags my writing, it isn’t flagging a machine. It’s flagging my stories. My lived experience. Saying AI wrote my work because AI helped me shape it is like saying an editor wrote a journalist’s story. Pulitzer winners have editors. Nobody hands the prize to the editor. So I stopped waiting for someone to build the missing half, and I built it. It’s called The CYNERSA Record™, and it now sits at the end of everything I publish. A short account of what I originated, what I rewrote, what I overruled, and what the AI actually did, with a link to the full itemized record behind it. Real counts, real decisions, taken from the actual work while the work is happening. I set up a rule in my AI’s own preferences that logs every interaction as I write, so the Record comes from the session itself and could not come from anywhere else. I know there are people who ask a machine for an essay, change nothing, and sign their name. Honestly, no detector is needed to spot them. The Record is for the rest of us, the ones whose fingerprints are all over the work: the stories we lived, the lines we rewrote, the advice we refused because we ran the numbers ourselves. Our contribution deserves to be on the record, especially now, because these AI police tools are just getting started, and they are only going to get more sophisticated. I built the Record to get in front of that. If these tools are starting to make you feel invisible inside your own work, you’re welcome to use mine. The format is free with attribution, the full explanation, including how to credit it, lives on my page, and if you can build on it and make it better, I want you to. Because the machine’s contribution already has a watermark. Ours has a Record now. Sources: Anthropic’s watermark announcement and its limits were reported by TechCrunch and TechTimes (August 11, 2026); Anthropic acknowledges the mark shows processing, not authorship. Bias in AI-writing detectors against writers whose first language is not English: Liang et al., GPT Detectors Are Biased Against Non-Native English Writers, Patterns (2023). Black students disproportionately and falsely accused of AI use: Center for Democracy & Technology survey, reported by Education Week (September 2024). The April 2026 ad-campaign account is from my own records. THE CYNERSA RECORD™ This piece began as two voice dictations of mine; the stories are lived and the argument is mine. AI stitched them with a story from my book, verified the research, and drafted from my direction. I made 13 edits and rejected 5 of the machine’s lines for not sounding like me. Full record: https://debrataylorwrites.substack.com/p/the-cynersa-record-what-working-with About the Record: https://debrataylorwrites.substack.com/p/what-is-the-cynersa-record If something here made you think, stay with me. Subscribe free and let’s keep going. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  7. Aug 3

