Manual Work is a Bug

Matthew Black and Hillary Black | Mav

We believe that Manual Work is a Bug. This is a podcast about the future of work and life with artificial intelligence. Hosted by Matthew Black and Hillary Black. A Mav Automation Ventures production.

  1. 2d ago

    Stop Using AI Like a Search Engine: MCPs & Connectors

    Mav Founder + CEO Matthew Black and Head of Brand Hillary Black are back at the same table, and they brought something to launch: Mav's official MCP server. Connect Mav to Claude or ChatGPT On the table: a company arcade with a global leaderboard (does this make us serious, or unserious?), why your AI is only as smart as what you connect it to, the personal agent that wants your credit card, a local model test that surprised Matthew, and the one prediction that could hand Meta the ads business, and get rid of the dreaded Meta Business Suite once and for all. Chapters 00:00 — Intro: back in the studio01:13 — The Mav arcade02:44 — How serious are we, really?04:39 — Use AI to do more, not less05:39 — Why taste matters now07:45 — Mav's MCP server is live08:42 — Asking Mav anything09:40 — Why we opened it up11:15 — Ten connectors, one dashboard12:11 — Treat AI like the gym12:55 — Otherwise, it's just a search engine13:36 — Static documents vs. live connections15:22 — OpenClaw and the personal agent race16:01 — Muse wants everything connected16:35 — From read-only to write17:23 — Can local models do this?18:18 — Consumer vs. business: who picked a lane19:35 — Matthew's favorite connector21:09 — Rippling and event ROI21:31 — Hillary's picks: Figma and Notion23:01 — Will Claude build what its connectors do?24:48 — Zapier, the original connector25:13 — Meta's opening in ads26:47 — Google, Microsoft and Apple27:42 — Is this interesting to anyone?28:44 — Episode 15 and the private jet Hosts Matthew Black — Founder + CEO, Mav Hillary Black — Head of Brand and Conversation Design, Mav Links and Resources The Mav arcade, home of Dialer Destroyer and the global leaderboard Matthew's earlier explainer on MCPs, the video Hillary mentions at 11:15 Meta Muse, the personal agent that only works once you connect everything Claude Duet, Anthropic's answer to Muse, announced the day this was recorded Granola, the meeting notes app behind Matthew's favorite connector Rippling, whose MCP gives Mav ROI on event spend Figma, Hillary's live asset library inside Claude Notion, where every transcript and testimonial lives Canva, whose Claude connector Hillary used to update every brand template Zapier, the original connector Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com. Spotify Amazon Music

    Stop Using AI Like a Search Engine: MCPs & Connectors

  2. Oct 22

    I Trust My AI More Than I Trust Myself with Charley Todd

    This week, it's Matthew's turn for a one-off. Mav Founder + CEO Matthew Black sits down with guest host Charley Todd, founder of Charley Insurance and friend of the show, for a conversation filmed three months ago that already feels like a time capsule. On the table: the AI model that vanished overnight, why Matthew trusts his model more than he trusts himself, the airplane that explains how he runs Mav, and the old-school answer he gives for surviving a founder's worst days. Chapters 00:00 — Intro: a three-month-old conversation01:27 — Matt, Matthew, or Matty?02:44 — Adopting AI as a new independent agency04:27 — Humans and technology: the next 18 months05:12 — Everyone is struggling for air05:56 — Loop engineering06:43 — The Fable that disappeared07:37 — It feels like the dot-com boom08:11 — Operate with AI every day10:01 — What consumers want: utility and context11:27 — Insurance inside GPT and Claude12:02 — The renewal letter nobody wants13:15 — What Siri should have done14:53 — I trust my model more than myself16:20 — Why Mav switched to Rippling17:13 — Where the human touch still wins18:28 — Learning vs. building19:21 — Learn by solving real problems21:24 — Boeing vs. Airbus: how Matthew runs Mav22:51 — Give your model context23:40 — Start with weekend vibe coding24:20 — Nine problems and one gem25:20 — The hardest day as a founder27:29 — Go in with a hundred percent love30:02 — "Clip it. Publish it."32:21 — Expanding, not contracting33:46 — Close Hosts Matthew Black — Founder + CEO, Mav Charley Todd — Guest host, insurance agency founder, Charley Insurance Links and Resources Anthropic's statement on Claude Fable access, the model that disappeared mid-session Rippling, the HR and finance platform Mav moved to for its MCP Jack Dorsey, "From Hierarchy to Intelligence," on running Block through AI instead of org charts Hacker News, where Matthew watches what's coming down the pipe Obsidian, the note app Charley uses for his master documents Art of Accomplishment, Joe Hudson's coaching company, home of "How could I enjoy this 10% more?" Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com. Spotify Amazon Music

