Earley AI Podcast

Seth Earley

In this podcast hosts Seth Earley invites a broad array of thought leaders and practitioners to talk about what's possible in artificial intelligence as well as what is practical in the space as we move toward a world where AI is embedded in all aspects of our personal and professional lives. They explore what's emerging in technology, data science, and enterprise applications for artificial intelligence and machine learning and how to get from early-stage AI projects to fully mature applications. Seth is founder & CEO of Earley Information Science and the award-winning author of "The AI Powered Enterprise." 

  1. 4d ago

    Earley AI Podcast - Episode 95: Contract Intelligence, Context Engineering, and Building AI That Scales with Deepak Bapat

    Why Making Complex Revenue Simple at Scale Requires More Than Throwing Contracts Into a Chat Interface Guest: Deepak Bapat, Co-Founder and CTO at Tabs Host: Seth Earley, CEO at Earley Information Science Published on: July 30, 2026 In this episode, Seth Earley speaks with Deepak Bapat, Co-Founder and CTO at Tabs, a revenue and accounts receivable management platform built for B2B companies. They explore why dropping contracts into a general-purpose AI tool is not a strategy for enterprise scale, what generative AI unlocked that OCR and legacy machine learning could never solve, why context engineering beat fine-tuning for contract extraction, and why newer and larger models are not always better for specialized tasks. Deepak shares candid and specific insights on building atomic AI pipelines, the provability requirement that financial compliance demands, and what finance and data leaders consistently underestimate before deploying AI on their contracts. Key Takeaways: Dropping contracts into a chat interface is a reasonable experiment but not an enterprise strategy - doing things at scale requires specific tooling, specific expertise, and integration across systems.The SaaSpocalypse framing misses the point - the more interesting question is not whether chat replaces UI, but how platforms can understand intent and preempt the actions users would otherwise have to click through manually.Generative AI solved the contract problem by reasoning over ambiguous natural language at document level - something OCR and rules-based systems fundamentally could not do.Context engineering beat fine-tuning at Tabs because merchant preferences vary so significantly that fine-tuning per merchant became cost-prohibitive - a well-prompted generalized model proved faster and more elastic.Newer and larger models are not always better for specialized tasks - Deepak's eval sets show that models from six months ago outperform newer versions on certain contract extraction jobs, likely due to overfitting on coding.Provability is the non-negotiable requirement in financial AI - it is not enough to produce correct output, you must be able to prove the output is correct and traceable back to the source contract.Organizations that want to deploy AI on their contracts first need to standardize internally on what outcomes they actually want - two people on the same team asking the same question about the same contract should not produce two different answers.Insightful Quotes: "The misconception is that difficult problems can just be solved by throwing something into ChatGPT and having the answer come out the other side. In our case, the at-scale piece is everything. Those intelligence tools are still individualized tools - to do things at scale for an entire enterprise still takes specific tooling, specific thought, and specific expertise." - Deepak Bapat "What we're trying to do is move from a place of unstructured data to provable and correct structured data. That is what Tabs is built around - and that is what most of these other systems simply cannot handle." - Deepak Bapat "When you think about the legacy players that were more rigid SaaS tools with manual entry and brittle connectors - what was intractable about that model is exactly what generative AI made solvable. The ability to reason over the words in a document, understand what they meant, and understand what the output should be - that changed everything." - Seth Earley Tune in to discover what it actually takes to build AI that is accurate enough, auditable enough, and elastic enough to handle enterprise revenue data at scale - and what most organizations underestimate before they start. Links LinkedIn: https://www.linkedin.com/in/deepakbapat/ Website: https://www.tabs.inc Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  2. 6d ago

    Earley AI Podcast - Episode 94: Cybersecurity, AI Risk, and Why Security Is a Sales Motion with Taylor Hersom

