Disambiguation

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Chris Fitkin, Co-Founder and Partner at Metacto, about why most companies are spreading AI experiments too thin across every department, why that wide-and-shallow approach produces shelfware instead of results, and what it actually takes to get AI into production in the mid-market.
Chris has 25 years in software engineering, a master's in software engineering, and is an AWS Certified Solutions Architect. He has been a CTO in cybersecurity, led due diligence at private equity firms, and co-founded Metacto with Garrett Fritz. The firm started in mobile app development and fractional CTO work, then pivoted to operational AI for mid-market and private equity-backed companies after seeing how dramatically the landscape was shifting.
The conversation covers the AI sprawl problem (88% of companies are using AI in at least one function per McKinsey, but only 5% see measurable impact per MIT), the repeating cycle of cool demo to mixed results to low adoption to shelfware, why companies that go narrow and deep are more than twice as likely to see measurable bottom-line impact compared to those that scatter experiments across the organization, the Five Signals framework for picking your first AI workflow (email, spreadsheets, copy-paste relays, contactless approvals, repeat expert answers, report factories), why mundane workflows are the right starting point, the 5%/95% gap between demo and production (access control, business rules, quality checks, human review, audit trails, monitoring, versioning, ownership), failure modes that compound when nobody catches a bad LLM decision for five days, context engineering and the three parts of building good context (transactional data, document corpus, codified business rules and domain knowledge), why business first has to replace AI first, the mid-market pricing reality (Anthropic offered one client $1.2 million a year in token minimums, Mars Inc pays $600,000 a month to Google Gemini), the AI Engineering Maturity Index assessment tied to EBITDA and enterprise value, and practical advice for stuck leaders.
Timestamps:02:07 - Change is the only constant across 25 years of technology cycles03:19 - App Store submissions doubled while downloads decreased04:16 - AI sprawl: 88% using AI, only 5% see measurable impact05:09 - The shelfware cycle: cool demo, mixed results, low adoption07:20 - Shadow AI is the new shadow IT07:43 - Narrow and deep is 2x more likely to produce bottom-line impact08:43 - Going deep: problem first, define success metrics before you build11:14 - Solution looking for a problem versus problem looking for a solution12:27 - Five Signals framework for picking your first AI workflow13:03 - Mundane workflows are validated by human capital investment14:31 - The 5%/95% gap: demo is 5% of the work, production is 95%16:05 - Failure modes: bad decisions compounding, admin database access exposed16:59 - Context engineering versus prompt engineering19:06 - Transactional data: clean, current, deduplicated data warehouse19:27 - Document corpus: proposals, QBRs, deliverables tagged with recency20:04 - Business rules and domain knowledge: codifying what lives in people's heads21:32 - Business first, not AI first: product managers lead engagements22:25 - Requirements engineering: the discipline everyone is rediscovering24:01 - The mid-market gap: too small for McKinsey, too complex for license distribution26:42 - Start small, measure lift, use wins to fund the next projects27:51 - AI Engineering Maturity Index: 30-day assessment tied to financial metrics33:54 - Focus on people: dedicated time, builders and advocates, adoption training36:52 - Leading a hybrid workforce: managing human and digital workers38:33 - Recommendation: Reid Hoffman's Masters of Scale with IBM CEO Arvind Krishna
Guest: Chris Fitkin, Co-Founder & Partner, MetactoHost: Michael Fauscette, CEO & Chief Analyst, Arion Research

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Robyn Bolton, Founder and Chief Navigator of MileZero, about why the same organizational forces that have been killing innovation for decades are now killing AI adoption, and what leaders can do differently.
Robyn's career started at Procter & Gamble, where she helped develop Swiffer, one of the most successful consumer product launches in recent history. From there she became a partner at Boston Consulting Group and then worked at Innosight, Clayton Christensen's innovation firm. She now runs MileZero, where she works with Fortune 500 companies like Medtronic, Nike, and Nestle to build pragmatic innovation capabilities. She published Unlocking Innovation last year and teaches at Massachusetts College of Art and Design.