    I Trust My AI More Than I Trust Myself with Charley Todd

  3. Oct 15

    How to Send Texts Your Customers Will Reply To: Conversation Design 101

    This week, it's a solo episode. Mav Head of Brand and Conversation Design Hillary Black takes the mic to share what eight years of writing for bots has taught her about one thing: getting people to actually text you back. On the table: what a conversation designer really does, why ChatGPT made your customers harder to impress, the texting mistake she sees brands make constantly, and her top three fixes you can use today, no matter which AI you run. Chapters 00:00 — Intro 00:15 — A solo episode 01:21 — What is conversation design? 02:05 — Why the good bots are good 02:43 — From writing bot scripts to writing system prompts 03:48 — Why conversation design matters more than ever 05:40 — When AI support chat goes off the rails 06:46 — If you're texting your leads 08:02 — Top three ways to improve your conversations 08:31 — Tip 1: Make it a two-way conversation 09:56 — Tip 2: Ask one thing per message 11:24 — Tell people how to respond 12:22 — Close the loop 13:23 — Tip 3: Test it yourself as a user 14:57 — What to look for in your conversation logs 16:07 — Close Host Hillary Black — Head of Brand and Conversation Design, Mav Links and Resources ChatGPT, OpenAI's assistant and the reason your customers expect more from every chat Claude, Anthropic's assistant, home of the Opus and Sonnet models Hillary mentions SmarterChild, the AIM chatbot that couldn't answer much of anything Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com. Spotify Amazon Music

  4. Oct 8

    How Insurance Agencies Are Using AI, Live From ITC Vegas 2026

    This week we have a special episode - Live from the ITC 2026 expo floor. Mav Founder + CEO Matthew Black and Head of Brand Hillary Black grab ten minutes in the official podcast studio between booth shifts to report on what actually changed in insurance this year. Every vendor is still pitching their version of AI for insurance. But what’s changed this year is the agency owners. On the table: why GPT helped the first five minutes of every sales conversation finally disappear for companies in AI, the commercial agency strategy for quickly moving into personal lines without hiring, evidence that Mav has the coolest booth in the tradeshow circuit, and Matthew's prediction for where the insurance product itself is headed in the next year. Chapters 00:00 — Intro 00:16 — Live from the ITC 2026 expo floor 01:17 — The coolest booth on the circuit 02:45 — "X for AI," three years later 03:40 — The agency owner has changed 05:07 — The old guys in the back 06:19 — From "I don't text" to "how do I grow?" 07:04 — Commercial agencies moving into personal lines 08:28 — A prediction for next year 09:37 — Why doesn't insurance just come with the car? 10:05 — Insurance like shipping on Amazon 10:44 — Close: see you at ITC 2027 Hosts Matthew Black — Founder + CEO, Mav Hillary Black — Head of Brand and Conversation Design, Mav Links and Resources ITC Vegas (InsureTech Connect), the world's largest insurance innovation event Toyota Insurance, the automaker-owned agency selling coverage alongside the car Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com. Spotify Amazon Music

  5. Oct 2

    How We Rebuilt Our Entire Website in Two Weeks with Claude Code (and Actually Made it Look Good)