    How the Threat Landscape Is Being Rewritten and What Organizations Need to Do Before It Gets Ahead of Them Guest: Taylor Hersom, Founder of Eden Data and Managing Director at Riveron Host: Seth Earley, CEO at Earley Information Science Published on: July 28, 2026 In this episode, Seth Earley speaks with Taylor Hersom, Founder of Eden Data and Managing Director at Riveron, a cybersecurity and compliance firm he built and grew before its acquisition in 2025. They explore why security is still treated as a cost center when it should be treated as a sales motion and competitive differentiator, how AI has exponentially expanded the attack surface, why most organizations have adopted AI with almost no security program around it, and how the subscription model Taylor pioneered is now reshaping how professional services firms price and deliver work. Taylor shares candid and specific insights on AI governance standards, the limits of automated threat detection, and why information architecture is the foundation security professionals are finding missing everywhere they go. Key Takeaways: Security is still treated as a cost center by most organizations when it should be viewed as a trust-building and sales motion that directly impacts revenue and brand reputation.Pre-revenue startups are now arriving with a million lines of AI-generated code - the attack surface has expanded exponentially and security programs have not kept pace.Most organizations have adopted AI across the enterprise without any AI-specific security program, controls around LLM access, or governance over what models are allowed to do.ISO 42001 and the NIST AI Risk Management Framework are the clearest starting points for organizations that want to de-risk their AI environment without reinventing the wheel.AI in security has shifted the human role from doing the work to supervising it - but final judgment on whether a threat is legitimate still requires a human and always will.Unorganized, incorrect, or inaccessible data creates systemic risk in AI environments - poor information architecture is what leads to the snowball effect of cascading security failures.The subscription model for security services - pricing for outcomes rather than hours - has proven durable across market conditions and is now becoming the industry expectation.Insightful Quotes: "We naturally leaned into AI from a technology standpoint, and there is almost no security around it to speak of. If you go ask the average company that's using AI across their enterprise, they probably don't have an AI-specific program where they have controls around their LLM and their processes and their access - and that is terrifying." - Taylor Hersom "Rather than go the FUD route - fear, uncertainty, and doubt - you can look at security as a way to build your brand and make it a part of your identity, and be proactive in how you use this when educating customers about how you protect their data." - Taylor Hersom "There's no AI without IA. Security requires information architecture - access controls, data organization, knowing what you have and where it lives. When you start losing control of your data, you start to create risks you don't even know about." - Seth Earley Tune in to discover why cybersecurity in the AI era is no longer just a technical problem - and what organizations need to put in place before the threat landscape gets ahead of them. Links LinkedIn: https://www.linkedin.com/in/taylorhersom/ Website: http://www.riveron.com  Website: https://www.edendata.com   Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  3. Jun 17

    Earley AI Podcast - Episode 93: AI Translation, Brand Voice, and Global Content with Olga Beregovaya