    Guilty Until Proven Human

    Let me back up a second. You are not writing on a 1920s typewriter with a stack of paper and a bottle of white-out. You are using technology, like spellcheck, autocorrect, and the little red line that tells you when you butchered a word. Every one of us writes with tools, and Substack never once put the spellcheck police on the platform. Nobody ever got flagged for using a thesaurus. A new tool shows up, and suddenly we are the problem. I will be honest about my first reaction, because I doubt I am the only one who had it. When I saw this, my stomach dropped. Not anger at first. Sadness. I have loved writing here. I have loved reading other people’s work here. And my very first instinct was to run, to pack up my words and go find somewhere else to write, because AI is now part of my writing process and I did not want to wear a fraud label for it. I did some mental marination to make sure I didn’t let myself get too emotional and give up my new writing home too quickly. Then I got mad because for me fear is always the first stop before anger. But the more I marinated, the more pissed I got! A tool that makes writers want to run is not protecting writing. It is punishing it. Shaming somebody for using AI is like me driving to visit my mom in Illinois and walking in the door bragging. Hey Mom, I made it in ten hours, and I did it with a paper map instead of the navigation! She would tilt her head, give me that look only a mother can give, and say, baby, you could have been here an hour ago if you’d used the GPS. AI does not define the finish line. It just helps you get there faster. Here is how I write, since apparently that is everybody’s business now. I get an idea. I word vomit it onto the page in my own voice, with my own brain, with my own two hands. I shape it myself. Then I run it through AI to check my references, help me clarify things when needed so that my readers clearly understand what I’m trying to communicate and help me make sure I am not passing along uninformed perceptions and misinformation. That is my whole process. AI does not tell me what to think. It does not hand me my opinions. It checks my work, the way spellcheck checks my spelling. I am the writer; it is the tool. End of story. Pangram does not even work, and they know it, because somewhere there is a disclaimer admitting it is not always right. They are running a tool they admit is unreliable and using it to brand people with a label that damages their name and tells their own readers they are frauds. If your machine is wrong even some of the time, even one time, the price of being wrong is somebody’s reputation, and that is despicable to say the least! The more intelligent and well-crafted your writing is the more you get flagged. Write something thoughtful, introspective, well spelled, properly built, and the machine decides no human could have done that. The reward for writing well is an accusation. And guess who gets flagged the most. Study after study shows these detectors flag Black writers, neurodivergent writers, and people who learned English as a second language at higher rates than everyone else. The people who already spend their lives being told that they are not good enough now get a machine that doubts them by default. Write with too much polish and it decides nobody like you could have done it alone. That is not a glitch in the system. That is the oldest bias there is wearing a new software hat. The detection tool is trained on the same bias as the a******s who created it! Guilty until proven human, and in its eyes some of us are guiltier than others before we type a word. If Substack does not want AI anywhere near this platform, fine. Own it. Grow a pair and say it plainly. No AI assisted writing allowed here and stand behind that rule in the open. That would at least be honest. Instead, they hired a machine to police a machine, and they hide behind a disclaimer every time it is wrong. They forgot something simple. People do not read our writing because of what we typed the words with. They read us because they connect to what we write. Nobody in the history of reading has ever loved a piece of writing and then taken it back when they found out the writer used a word processor. Readers care about the idea and the voice. They do not give a crap what we wrote it with. Right now this is small. A scan button on a blogging platform, easy to shrug off. But small and reasonable is exactly how policing gets in. It starts somewhere low stakes, it becomes normal, and one day we stop noticing that a machine is deciding who counts as real and who counts as fake. We get numb to it. Pangram is not the end of that road. It is the on ramp. I am going to take that apart in a piece of its own, because it’s too complex to just write a paragraph in this piece. So, flag me if you want. I have thoughts. I have things to say. I am going to use every tool available to say them clearly, and I will not waste another word or breath on your biased, ineffective and unnecessary AI police. If something here made you think, stay with me. Subscribe free and let's keep going. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  8. Jul 26