    This week we're walking through the entire process of rebuilding and redesigning the ⁠hiremav.com⁠ website using Claude Code, the Figma MCP, and two humans. ⁠Mav⁠ Founder + CEO ⁠Matthew Black⁠ and Head of Brand ⁠Hillary Black⁠ walk through the exact process from ideas to setup, iterating, and launching, plus where humans stepped in to make it not look like every other AI-generated website. We think it's looks pretty cool. Hillary, a non-engineer, was able to build the website of her dreams the exact way she wanted without waiting weeks to fit it into a sprint cycle, and the Claude Code "other tab" even convinced Matthew to use the Claude App instead of terminal for his work. On the table: why buying a hundred old insurance agencies and trying to AI-enable them doesn't work, the one rule that kept the site from looking AI generated, the tells that give an AI website away, a dev setup that went from four hours to thirty minutes, whether a website should even be the same for every visitor, and an Easter egg worth a free year of Mav. 🥚 There's an Easter egg hidden on the new ⁠hiremav.com⁠. Find it, screenshot it, and DM Matthew on LinkedIn. First one gets a free year of Mav. Chapters ⁠00:00⁠ — Intro⁠00:15⁠ — Cold open: we're live⁠01:01⁠ — The TV nobody could find⁠04:21⁠ — Too many humans in the way⁠04:49⁠ — Why the AI roll-up doesn't work⁠06:29⁠ — Buying the business for parts⁠07:11⁠ — Today: replacing ⁠hiremav.com⁠⁠08:08⁠ — Twenty new pages, two people⁠09:35⁠ — No staging server, no CMS⁠10:45⁠ — AI started, human finished⁠11:28⁠ — "I'm not a designer"⁠13:04⁠ — Was the wireframe holding it back?⁠14:17⁠ — The tells of an AI website⁠15:43⁠ — A four-hour setup in thirty minutes⁠16:57⁠ — Claude Code in the app⁠18:28⁠ — Every page, tracked⁠20:42⁠ — What even is a website anymore?⁠22:01⁠ — The new AI use case: create constraints⁠25:08⁠ — Using humans where it counts⁠26:01⁠ — One button and it's live⁠27:06⁠ — The Easter egg Hosts ⁠Matthew Black⁠ — Founder + CEO, Mav ⁠Hillary Black⁠ — Head of Brand and Conversation Design, Mav Links and Resources The new ⁠hiremav.com⁠ ⁠Claude Code⁠, what Matthew pointed at the README and what built the pages ⁠Claude Code and Figma⁠: setting up the MCP server, the connection that let Claude read the actual design file ⁠Guide to the Figma MCP server⁠ ⁠Model Context Protocol⁠, the open standard behind the Figma connection ⁠ngrok⁠, how Hillary's local build gets shared with anyone ⁠Docker⁠, the self-contained dev environments ⁠micro1⁠, the AI training-data company Matthew mentions Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: ⁠hiremav.com⁠. ⁠Spotify⁠ ⁠Amazon Music⁠

    How We Rebuilt Our Entire Website in Two Weeks with Claude Code (and Actually Made it Look Good)

  6. Sep 25

    Where AI Outperforms Humans, and What the Best Use Cases Have in Common

    After two weeks of AI getting it wrong, let's talk about where AI gets it right. Founder + CEO Matthew Black and Head of Brand Hillary Black go looking for the places AI genuinely outperforms a human, and what those use cases have in common.It starts with a car that drives itself better every month, and an insurance company that charges you less the more you don't do your own driving.We talk: why Tesla's lead in autonomous driving may be impossible to catch, the difference between selling tokens and selling outcomes, why everyone's AI flyers look the same, a data broker's offer that stopped Matthew cold, why CRMs are graveyards, and his MCP setup that replaced a personal trainer. Chapters 00:00 — Intro 00:15 — Cold open: ten episodes in 01:10 — Starlink at cruising altitude 03:40 — The chairs: a final update 05:04 — When nobody can override anything 07:01 — Tesla self-driving, six years later 08:24 — Insurance that pays you to let the AI drive 09:46 — Why Tesla wins: the data nobody else has 10:54 — Sell outcomes, not tokens 12:22 — Why all the flyers look identical 13:29 — Data is being brokered like oil 14:04 — You can't be an AI wrapper anymore 16:37 — Tesla isn't a car company 17:57 — The company that can't use its own data 19:53 — Choosing tools by what they connect to 21:03 — Why do you even need a CRM? 21:38 — No steering wheel, no pedals 22:48 — What agencies can do at a smaller scale 24:35 — The fitness stack that replaced a trainer 27:16 — The bridge is hard 29:40 — Why Mav went all the way into insurance Hosts Matthew Black — Founder + CEO, Mav Hillary Black — Head of Brand and Conversation Design, Mav Links and Resources Tesla Insurance: how the Full Self-Driving discount works Lemonade's Autonomous Car insurance, roughly 50% off every mile driven on FSD Tesla, Uber and Waymo all get the OK to run thousands of robotaxis in Nevada Model Context Protocol, the open standard behind connecting your tools to Claude athletedata, a single MCP server that connects WHOOP, Hevy, MyFitnessPal and 30 more fitness apps to Claude Rippling, the not-a-sponsor Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI.