    Why the Gap Between an AI Translation Demo and Enterprise Production Is Wider Than Most Organizations Realize Guest: Olga Beregovaya, VP of AI at Smartling Host: Seth Earley, CEO at Earley Information Science Published on: June 17, 2026 In this episode, Seth Earley speaks with Olga Beregovaya, VP of AI at Smartling, who brings 25 years of experience across every major evolution in natural language processing - from rules-based systems through statistical models, neural translation, and now LLMs. They explore why plugging into a commercial model at token-level pricing is not a translation strategy, how brand voice fractures at 300,000 employees, why information architecture is just as essential for language pipelines as it is for retrieval, and what it actually takes to deliver consistent, on-brand, multilingual content at enterprise scale. Olga shares candid and specific insights on language complexity, the human-in-the-loop imperative, and why the organizations that are finally succeeding with AI have stopped treating it as art for art's sake. Key Takeaways: The price of a commercial model's tokens is not the cost of enterprise AI translation - data integrity, pipeline architecture, linguistic assets, and human review are the real cost drivers. Brand voice fractures the moment every employee can generate content autonomously - a Fortune 10 company discovered it had 300,000 voices overnight after deploying a co-pilot tool. Information architecture is equally essential for language pipelines as for retrieval - nested HTML tags, tokenization failures, and unstructured content break translation before the model ever sees the text. LLMs unlocked context that neural machine translation never had - resolving pronouns, disambiguating terminology, and working at document level instead of sentence by sentence. The assumption that AI translation works equally across all languages is one of the most dangerous misconceptions in the space - morphological complexity, writing systems, and training data representation vary enormously. Human review is not optional even in fully automated pipelines - it is how models learn, how ground truth is established, and how brand consistency is maintained over time. The organizations now succeeding with AI translation have moved from implement-and-fail to measured deployment - defining use cases, respecting prerequisites, and matching tooling to actual requirements. Insightful Quotes: "Yes, you can totally consume your million tokens at a super low price point, but what exactly are you buying for this money? Everybody can totally produce a translation or generate copy, but is it going to represent your brand? That's a different question." - Olga Beregovaya "He installed a co-pilot tool and said, it's great, except my company has 300,000 employees and now my company has 300,000 voices. That's not necessarily what I was prepared for in different countries." - Olga Beregovaya "If you want your models to evolve, and if you want your models to learn, you obviously need somewhere for these models to learn from - and this is where human review comes in. It is always twofold: guaranteeing the quality to your customers, and helping your models evolve." - Olga Beregovaya Tune in to discover why AI translation at enterprise scale requires far more than a model and an API key - and what the organizations getting it right have built that their competitors have not. Links LinkedIn: https://www.linkedin.com/in/olga-beregovaya-04b5/ Website: https://www.smartling.com Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  4. Jun 1

    Earley AI Podcast – Episode 92: Supply Chain Intelligence, Knowledge Graphs, and the Limits of the Easy Button with Ilya Levtov

    Why Supply Chain Visibility Is One of the Most Consequential and Underestimated Applications of AI in the Enterprise Guest: Ilya Levtov, Founder and CEO at Craft.co Host: Seth Earley, CEO at Earley Information Science Published on: June 1, 2026 In this episode, Seth Earley speaks with Ilya Levtov, Founder and CEO of Craft.co, a supplier intelligence platform that uses AI and knowledge graphs to give enterprises and government agencies visibility into their full supply networks. They explore why most organizations believe they have adequate supply chain visibility when they do not, why a simple risk score will always mislead, and how cross-correlating data streams surfaces risks that no human - and no generic LLM - would ever find alone. Ilya shares candid and specific insights on building knowledge graphs for mission-critical infrastructure, why only one percent of enterprise knowledge exists inside today's LLMs, and how the give-to-get model is turning supply chain intelligence into a shared strategic asset. Key Takeaways: Most enterprises believe their top-supplier relationships give them adequate visibility - but the middle and long tail of a supply network, which can run to 20,000 or 30,000 suppliers, remains almost entirely opaque.Supply chain is a misnomer - it is a complex, multi-dimensional network where companies are simultaneously suppliers, customers, and competitors to each other.A simple risk score is not meaningful and not actionable; supplier risk is deeply contextual and requires human judgment to weigh cost, probability, and consequence together.Cross-correlating data streams reveals hidden risks that no single source can surface - including correlations between employee morale and cybersecurity vulnerability that have proven highly predictive.Only approximately one percent of enterprise knowledge exists inside today's LLMs - which is exactly why a specialized knowledge graph grounded in proprietary data is essential before applying AI.AI has compressed analyst work on a supplier report from eight hours to under 30 minutes - but the decision of what to do with those findings still requires human judgment and always will.The give-to-get model and supplier passporting allow enterprises to share intelligence across a shared supply network without compromising their own competitive position.Insightful Quotes: "Only 1% of enterprise knowledge approximately exists inside the LLMs today. Companies don't want to give all of their data to the LLMs. Data providers don't want to give it for free either. That's why you need a specialized approach - leverage the power of the models on your own data set and on your knowledge graph." - Ilya Levtov "A financially vulnerable supplier becomes a target for adversarial capital - entities coming in from unfriendly nations looking to survive. You're connecting two different data sets, connecting entities, and getting to a very significant risk insight you need to act on before it becomes a problem for your enterprise." - Ilya Levtov "Organizations compete on their knowledge - knowledge of customers, knowledge of solutions, knowledge of supply chains, knowledge of routes to market. Those are competitive advantages. You do not want those inside an LLM. That is why doing this in a way that is internal and proprietary is so important." - Seth Earley Tune in to discover why supply chain visibility is one of the most important and most underestimated applications of AI in the enterprise today - and what it actually takes to build intelligence at the scale the problem demands. Links LinkedIn: https://www.linkedin.com/in/ilya-levtov/ \Website: https://www.craft.co Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  5. May 26