The conversation covers why innovation is a leadership problem and not an ideas problem (organizations are full of ideas, but leadership behaviors train people to stop sharing them), the five organizational antibodies that neutralize anything new and unfamiliar, how those same antibodies show up in AI adoption, the Revenge of Clippy (a Fortune 500 company launched a custom chatbot and employees responded with malicious compliance, asking questions they already knew the answer to just to check the box), why Robyn now defends innovation theater when companies commit to the season rather than just the show, the ABCs framework (Architecture, Behavior, Culture) and why 30 years of focusing on architecture alone has produced no improvement in corporate innovation results, the AI pilot trap and why falling in love with the solution instead of the problem is the root cause, Jobs to be Done applied to AI adoption, why automating a broken process just makes it fail faster, continuous change versus project-based change management, why scale needs to be defined at the start of a pilot because not everything should go company-wide, human infrastructure as the missing budget line in AI readiness, the replacement mistake versus AI-augmented humans, the sycophancy problem in LLMs and why manufactured trust is dangerous, and practical advice for stuck leaders.
Timestamps:00:00 - Introduction00:31 - Episode title and guest intro00:47 - Robyn's path: P&G, Swiffer, BCG, Christensen's firm, MileZero02:08 - Choosing process over product: the deeper root cause03:57 - Innovation is a leadership problem, not an ideas problem05:46 - Logical responses with unintended consequences07:57 - Organizational antibodies and AI adoption09:09 - The Revenge of Clippy: malicious compliance with a company chatbot11:13 - Defending innovation theater: the season versus the show14:33 - Change management as a checklist versus ongoing behavior change15:12 - The gym analogy: one visit does not make you healthy16:00 - The ABCs framework: Architecture, Behavior, Culture16:32 - 30 years since the Innovator's Dilemma, results have not changed19:23 - The AI pilot trap: fall in love with the problem, not the solution21:22 - People, workflow, culture, technology: in that order21:53 - Incentives determine behavior: change the metrics, change the outcome23:13 - Jobs to be Done applied to AI adoption24:12 - Blank sheet of paper: stop cramming AI into broken processes25:45 - Continuous change versus project-based change management26:57 - Learning mindset: reassess after every step28:52 - Scientific method applied to business: test hypotheses individually32:06 - What scale really means: not everything goes company-wide33:16 - Human infrastructure and AI readiness34:06 - The replacement mistake: AI augments, not replaces35:53 - Sycophancy and manufactured trust in LLMs38:01 - Practical advice: find the problem, ask why, use the five whys39:48 - Recommendation: The Coaching Habit by Michael Bungay Stanier
Guest: Robyn Bolton, Founder & Chief Navigator, MileZeroHost: Michael Fauscette, CEO & Chief Analyst, Arion Research

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Justin Watt, Co-Founder and CEO of Switchboard, about why mid-market companies cannot just layer AI onto broken processes and scattered data and expect results, and what they need to do first.
Justin's path ran from IBM, where he worked on large government projects and learned more about what not to do than what to do, through MetaLab working with Silicon Valley companies like Amazon and Uber, through project management and IT leadership, to co-founding Switchboard, which focuses on helping mid-market non-tech companies modernize and adopt AI and automation.
The conversation covers why "digital transformation" has run its course as a term (most companies replaced tools but never actually transformed how they work), why Justin uses "modernization" instead, the 2006 problem (90% of mid-market leaders think their organization is technically capable because someone can build a pivot table in Excel), the "Data Lake" Excel file (a real client who named their spreadsheet that), why Excel came out the same year as Back to the Future and many companies still run their data on it, human duct tape (paying people to move spreadsheet cells between files), the 14,000-row rate sheet across four Excel files maintained by different people, why automating a broken process just automates the brokenness, mapping processes before touching technology, the governance gap (most companies treat governance as a log instead of a framework), the HR chatbot disaster (a company rolled out a chatbot that let interns look up everyone's salary and performance reviews), goal-based AI risks (the Anthropic vending machine story where someone got five iPhones for $500), board pressure without a defined outcome, margin driving versus revenue driving, the crawl-walk-run approach, and practical first steps for mid-market leaders.