    The 700

    A company told the world its AI could do the work of 700 people. A little over a year later it started bringing the people back. This is what I call The Replacement Fallacy, and this is what it looks like out in the real world. There is a story we are all being told right now, and it goes like this. The machine is coming for your job, so you had better make peace with it. It gets repeated so often, by people who sound so certain, that most of us have started to believe some version of it. I understand why. When something is said with enough confidence and enough frequency, it starts to feel like weather. Just a fact about the world that you have to live with. I want to give that story its real name. The Replacement Fallacy. It is the belief that a human being is a cost, and that intelligence is a thing you can lift cleanly out of a person and drop into a machine without losing anything on the way. It sounds reasonable in theory. It falls apart the moment it touches a real customer, a real patient, a real reader, anyone who can tell when there is no one home. You do not have to take my word for any of this. It has already happened in public, with a name and a date attached. The company is Klarna, the big buy now, pay later firm. In December of 2023 they froze hiring. The plan was to shrink the workforce and replace people with AI. I would like to have been a fly on the wall at Klarna to watch exactly how that unfolded in real time. What business problem were they trying to solve, what outcome did they call success, and what methodology did they use to put the decision into motion? In any event, in February of 2024 they made the announcement that got everyone’s attention. Their new AI assistant, built on OpenAI’s technology, was doing the workload of 700 full time customer service agents. It had handled roughly two thirds of all customer chats in its first month. It worked in more than 35 languages, around the clock, and never took a break. On paper the company projected the move would add about 40 million dollars to its profit that year. The CEO, Sebastian Siemiatkowski, became something close to a spokesman for the whole idea. This was the future, and the future needed fewer people in it. Hold that image for a second, because here is the part the original headlines left out. A little over a year later, in the spring of 2025, the same CEO sat down with Bloomberg and said, in plain terms, that they had gone too far. To paraphrase, he said the decision to value saving money over people’s jobs had been weighted too heavily when they chose to replace people with AI. He said the quality had suffered for it. And then he said the line I keep coming back to. There will always be a human if you want. Klarna began investing in human support all over again, bringing people back into the very work it had been so proud to hand off. Investing in the quality of human support, he said, is the way of the future for us. Read that one more time. The human support he now called the way of the future was the very thing they had spent the past year removing. The AI did handle an enormous volume of work. It was fast, it was cheap, it answered at three in the morning without complaint. Every number on the spreadsheet moved in the direction they wanted. If a customer service conversation were only information traveling from one place to another, the machine would have won the argument outright and kept it. A customer service conversation was always more than that. When a person reaches out to a company, they are almost never just asking a question. They are frustrated, or worried about money, or stuck inside a situation that does not fit any of the menu options. What those 700 people carried into every interaction went far past the answer itself. They brought judgment. They brought reassurance. They brought the read of an upset person who needs to be slowed down and handled with a little care. That is the part that never shows up on the spreadsheet, because nobody ever wrote a line item for it. I call it The Human Layer. It is the work a person is doing that you cannot even see until you take them out of the equation and watch what happens. There is a lesson sitting underneath the whole episode. Klarna set out to automate a task, and what they automated away was a whole set of things they had never initially assigned any value to. The person on the other end of the chat had been holding judgment, accountability, and trust inside every exchange. All of that left when the people left, and the company felt the absence of it before it fully understood what it had given up. Klarna is not alone in this, and I am not picking on them. If anything I respect that they said it out loud in both directions, first the boast and then the correction. Around the same stretch, Duolingo announced it was going AI first and walked straight into a wall of backlash so hard that the CEO spent the following weeks clarifying that he never meant AI would replace what his people actually do. The pattern keeps repeating because the fallacy is comfortable to believe. It lets a company tell itself that the people were the expense and the software was the asset. I am telling you this for a reason that has nothing to do with comfort. There is a decision here that we all have to make daily in this new age of AI. Every one of us is now being invited to do exactly what Klarna did, only at our own small and private scale. To take ourselves out of our own work because the machine can produce something faster without us. And every time we do it without thinking it through, we are placing the same bet they placed, that the part of us we just removed was not really doing anything. Klarna made that bet with 700 jobs and a public reputation on the line, and the real world mailed back the receipt. You get to make that same bet a hundred times a week, in what you write and build and decide. You do not have to lose yourself. You can use these tools and remain completely yourself while you do. It does not happen by accident. It happens on purpose, in the small decisions, one after another after another. What Klarna learned is that the 700 souls they let go had been carrying the judgment, the accountability, and in the end the value of the company itself. The people were the secret sauce the whole time. THE CYNERSA RECORD™ The premise, the framing questions, and the closing line are mine. AI researched the Klarna case, verified every figure and quote against published sources, and drafted from my direction. I made 18 edits, rejected 5 of the machine’s suggestions and its first cover image, and banned one AI writing pattern mid-draft. Full record: https://debrataylorwrites.substack.com/p/the-cynersa-record-the-700 About the Record: https://debrataylorwrites.substack.com/p/what-is-the-cynersa-record This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