    Earley AI Podcast - Episode 91: Real-Time Voice Intelligence, Fraud Detection, and AI Guardrails with Mike Pappas

    Why Voice Is Not a Solved Problem - and What Real-Time Audio Intelligence Changes for Enterprise AI Guest: Mike Pappas, CEO at Modulate Host: Seth Earley, CEO at Earley Information Science In this episode, Seth Earley speaks with Mike Pappas, CEO of Modulate, whose work began in gaming - one of the most demanding environments for real-time voice intelligence - and has since expanded to enterprise applications including fraud detection, customer abuse prevention, AI agent guardrails, and sales coaching. They explore why transcription is not the same as understanding, what gets lost when audio is reduced to text, and why voice is the most powerful tool fraudsters have. Mike shares candid and specific insights on deepfake detection, the fine line between safety and surveillance, and what organizations need to put in place before deploying voice AI at scale. Key Takeaways: Transcribing voice and understanding voice are not the same thing - intonation, emotion, cadence, and timbre carry information that transcripts cannot capture.Voice AI demos are typically built for pristine environments; the real challenge is building systems that hold up under noise, jargon, and emotional complexity in production.Real-time intervention changes behavior more effectively than after-the-fact review - feedback delivered in the moment produces measurable reductions in repeat offenses.Voice is the most powerful tool for manipulation because it bypasses rational judgment by triggering emotional responses - and AI is now making voice fraud scalable.AI voice agents cannot introspect - they cannot tell when a call is going wrong, which is why a separate supervisory layer is essential for any enterprise voice deployment.The line between safety systems and surveillance systems is real; collecting and storing only what is necessary for the specific risk being addressed is both a privacy and a trust requirement.Before deploying any voice AI, organizations need to define their KPIs clearly - if the system is driving customer satisfaction down, the deployment is failing regardless of what else it is doing.Insightful Quotes: "When you hear a voice, you hear the intonation, you hear the emotion, you hear pregnant pauses - there is so much information being carried in that audio that gets lost when you pull down to a transcript. And whenever we talk to someone who professionally works in a contact center, they are always saying, we know these transcripts are losing tons of good value." - Mike Pappas "If I am actively harassing you and the platform is able to come in and put a stop to it live in the conversation, that feedback actually systematically changes behavior. Getting an email 30 minutes later saying we noticed you did something wrong - that just infuriates people, it does not lead to change." - Mike Pappas "There is a fine line between safety systems and surveillance systems. How do you design voice AI that improves safety and trust but does not cross that boundary that makes users and employees uncomfortable?" - Seth Earley Tune in to discover why real-time voice intelligence is one of the most consequential and least understood frontiers in enterprise AI - and what organizations need to get right before they deploy. Links LinkedIn: https://www.linkedin.com/in/mike-pappas-9a30a858/ Website: https://www.modulate.ai   Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  6. May 26

    Earley AI Podcast – Episode 90: Federated AI, Decision Intelligence, and the Data Architecture Reset