Timestamps:00:00 - Introduction00:31 - Episode title and guest intro00:45 - Justin's path: IBM, MetaLab, Switchboard01:52 - Learning what not to do at IBM: meetings about meetings02:58 - McKinsey stat: 65% of digital transformation fails, 90% is change management04:02 - Why "modernization" instead of "digital transformation"06:13 - The 2006 problem: 90% of leaders stuck technically06:40 - Your A-player's tech skill is a pivot table09:19 - The Matrix Code moment: a client's breakthrough10:59 - The "Data Lake" Excel file11:47 - Human duct tape: people moving cells between spreadsheets14:09 - The 14,000-row rate sheet across four files15:18 - You cannot automate a broken process16:40 - Map the process: get everyone in the room18:10 - A third of steps exist because of software limitations19:15 - SaaS trust is gone: ten years of "it's on the roadmap"20:06 - Governance is a log, not a framework21:27 - The HR chatbot disaster: interns looking up salaries24:36 - Trust and connecting AI to internal data26:33 - Goal-based AI exposes bad goal definition27:22 - The Anthropic vending machine: five iPhones for $50029:15 - Board pressure without a defined outcome31:13 - AI for margin driving, freeing time for revenue33:14 - First three months is modernization, not AI35:22 - Pick the most broken area and map the process37:34 - Just start playing with AI personally39:48 - Recommendation: Ben Evans quarterly AI macro presentation
Guest: Justin Watt, Co-Founder & CEO, SwitchboardHost: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Eva Minkoff, Founder of Bold Being, about why AI adoption failures are not technology problems but identity threats, and why organizations cannot train or explain their way through them.
Eva spent two decades in healthcare across clinical research, bedside care, media, marketing, and startups as both a co-founder and early employee. She gave a TEDx talk called "Five Minutes to Fix Our Broken Healthcare System" about the patient-doctor relationship and the collapse of self-trust under relentless pressure. That same pattern of pressure-driven identity loss is now playing out across every industry as AI reaches professionals whose authority and sense of legitimacy are built on a specific kind of expertise.
The conversation covers why AI adoption is an identity threat rather than a change management problem, why employee sabotage of AI initiatives (30 to 70 percent in some studies) is survival behavior rather than insubordination, the Luddite parallel (skilled workers whose livelihoods were being erased, not irrational resisters), why you cannot explain someone out of an identity threat, compliance theater (employees using AI to produce outputs and then quietly redoing the work themselves), decision latency in leaders who were previously decisive, leadership brittleness (senior leaders going rigid or withdrawing under AI pressure), Eva's three-stage framework (instability, autopilot, durable agency), sustainable adaptability versus just staying current, regulated empathy in healthcare and its parallel in every industry, a client transformation story (Chief Patient Safety Officer who eliminated ED wait times and saved $50 million after doing identity work rather than operational work and was promoted to CMO), and three practical steps leaders should take right now: name the instability out loud, audit your own autopilot, and change the metric from adoption speed to adaptive capacity.
Timestamps:00:00 - Introduction00:41 - Eva's background: two decades in healthcare, TEDx talk, coaching03:20 - Daughter born on ChatGPT launch day: living in both disruptions04:04 - Identity threat versus change management06:26 - Employee sabotage: the Luddite movement of our era07:16 - Survival behavior, not insubordination09:09 - Skills training assumes a knowledge barrier, the real barrier is psychological09:59 - Not ethical unpreparedness but human unpreparedness10:09 - Decision latency, performative adoption, compliance theater11:13 - Leadership brittleness12:24 - Resistance is information, not a problem to suppress15:57 - Three-stage framework: instability, autopilot, durable agency19:22 - Sustainable adaptability: not staying current, building internal capacity22:07 - Holding ambiguity without defaulting to paralysis23:47 - Healthcare as the most human-dependent industry25:08 - Regulated empathy: suppressing emotional responses as default operating mode27:33 - Client story: Chief Patient Safety Officer to CMO in one year30:06 - Three practical steps for leaders30:49 - Step one: name the instability out loud31:40 - Step two: audit your own autopilot32:48 - Step three: change the metric to adaptive capacity34:45 - Recommendation: Brene Brown, Atlas of the Heart, shame versus guilt
Guest: Eva Minkoff, Founder, Bold Being
Eva is currently conducting research interviews for her book on how senior leaders are navigating AI-driven change. If that is your experience right now, request a private 30-minute conversation here: calendly.com/boldbeing/conversation To connect further: LinkedInWebsiteHuman Leadership Now Substack/Newsletter
Host: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Omid Pakseresht, CEO of Good Folio, about what it takes to build AI systems that actually work in enterprise settings, why model quality is no longer the bottleneck, and why adoption is a design problem that starts long before the technology.