    The 700
  9. Jul 12

    The Story I Was Handed

    You did not write the story of who you are. Someone else did, for reasons that served them, and they are counting on you never questioning it. I know, because I carried one. Mine was about my own faith. I am African American. My ancestors were enslaved by people who called themselves Christians while they bought and sold human beings, broke families apart, and built their fortunes on stolen lives. For a long time that fact sat in my chest like a stone. Why would I keep the religion of the people who did that to my ancestors? The hypocritical terrorists did not live by the Christian beliefs, they were the furthest thing from the God they claimed to worship. I carried that question for years without examining it. Then one day I noticed something uncomfortable. I spend my days telling people not to outsource their thinking. I tell them not to let a machine decide what is true for them, not to accept the easy answer just because it arrives quickly and sounds confident. The whole of my work rests on one idea. The human being has to stay in the loop. The human being has to do the thinking. And here I was, holding a belief about my own history that I had never actually examined. I had outsourced it. Not to a machine, to something older. I had accepted a story I was handed and never gone looking for the evidence myself. So I went looking. What I found did not match the story. It was stranger, harder, and in the end more freeing than the version I carried. Let me walk you through it, because many of us are carrying the same handed-down version, and we deserve the real one. There is a pattern underneath all of it, and once you see it you cannot unsee it. Those in power build a system first; a system that serves them. Then they wrap that system in a comforting story. The truth, if it shows up at all, shows up last. Hold onto that. It is the thread that ties everything together. The categories were written, not discovered Start with a fact almost no one is taught. Race, the hard and permanent kind that decides your whole life before you draw a breath, was not discovered in nature. It was written into law by men who needed it. In the early years of colonial Virginia, status was not yet locked to skin color. A man of African descent named Anthony Johnson worked his way out of servitude, bought land, and became a respected property owner. A woman named Elizabeth Key, wrongly held in bondage, took her case to the colony’s courts and won her freedom. For a brief window, the lines were not yet fixed. Then the window was closed on purpose. In 1662 the Virginia assembly passed a law with a Latin name, partus sequitur ventrem, meaning that which is born follows the womb. From then on a child inherited the legal condition of the mother. If the mother was enslaved, the child was born enslaved, no matter who the father was. This broke deliberately with English common law, which reserved that rule for livestock. The colony took a principle written for farm animals and applied it to human beings. And because the father was so often the white enslaver himself, the law did two things at once. It erased his responsibility, and it turned the bodies of enslaved women into a renewable source of profit. The next turn came after a fire. In 1676, in an uprising called Bacon’s Rebellion, poor white servants and enslaved Black laborers did the one thing the wealthy planters feared most. They joined together, took up arms, and burned the colonial capital to the ground. The rebellion itself failed. The lesson the elite drew from it did not. They understood that as long as the people at the bottom saw their shared condition clearly, they were dangerous. So the planters made a deliberate choice. They hardened the laws around race, locked Black Virginians into permanent and inheritable slavery, and handed poor whites just enough new privilege to feel superior. Give the poor white man a color to be proud of so he will not notice he is still poor. The hardening was written into the Virginia Slave Codes of 1705. The final layer wore a lab coat. Europe’s most respected thinkers came along afterward and gave the whole structure the dignity of science. In the middle of the 1700s the naturalist Carl Linnaeus sorted humanity into colored varieties and pinned little character descriptions to each one, calling Europeans one thing and Africans another. In 1795 the German scholar Johann Friedrich Blumenbach coined the word Caucasian, naming the supposed white race after a skull he admired from the Caucasus Mountains, and he placed that race at the top of his ranking. The categories millions of people still check on forms today were authored by men with open biases, after the system that needed them was already running. That is the order of events. The economics came first. The law came to lock it in. The science was fabricated to bless it. Truth was nowhere in the room. The truth is this: there is no natural hierarchy of human beings. There never was. Men built one and called it nature. Slavery did not start as a story about color The men who authored the story are dead. The story survives only as long as it keeps getting repeated by the next generation and no one questions it. Here is the part that took the most courage for me to sit with, because it complicates the clean story and because the people still benefiting from what slavery built love to use it to excuse what was done to enslaved Africans in America. Slavery was not invented by Europeans, and for most of human history it had nothing to do with skin color. It existed across the ancient world. The word slave itself comes from Slav, because so many Slavic people of Eastern Europe were captured and sold that their name became the name for the condition. Europeans enslaved other Europeans for centuries. Even during the same years as the Atlantic trade, pirates from the North African coast captured and enslaved somewhere between one million and one and a quarter million Europeans, taking them from coastal towns as far north as Britain, Ireland, and Iceland. Bondage was a human institution long before it became an American racial one. The Atlantic trade itself, the one that took my ancestors, was not something Europeans accomplished alone. They mostly could not push into the African interior. They depended on African suppliers. Powerful kingdoms, the Kongo, the Asante, Dahomey, and Oyo, captured people through war and raids and sold them at the coast in exchange for guns, cloth, and liquor, and the guns fed still more war and more captives. The port of Ouidah became the second largest point of departure in the entire trade. Across its full span, roughly twelve million Africans were carried across the ocean in chains. African hands were on it. But here is the precision that matters. The people doing the selling did not see themselves as betraying fellow Black Africans, because the idea of a shared Black race did not yet exist. They were selling enemies and rivals and the captives of other wars. The notion that all of those different peoples were one kind was not something they believed. It was manufactured later, in the Americas, by the very system that held them captive. The demand, the capital, the ships, and above all the invention of permanent racial slavery were European and American creations. Both of those truths live in the same room, and a serious person learns to hold them at the same time without using one to cancel the other. We are all African under the skin Then I reached the science, and the floor completely gave way. Every human being alive descends from Africa. The human species took shape in Africa around three hundred thousand years ago, and only around fifty to eighty thousand years ago did a small band migrate out and slowly populate the rest of the planet. Everyone who is not African today descends from that handful of travelers. For nearly the entire span of human existence, our ancestors were dark-skinned, because dark skin is what protects the… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  10. Jul 1