    Where AI Outperforms Humans, and What the Best Use Cases Have in Common

  7. Sep 18

    Why AI Customer Service Keeps Failing and What It Might Be Costing Your Business

    We're an AI company, but we won't sell you AI for customer service (yet). Here's why we're so bothered by it. Mav Founder + CEO Matthew Black and Head of Brand Hillary Black get into why the thing everyone is automating first for some reason, is also the thing that breaks most often, and potentially costs your business the most. Hillary is bothered. Three sets of chairs, a retailer with no phone number, and a bot that can talk all day without doing a single thing about it. On the table: the one company Matthew thinks actually gets this right, the two conditions that separate a support experience that works for customers, from one that wastes an hour of your life, the unit economics nobody considers before they switch off their phone lines, what happens when your agent starts talking directly to the store's agent, and a mall built entirely for AI. Chapters 00:34 — Cold open: three sets of chairs and no phone number02:35 — Why some AI support is good and most of it isn't03:21 — Why Mav stayed out of customer service on purpose04:10 — Sales is a formula. Support is one-out-of-one.06:34 — 27 entry points vs. 3, and the unit economics nobody runs08:43 — Cheaper per conversation, more expensive per customer09:01 — The one that actually worked10:50 — The two conditions12:06 — Onboarding a rep with no access to inventory12:30 — Saying the right thing doesn't matter if it can't do the job14:30 — Muse, and what happens when agents talk to agents16:24 — Meta bought the agent forum18:13 — MoltMall20:01 — Everyone is doing the best they can22:19 — The fold phone detour23:52 — Go win somewhere easier first24:25 — The worst possible place to put AI in insurance26:16 — Reinvest the savings at the bottom of the funnel too27:37 — Why Chewy gets away with it Hosts Matthew Black — Founder + CEO, Mav Hillary Black — Head of Brand and Conversation Design, Mav Links and Resources Meta's Muse — the personal agent launched September 8, described as OpenClaw for everyone OpenClaw — the open-source personal agent Muse is answeringMeta acquired Moltbook, the social network built for AI agents Model Context Protocol — the open standard behind agent-to-agent connection Stripe Support — the implementation Matthew holds up as the bar Chewy — Matthew's example of stellar customer service Follow the show Manual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com.

    Why AI Customer Service Keeps Failing and What It Might Be Costing Your Business

  8. Sep 11

    How People Actually Shop for Insurance With AI & How Agencies Can Show Up

    Your customer is about to ask an AI whether they're overpaying you. Mav Founder + CEO Matthew Black and Head of Brand Hillary Black get into what happens when people end up shopping for insurance when all they did was ask AI to fix their budget.On the table: the prop plane Matthew almost bought, why comparison sites can't help you with this, what Bitcoin taught him about how consumers trust, how a small agencies compete with big budgets, how to get your agency recommended by ChatGPT, what insurance will look like a decade from now, and a million dollar idea to win over your clients he gives away at the end.Hosts- Matthew Black — Founder + CEO, Mav- Hillary Black — Head of Brand and Conversation Design, MavLinks and Resources- The Insurance Buzz with Michael Weaver and Courtney Weaver — the agency-side version of this conversation, with Matthew- Insurify's ChatGPT app — real-time personalized auto quotes inside the chat, the execution Matthew points to- How Insurify's app started, back in February — rate estimates and side-by-side comparisons from a database of 196M quotes- The insurer ChatGPT apps directory — who else is already quoting inside an assistant- Why quoting inside the assistant is the whole game, including Liberty Mutual making the first move among US carriers- Model Context Protocol — the open standard that makes any of this connectable- Tesla Insurance — insurance as a line item in the purchase wizardFollow the showManual Work is a Bug is Mav's podcast, now on video. Watch on YouTube and follow wherever you listen, so you don't wait another six years for the next episode. Mav helps insurance agencies grow their book with AI: hiremav.com.

    How People Actually Shop for Insurance With AI & How Agencies Can Show Up
See All (20)

Trailer

Ratings & Reviews

5
out of 5
4 Ratings

About

We believe that Manual Work is a Bug. This is a podcast about the future of work and life with artificial intelligence. Hosted by Matthew Black and Hillary Black. A Mav Automation Ventures production.

Information

  • Creator
    Matthew Black and Hillary Black | Mav
  • Years Active
    2020 - 2026
  • Episodes
    20
  • Rating
    Explicit
  • Copyright
    © 106556
  • Show Website
    Manual Work is a Bug