    Why Centralization Is the Wrong Foundation for AI - and What Organizations Need to Build Instead Guest: Todd Barr, CEO at Axonis.ai Host: Seth Earley, CEO at Earley Information Science Published on: May 13, 2026 In this episode, Seth Earley speaks with Todd Barr, CEO of Axonis.ai, a company spun out of a government defense integrator that is bringing federated AI and decision intelligence to high-consequence enterprise workflows. They explore why the demo-to-production gap is one of the most costly misconceptions in enterprise AI today, why centralization was built for business intelligence and not for AI, and what it really means to send your AI to your data rather than the other way around. Todd shares a candid and direct perspective on decision artifacts, AI cost exposure, the risks of vendor lock-in, and why enterprises that give away how they make decisions may be giving away the most valuable thing they own. Key Takeaways: The demo-to-production gap is a form of malpractice - polished AI demos built on curated data create executive expectations that production reality cannot meet. Centralized data infrastructure was built for business intelligence, not AI - it is optimized for reporting, not reasoning or prediction. The premise of agentic AI is decentralization - if agents have to wait for data to be synced and centralized before acting, the architecture is working against itself. Data resists centralization for three distinct reasons: technical constraints, regulatory and compliance requirements, and organizational politics. Decision artifacts - cryptographically sealed records of data used, model applied, and reasoning followed - turn AI-assisted decisions into auditable, improvable corporate assets. Enterprises now face a clear choice: pay in tokens, pay in vendor lock-in, or invest in owning their own AI infrastructure through open source models. How an organization makes decisions is its most proprietary asset - giving that context to a third-party AI platform may be the most consequential thing enterprises are doing right now without fully understanding it. Insightful Quotes: "The misconception is really the gap between prototype and reality, and that's where a lot of these things are falling down right now. Getting people excited about something they can't have is almost malpractice." - Todd Barr "Centralization is almost a fallacy in itself. Whenever you are using data you are changing it, enriching it, doing something with it. It is a fractal nature of data that defies the whole concept of centralization." - Seth Earley "If I'm an enterprise, what do I own in this day and age? I own how I make decisions. Which data I use to make those decisions. If we are just going to give that away, that is like giving our brain away." - Todd Barr Tune in to discover why the most important AI infrastructure decision an enterprise can make right now is not which model to use - but whether they are building a foundation they actually own. Links LinkedIn:   / tbarr  Website: https://axonis.ai Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  7. May 6

    Earley AI Podcast Episode 89: Memory, Power, and the Hidden Constraints of AI Infrastructure

    Guest: Steven Woo, Fellow and Distinguished Inventor at Rambus Host: Seth Earley, CEO at Earley Information Science Published on: May 5, 2026 In this episode, Seth Earley speaks with Steven Woo, Fellow and Distinguished Inventor at Rambus, where he has spent over 30 years at the frontier of memory technology. They explore why memory - not compute - is the binding constraint on AI performance today, how moving data between chips consumes more than half of all power in a high-end AI processor, and what the rise of agentic AI means for infrastructure planning.  Steven shares a rare long-view perspective on the innovation curve for memory technology, the supply-demand dynamics driving prices higher, and the questions enterprise leaders should be asking before signing their next infrastructure contract. Key Takeaways: Memory, not compute, is the primary bottleneck limiting AI performance - and the gap between processor speed and memory speed is widening, not closing. Over 50 percent of the power consumed by high-end AI processors is spent simply moving data on and off the chip, not performing computation. Stacking memory components closer together can reduce energy costs dramatically but introduces new challenges around heat dissipation and power delivery. Training and inference have very different memory profiles - understanding both is essential for organizations architecting AI infrastructure at scale. Agentic AI compounds the memory challenge significantly, because one user can spin up multiple agents that each spawn further agents, multiplying context and capacity demands. Memory prices have risen sharply due to supply-demand imbalance - organizations are now signing long-term supply agreements to lock in capacity, just as they do for power. The most important question enterprise leaders can ask their infrastructure providers is how much experience and demonstrated reliability they have - downtime during model training can be catastrophic. Insightful Quotes: "Memory has become a big bottleneck. In many cases, in AI, your speed at which you can actually process information and create new large language models is really gated by the speed and availability of memory." - Steven Woo "More than 50 percent of the power is spent in circuits just trying to move data on and off the processor. It's pretty astounding to think that as companies plan how much power they need, a lot of it is really related to simply moving data back and forth." - Steven Woo "People think of compute in terms of gigawatts. But it turns out it's really the movement of that data - and nobody talks about that. It's the silhouette behind the curtain that's actually constraining everything else." - Seth Earley Tune in to discover why the future of AI depends as much on memory engineering as it does on model development - and what enterprise leaders need to understand about the infrastructure constraints shaping every AI investment they make. Links LinkedIn: https://www.linkedin.com/in/stevencwoo/ Website: https://www.rambus.com Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