    Your Layoffs Are Breaking Your AI

    Watch how a company brings in AI, and you can usually tell whether it will work before a single model goes live. The signal is in the language. They talk about rolling it out. They put it on the IT roadmap, budget for licenses and one training session, block off a Tuesday afternoon to walk everyone through the new tool, and then wait for the productivity to arrive. They are running the same playbook they would run for a new payroll system. That playbook is the reason it fails. The mistake underneath most failed AI projects is easy to recognize once someone calls it out, and companies do it anyway, because it looks like the logical thing to do. They treat artificial intelligence as though it were another piece of software. It isn't, and the difference is the whole story. Software is a finished thing. You buy it, you install it, and it does the same task the same way every time you ask. The people who use it train themselves around it. They learn its buttons, adapt their habits to its logic, and the software never changes based on who is operating it. That predictability is the entire point, and it is what makes the install-and-train approach sensible. You are placing a fixed tool into a workforce. Artificial intelligence works from the other direction. It is not fixed and it is not finished. It learns, and what it learns from is a record of how people already do the work. The data you feed it is not raw material sitting in a warehouse. It is the accumulated judgment of the people who built your business, encoded in the decisions they made, the exceptions they caught, the things they knew to do that were never written in any manual. When a company treats those people as a cost to remove and the AI as their replacement, it is destroying the source the system was meant to learn from, and then wondering why the results come back thin. None of this is anecdotal. The failure rate for AI projects runs above 80%, roughly double the rate for ordinary IT work, and by one widely cited MIT estimate, about 95% of generative AI pilots deliver no measurable return. Sit with that number for a second. Nineteen out of twenty. An entire field of companies pouring money, talent, and executive attention into these projects, and almost none of it comes back. A genuinely broken tool would fail everywhere, every time. This is something else: a technology being handled the wrong way. Look at the companies that do get a return and the same pattern surfaces, most of the work that produces success comes down to people and process, not the technology itself. A company that cuts the people first is cutting the thing that was going to make it work. Nowhere is this playing out faster than in media. The layoffs are already here, framed as the cost of becoming an AI-first company, and the logic sounds airtight in a board deck. Replace the editors, the coordinators, the localization teams, the people who tag the archive and clear the rights, and let the models do the work for a fraction of the cost. But those people were not the expense standing in the way of the technology. They were the ones who knew which archival clip would actually clear, which cut would land, which exception the rulebook never covered. That knowledge was the raw material the AI needed, and it is walking out the door in the same restructuring that was supposed to make the company modern. The industry is not becoming AI-first. It is becoming knowledge-poor, and calling it progress. There is a way through this, and it starts by putting the people back at the center of the work instead of at the edge of the budget. I call it The Human Layer. The idea is straightforward, though almost nobody does it: before you touch the technology, you map the actual workflow, the real one, the way the work truly moves and who holds the judgment at each step. You find where the knowledge lives, the parts that were never written down. You bring the people who own that knowledge into the change instead of routing around them, because they are the ones who can tell you which data matters. Only then does the technology have something true and valuable to learn from. The order is the method. Get it right and AI has a real chance of working. Get it backward, tool first and people last, and you get the 95% failure. The companies that come out of this ahead will be the ones that understand what they are working with, and treat their people as the asset the technology runs on rather than the cost it eliminates. This is a commercial argument before it is anything else. It is the difference between a project that returns something real and one that becomes part of the 95%. The tools will keep getting better every quarter. The technology was never the question. The question is whether the company is ready to lead its people through the change. If something here made you think, stay with me. Subscribe free and let's keep going. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  11. Jun 7