Omid studied math at Oxford and quantitative finance, built risk management tools, and was the first product person at two AI startups focused on knowledge graphs for finance and supply chain. He founded Good Folio about five years ago as an applied AI venture platform that sits between a platform company and a venture studio, where each deployment improves the next through cross-domain learnings and reusable infrastructure.
The conversation covers the Unilever engagement (computer vision for small retail stores across Indonesia, Philippines, and Pakistan where markets went from pen and paper straight to agentic WhatsApp workflows, skipping digitization entirely), the Inspector product (AI-driven financial promotion compliance that redesigned the function rather than automating it), why better models no longer directly lead to better outcomes, why the constraint is now system design rather than model performance, the discovery process for mapping workflows and identifying leverage points, why building agents is like hiring new people for your organization, architectural choices driven by where workflow risk sits (small distributed models for emerging markets vs. strict guardrails for compliance), the cross-vertical pattern of systems over models, why incentive alignment is the most underappreciated adoption bottleneck, why too much focus on efficiency crowds out growth thinking, why adoption is a design problem, pilot fatigue, and practical advice for business leaders.
Timestamps:00:00 - Introduction00:44 - Omid's background: Oxford, finance, AI startups, the hospital backroom moment02:39 - Good Folio: applied AI that works in enterprise settings03:46 - The venture platform model: most AI problems are system-level problems06:01 - Unilever: computer vision in emerging-market retail06:51 - From pen and paper straight to agentic WhatsApp workflows08:42 - Vision AI and agent systems for orders and replenishment09:34 - 5-8% sales increase in fast-moving consumer goods10:20 - Inspector: financial promotion compliance11:19 - AI-generated content creates more marketing but also more risk11:51 - Continuous monitoring agents trained with compliance officers12:53 - Marketing happier, compliance can sleep at night14:24 - The shift from model performance to workflow design15:05 - Better models don't directly link to better outcomes15:55 - Impressive pilots that don't scale17:22 - Discovery: mapping workflows and identifying leverage points18:25 - Building agents is like hiring new people19:03 - Building is the fast part; understanding requirements is the hard part21:15 - Architectural choices driven by workflow risk23:18 - Same technology, deployed differently for compliance25:17 - Cross-vertical patterns: systems over models26:50 - The adoption bottleneck: incentive alignment28:41 - Adoption as a design problem31:00 - Don't start with "what can I do?" Start with "where does the system break?"33:23 - Pilot fatigue and why the first success unlocks the rest34:38 - Recommendation: Matt Lerner, Growth Levers
Guest: Omid Pakseresht, CEO, Good FolioHost: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Gavin McMahon, Co-Founder and CEO of fassforward, about why AI adoption is not a technology problem but a human problem, why traditional change management is broken, and how storytelling can move people through the fear and uncertainty that AI has created.

Gavin is an engineer by training with 30 years of consulting experience across automotive, defense, and technology. He worked at Gartner during the early internet era, helping move the company from paper delivery to online. In 2001, he co-founded fassforward, which has grown into a leadership and storytelling consultancy serving clients like Verizon and Mastercard. He recently published Story Business.

The conversation covers why change management is a power dynamic problem (change managers have responsibility but no power), the shift from change management to change leadership, why the rate limiter on AI adoption is the organization's appetite for change, Hemingway's iceberg theory and how people fill in scary narratives when leaders leave gaps, the staircase problem (going down toward productivity is a race to the bottom), the shopping mall to Amazon value shift and why it is happening in dog years with AI, the motive triangle of hope, fear, and reason, why employees are sabotaging AI initiatives and the IKEA retraining model as the right approach, why AI works like an army of ants at the word and sentence level while humans think at book and chapter level, the production / coordination / judgment framework for splitting work, the judgment pipeline gap, why leaders should ask "am I using traditional thinking to solve a nontraditional problem," the tragedy of the commons across individual, organizational, and societal competition, the social media parallel, and why there is no AI strategy (just AI accelerating your business strategy).