  8. Apr 27

    Earley AI Podcast - Episode 88: Digital Twins, Agentic AI, and the Future of Work with David Shim

    From Meeting Intelligence to Personal AI: How Digital Twins Are Reshaping How We Work Guest: David Shim, Co-Founder and CEO at Read AI Host: Seth Earley, CEO at Earley Information Science Published on: April 27, 2026 In this episode, Seth Earley speaks with David Shim, Co-Founder and CEO of Read AI, the fastest-growing meeting intelligence platform globally with over 5 million monthly active users. They explore how AI is moving beyond summarization toward recommendation and autonomous action, what it really means to build a digital twin grounded in your actual work history, and why the organizations getting the most from AI are the ones that treat it like a trainable intern rather than an out-of-the-box solution. David shares candid insights on agentic guardrails, data privacy, workforce transformation, and why access to personal AI may one day be considered a basic human right. Key Takeaways: AI is moving from task execution to recommendation - the next frontier is AI that proactively surfaces what you should do next.A digital twin is only as good as its context; weighting recent activity more heavily produces responses that actually reflect how you think and work today.Treating AI like a trainable intern - feeding it your emails, files, meetings, and tools - is what separates high-value users from disappointed ones.Native permissions are the cleanest foundation for digital twin privacy; building new rules for every edge case creates the vulnerabilities you are trying to avoid.Agentic guardrails should be built in from the start, not bolted on - autonomy without oversight erodes trust and adoption faster than it builds them.The tension between organizational IP and individual work style is real; your tone, voice, and preferences belong to you, even when the content belongs to the company.AI is a great leveler - emerging markets and individuals with access to these tools are already competing on equal footing with developed market counterparts.Insightful Quotes: "It's not plug and play today. You have to give it more context - your emails, your files, your CRM, your meetings. When you have all that data, now your intern is learning as you go, and it's pulling from your experience as the mentor." - David Shim "Your digital twin knows I hate meetings after three hours straight. After three hours, my engagement goes down, my sentiment goes down - so it puts in a buffer. That's the first part. Then it starts asking: what happens when people ask you a question?" - David Shim "You can't take the AI's version of the world as a representation of your version of the world. What's more valuable is your secret sauce, your knowledge, your expertise - you have to give it examples of your work, give it your perspective, not just take the LLM's." - Seth Earley Tune in to discover how digital twins and agentic AI are transforming the way individuals and organizations work - and what it takes to get real value from the technology before it gets ahead of you. Links LinkedIn: https://www.linkedin.com/in/davidshim/ Website: https://read.ai Thanks to our sponsors: VKTREarley Information ScienceAI Powered Enterprise Book

See All (95)

Ratings & Reviews

4.3
out of 5
7 Ratings

About

In this podcast hosts Seth Earley invites a broad array of thought leaders and practitioners to talk about what's possible in artificial intelligence as well as what is practical in the space as we move toward a world where AI is embedded in all aspects of our personal and professional lives. They explore what's emerging in technology, data science, and enterprise applications for artificial intelligence and machine learning and how to get from early-stage AI projects to fully mature applications. Seth is founder & CEO of Earley Information Science and the award-winning author of "The AI Powered Enterprise." 

Information

  • Creator
    Seth Earley
  • Years Active
    2021 - 2026
  • Episodes
    95
  • Rating
    Explicit
  • Copyright
    © 2026 Earley AI Podcast
  • Show Website
    Earley AI Podcast