    Lost In the Matrix

    I have been writing songs for decades. Hundreds of them. Many are recorded, hundreds more are registered with BMI, and I have songwriter credits across more than one album on major music labels. This was never a hobby. It was a career. But a finished song is not the same as a song the world gets to hear. So a large chunk of my catalog sat in notebooks and voice memos and on hard drives, finished and waiting. Then a friend told me about Suno. I want to be honest about the feeling, because the feeling is the whole reason this story matters. The first time, I gave SUNO nothing but a chorus and it handed me back something I actually loved, I felt a kind of excitement I had not felt in years. All those songs I had been holding onto, all those years, and finally a way to produce them myself and put them into the world. So I went in. Deep. I should tell you I am not a person who just pushes a button and hopes. I am a professional singer and songwriter. I trained at the Musicians Institute in Hollywood for two years and back in 1995 I sang in an R&B group called Vybe, signed to Island Records, and I co-wrote on our album. My most notable credit came later, a song called “What Goes Around” on a K-Ci & JoJo record, the platinum duo who came out of Jodeci. If you go looking, my older credits read under earlier versions of my name, Debra Mitchell and Debra Mitchell-Adams, before I became Debra Taylor. So, when I say I understand songwriting, I’ve spent a lot of years doing it. Which raises a fair question. If I am a real songwriter, why did I need Suno at all? Here is the honest answer. I write songs, but I do not produce them and I do not read music. For twenty years, turning one of my songs into something you could actually hear meant the same climb every time. I would demo a song a-cappella, find a producer to build the track, and then the two of us would work out what kind of voice the song needed and go find a singer. One song could take months. In twenty years of doing it that way, I got maybe fifty to a hundred of my songs fully produced. And now I live in Atlanta, not Los Angeles, without the network I spent years building, so doing it the all human way today would mean rebuilding those relationships from scratch before I heard a single finished track. Suno collapsed all of it. I gave it the same direction I used to hand a producer, the tempo, the key, the chord progression, the feel. I sang my song into it to guide it. Then I shaped it, prompt after prompt, until it sounded like what I had been hearing in my head for years. Six songs in one week. I was astounded! I played those six songs for a small group of people I trust. Every single one of them fell in love. Every single one of them told me these were hits. And then it hit me. I am not the one singing these songs. That voice, the one people were falling for, came out of Suno. It was beautiful and it was not mine, so how was I supposed to put this into the world? I did not have an artist. So I made one. I called her LIYA and I poured decades of my own songwriting into her. She was not a gimmick. She was the voice for work I had been building my whole career. I released her first song, and people loved her! She is out there right now getting love from listeners, and I am not even promoting her. Then I went back to make more. And I could not get into my account. Sit with that for a second. I had a brand-new artist that people were responding to, real songs, real momentum, and I was locked out of the only place she could sing. For weeks. Then months. I tried everything. I emailed. I messaged the founders directly on LinkedIn. I eventually decided I just wanted a refund, because at least that would be something. Here is what I found when I went looking for help. There was no one there. Suno had handed its entire customer service over to AI. Every message I sent came back as a canned response. No human to troubleshoot with. No one to walk me through anything. No one to say I see what happened, let me help. Just a polite machine, repeating itself. That is the moment my stomach dropped, because that is the exact thing I warn people about. I write about something called, “The Human Layer.” The idea is simple. AI is powerful, but the moment you pull humans out of the places where judgment and care actually matter, the whole thing collapses on the people who depend on it. I talk about this as it relates to enterprises. I believe this is a pivotal issue for companies using AI. And there I was, watching my own AI artist disappear into a company that had removed humans from the one place I needed one. LIYA did not get lost in space. She got lost in the exact gap I warn enterprises about every single day. Now let me be clear about what I actually lost, because it is not what you might think. I did not lose the music. The songs I made while I was paying for the service are mine. I downloaded the stems. I can release them and thankfully I have a few more songs from LIYA that I can follow up with. But what I lost was the means to make more. LIYA’s voice was never something I could pick up and carry out the door. It lived on their platform, tied to an account I cannot reach, and even if I open a brand new one, I cannot rebuild what was on the old one. That is where LIYA lives now. In a room I no longer have the key to. It was never an ownership problem. It was a control problem. And that distinction is the whole lesson. I didn’t feel like I was taking a shortcut at the time. It honestly felt like I had found a tool that could finally bring my songwriting to life. And choosing a voice that was not mine was not a compromise. It was a craft decision. A songwriter picks the voice a song calls for, and this one called for LIYA, her runs and her tone. That is a big part of why people loved it. But somewhere in the middle of it, a quieter instinct kept tapping me on the shoulder. It said, make a version that is fully yours too, something with your own voice on it that no platform could hold hostage. I did not listen. And what I understand now is that the instinct was right but my reasoning was off. Protecting my work was never going to come down to whose voice was on the song. It came down to whether I controlled the ground the song was built on. And I did not. So here is what I keep coming back to. I did not lose LIYA because the songs were not good. People loved them. I lost her because when the tool broke, there was no human on the other side of it. Just a machine answering for a machine, repeating itself while my artist sat locked in a room I could not reach. That is the part every company should sit with, because it is exactly what many of them are racing to do right now. They are handing their entire customer service over to AI. Not using it to help their people. Replacing their people with it. And it works beautifully, right up until something goes wrong that the system was never built to handle. Then the customer is me. Standing outside a locked door. Sending message after message into a void while the thing that was supposed to help hands back one more polite, useless reply. Here is the distinction that matters. AI has a place in customer service. It can answer the easy questions in seconds and free your people for the ones that actually need a person. But the moment you remove humans entirely; you have not made your service efficient. You have made it unreachable. And your customers will remember exactly how that felt, the way I remember it now. So if you take one thing from my expensive little story, let it be this. Keep a human in the room. Build on ground you control. And never hand the whole thing over to a machine that will never grasp the nuances involved in superior customer service for humans. That is what it cost me to learn. My hope is that it costs you a lot less. If something here made you think, stay with me. Subscribe free and let's keep going. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