Timestamps:00:00 - Introduction00:44 - Gavin's background: engineering, Gartner, fassforward, Story Business01:41 - Why AI is about human engineering: decision and judgment03:07 - Execution validates strategy04:44 - Why "change management" is the wrong term05:39 - The power dynamic: responsibility without power06:50 - Change leadership, not change management07:30 - The rate limiter: organizational appetite for change09:08 - You lead people, you manage work10:06 - The river and rapids metaphor: pools of stillness11:27 - Storytelling as a mechanism for AI adoption11:52 - Hemingway's iceberg theory: people fill in the scary parts13:26 - AI as productivity hack vs. real workflow change15:50 - The staircase: going down toward productivity is a race to the bottom17:22 - Shopping mall to Amazon: a 20-year value shift in dog years20:00 - The motive triangle: hope, fear, and reason23:38 - Employee sabotage and the origin of the word "sabotage"25:20 - IKEA's retraining model: the right way to activate AI26:38 - AI-sized peg in a square hole28:52 - AI is an army of ants, not a human-shaped replacement31:15 - Production, coordination, and judgment: the three types of work33:27 - The judgment pipeline gap: pig in a python35:19 - Am I using traditional thinking for a nontraditional problem?36:32 - Tragedy of the commons: competition at every level42:17 - The social media parallel: same path, faster45:25 - What leaders should do first: business strategy, not AI strategy47:53 - The Chief Blank Officer as a telltale sign49:09 - Recommendation: Bryce Hoffman and Red Teaming
Guest: Gavin McMahon, Co-Founder & CEO, fassforwardHost: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Justin Bolles, CTO of Resultant, about the critical gap between being AI-enabled and being truly AI-ready, and why most enterprises are skipping the architecture, governance, and security work that determines whether AI deployments succeed or fail at scale.
Justin brings over 15 years of experience across software development, cloud infrastructure, and AI consulting. He studied computer information systems at Purdue, founded three startups, and designed software across mobile, web, and embedded systems for both government and private sector clients before becoming CTO at Resultant, a data, technology, and AI consulting firm.
The conversation covers the AI-ready vs AI-enabled distinction (turning on tools vs building the security and governance to use them safely), why security through obscurity died the moment AI could traverse file trees in milliseconds, the internal data leakage risk most companies overlook, metadata-rich ecosystems as prerequisites for AI at scale, the principle of least privilege for AI agents, governance by design, multi-model validation for hallucination safeguards, the Pivot workforce recommendation engine that won State IT Innovation of the Year, AI economics paralleling the early cloud era, and why suggestions over decisions is the right AI autonomy level today.
Timestamps:00:00 - Introduction00:42 - Justin's background: Purdue, startups, Resultant01:48 - The low-code parallel: history repeating with AI03:19 - AI as an intelligent new college grad04:39 - AI-ready vs AI-enabled: the critical distinction05:51 - Security through obscurity is dead07:07 - Hallucination risks: fake citations and deleted databases08:12 - People, process, technology: in that order09:49 - Metadata-rich ecosystems as AI prerequisites13:05 - Institutional knowledge: processes never written down16:25 - Access controls and permission models for agents18:12 - Silicon Valley: AI deleted the UI to fix bugs20:04 - The 100-refunds story: keep customers happy gone wrong22:33 - Governance by design: build it in from the start24:25 - Generational workforce shifts and destination employers27:01 - Cost governance and AI billing surprises29:14 - Multi-model validation: never let a developer test their own code31:10 - AI subsidies will end: Wall Street demands profitability35:21 - Pivot: AI workforce engine for Indiana38:31 - 95%+ satisfaction, outcomes exceeding predictions41:07 - AI economics: pricing mirrors early cloud uncertainty43:08 - Start with the outcome, not the technology45:32 - Don't boil the ocean: narrow use cases to production48:03 - Assessment framework: security, use case, build, scale50:20 - Suggestions over decisions: the right autonomy level51:07 - Recommendation: Chase AI on YouTube
Guest: Justin Bolles, CTO, ResultantHost: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Jackson Lynch, Founder and President of Talent Sherpa, about why most organizations are not ready for AI agents, not because the technology is lacking but because the work itself was never properly designed. Jackson's thesis: AI does not make a system work, it just reflects how broken the system already was. Faster and at scale.