  12. May 24

    I Came to AI in Fear

    I didn’t come to artificial intelligence with curiosity. I came to it with fear. In early 2024, I started watching dozens of podcasts and taking online courses with the mission of learning everything I could about AI. And somewhere in all of them ran the same thread. People’s jobs, especially positions like mine, paralegals, contract managers, were going to be among the first to be replaced by AI. Not might be. Would be. And in a very short period of time, two to three years from 2024. The fear was immediate, and it was specific. What I was afraid of losing wasn’t abstract. It was my livelihood and my purpose, the two things that, taken together, make me feel like myself. I need to build, to contribute, to make something that matters. Take away my ability to earn and the sense that my work counts, and I wasn’t sure what would be left of me. But I don’t let fear sit there as a headline. When something frightens me, my first instinct is to understand it, then to employ critical thinking to position myself appropriately. I took a deep dive into AI, and the more I worked with it, the more I understood the reports weren’t wrong. This wasn’t only about paralegals and contract managers. It was something that could reach almost everyone, including CEOs and founders. Not someday. Soon. Here’s the part I hadn’t told many people. I had been here before. In the 1990s I was chasing a different life, a singer/songwriter signed to two major labels, certain I was on my way. Then the internet arrived and the music business changed underneath me almost overnight. The money dried up and as a singer in the R&B group Vybe, signed to Island Records, with two Billboard-charting singles, one that reached the Hot 100 my career stalled. I didn’t understand the size of what was happening until it was too late to get in front of it. By the time I saw it clearly, it had taken everything. When I heard the same kind of prediction about AI, something in me went cold. I knew that feeling. I had lived the cost of not seeing a shift in time. Except this time was harder. Because this time, my mind wasn’t the same. In 2023, I was working as a realtor again, rebuilding. It was a Sunday. I was out front of a subdivision trying to push an open house sign into the ground, something I’d done a thousand times, and it just wouldn’t go in. My husband sat in his truck behind me, watching. Then my life changed in the blink of an eye. A nineteen-year-old, going about sixty miles an hour, was racing another car and lost control. My husband saw it coming and couldn’t move fast enough to stop it from hitting me. The car hit me and threw me so far that my husband couldn’t see where I’d landed. Broken arm in two places. Broken leg in two places. A brain bleed. I woke up in the hospital with no idea where I was or why. When I got home seven days later, I knew something was wrong. I’d be speaking and suddenly the word I wanted wasn’t there. I stuttered. My sentences came out garbled. The fatigue hit like a wave every afternoon. My thoughts wouldn’t organize the way they always had. Finding a word felt like reaching into a dark drawer and retrieving nothing. The neurologists ran scan after scan and kept telling me the same thing. Your brain looks fine. For your age, the scans are fine. But I wasn’t fine. The scans didn’t show what I was living. I did the only thing I’ve ever known how to do when something tries to take me down. I refused it. I’d walked two to five miles a day before the accident. Afterward I could barely get from the bed to the bathroom. I started with a walker, moved to a cane, used it for three months, and then one morning I reached for it and decided I was done. I put it in the garage. And I walked. It was excruciating, but I did it anyway, because walking had always been my joy and my escape. I wasn’t going to let anything take that away from me. Eventually, I walked again without pain or a limp. It took two years of self-talk and dogged determination to walk again on my own terms. No lingering physical signs from the accident. Not on my watch. That same refusal is what I brought to the idea of being replaced by AI. Not on my watch. I started using AI in 2024, about a year after the accident. At first just questions. My friends always called me the question lady. But the questions got bigger, and somewhere in there AI stopped being a curiosity and became something I hadn’t expected. It became the bridge between the thoughts in my head and the words I’d lost. For the first two years after the accident, even though the neurologist kept telling me my brain was healed, I was still living with cognitive difficulties. Nothing anyone else could observe, but I knew my mind wasn’t what it had been. Ordinary conversations didn’t flow with ease. I couldn’t always find my words or communicate my thoughts the way I meant to. I struggled to hold onto details, especially long-term memories. And then something unexpected happened. The thing I had approached with fear evolved into curiosity, then discovery, then marvel. When I couldn’t find a word, it helped me find it. When I couldn’t organize what I was thinking, it helped me hold the shape of it. When I couldn’t reconstruct what I’d been trying to say, it helped me build it back. I spent my days translating what was in my head into something I could understand clearly again, and then into something valuable. I began to learn, on a deeper level, about things I’d always been curious about but never had the time to explore. I began to take ideas I’d carried for decades and turn them into real intellectual property. Songs, business concepts, and books. It gave me back not only what the accident had taken, but something more. Access. Access to information, processes, and resources I’d never had before. And out of that, I built something. I took thirty years of contracts, dealmaking, and negotiation, work I’d done as well as anyone in the room and been told I could only execute, never lead, and turned it into a company. Built on top of AI, in a way that meant I could never be erased again. That was the moment the fear changed shape, because I finally understood what was actually happening to me. AI didn’t replace me. It amplified me. And it gave me back what the accident could have taken, and more. AI didn’t heal my brain; that happened naturally, over time. But it amplified what I already had, and became my business partner in a way I never imagined. That distinction is the whole thing. And it’s the thing almost everyone is getting wrong right now. In 2023, the people building this technology signed a public statement warning that AI could end us, naming extinction alongside pandemics and nuclear war. Then they went back to building. They named the risk, then went back to monetizing it and making it even more powerful. They never named who leads through it. That’s the gap I live in. While we argue about whether AI ends the world in 2050, organizations are mishandling it today. They cut their most experienced people, hand the work to a model, discover the model can’t do what those people did, then either rehire at a premium or live with the damage. I call that The Replacement Fallacy™: treating AI as a tool to shrink the workforce instead of transform the work. What they’re all missing is that the judgment was never in the process. It was in the person. And that judgment, which is a product of our humanity, should never be transferred to a machine. The alternative is the thing I lived. Human judgment kept in command, with AI as the workforce underneath. I call it The Human Layer. The tool can draft, generate, accelerate. But someone still has to be the critical thinker behind the machine. We have to direct, correct, manage, and decide how and when to deploy the outcome of every phase of our interaction with AI. The builders told us the end is possible. I’m telling you the middle is here. The middle is where we decide whether AI amplifies what’s human in us or erases it. I came to… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit debrataylorwrites.substack.com

About

The AI conversation happens in a room most of us were never invited into. Debra Taylor, author of The Human Layer, reads her essays for everyone standing outside it. We can go further with this thing than anyone imagined, as long as a human stays in charge of the direction. Stay in the room with me. debrataylorwrites.substack.com