Jackson started in industrial engineering at Boeing on the 777 program, then moved through HR leadership at International Paper, PepsiCo, Nestle, Clearwater Paper, and Sonoma Energy. Now he runs Talent Sherpa as a consigliere for CHROs and CEOs trying to treat human capital as infrastructure rather than overhead.
The conversation covers why AI exposes lazy management (vague job descriptions and activity-based metrics worked because humans filled gaps, but agents cannot), activity-based vs. outcome-based role definitions (the accounts receivable example), work disaggregation before selecting a tool, the freed-up time problem (8 hours saved, 1 hour of productivity), the customer service agent misstep ("keep the customer happy" led to millions in refunds), why "human in the loop" signals the work has not been rethought, pivotal roles (the 5-7% with outsized influence on performance), the CHRO as Chief Workforce Optimization Officer, the Fortune 50 vs. SMB bifurcation, and practical advice for leaders who have not started.
Timestamps:00:00 - Introduction00:47 - Jackson's background: Boeing, HR at PepsiCo, Nestle02:19 - Systems thinking, Deming, operating model over org chart03:03 - AI exposes lazy management03:48 - When AI hits fog, it scales the fog05:11 - Lazy management was invisible because humans filled the gaps06:02 - Automating a broken process breaks it faster and at scale07:10 - Activity-based vs. outcome-based role definitions08:53 - Outcomes enable creativity; activities drive blame10:06 - Agentic AI requires goals, and we are not good at them11:18 - The 1000 cold calls problem12:17 - Customer service agent misstep: millions in refunds13:02 - Work disaggregation: tasks, processes, high vs. low value15:49 - The freed-up time problem: 8 hours saved, 1 gained16:29 - Who owns work design?18:18 - Guardrails: deterministic vs. probabilistic boundaries20:35 - A misaligned agent damages a thousand relationships23:05 - The CEO-CHRO relationship24:11 - Fortune 50 vs. SMBs: company-size bifurcation27:14 - The hybrid workforce: employees, agents, fractional talent28:12 - Pivotal roles: the 5-7% with outsized influence30:05 - Engineering lens vs. change management lens31:31 - Undefined handoffs are where value leaks33:12 - Human in the loop as a transitionary stage36:00 - "Human in the loop" means you have not rethought the work37:17 - Three misconceptions about AI and workforce39:51 - 75% of AI strategies are for show; 90-95% no ROI40:46 - Start with pivotal roles, not tools42:15 - One team first, disaggregate, experiment fast43:07 - Onboarding as a safe starting experiment44:12 - Recommendation: Tony Sarsen
Guest: Jackson Lynch, Founder & President, Talent SherpaHost: Michael Fauscette, CEO & Chief Analyst, Arion Research
Subscribe and turn on notifications so you never miss an episode.

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Dr. Alex Mehr, Founder and CEO of Famous Labs, about why the most important competitive advantage in the AI era is no longer engineering skill but taste, judgment, and knowing what to build. Alex argues that AI has made execution so much easier that the bottleneck has moved upstream: the people who will win are the ones with the strongest product instincts and the clearest sense of what the market actually needs.
Alex's path is distinctive. He grew up in an academic family with plans to become a physicist and university professor. He earned a PhD in mechanical engineering and worked at NASA Ames Research Center in California. Through proximity to Silicon Valley, he caught the entrepreneurship bug and co-founded Zoosk, a dating platform that grew to include a large engineering team. He describes becoming an entrepreneur as crossing the Rubicon: once you do it, there is no going back. He even kept publishing research papers during Zoosk's early years just in case he wanted to return to academia. He never did. Now he runs Famous Labs, which he describes as the ultimate playground for really smart people, constantly launching new AI-powered products.
The conversation covers the taste shift (why engineering thinking still matters but judgment and product instinct now move the needle more than raw coding ability), how Famous Labs hires (they no longer ask technical questions but instead ask "what have you built and why does it look that way?"), the junior engineer pipeline gap (real but short-term and already dissolving as younger engineers pick up AI tools), why layoff narratives are overblown (companies have always right-sized and AI is just the latest excuse), Famous Labs' multi-product model and why AI enables "idea machines" who can pursue multiple products because execution costs have dropped, the death of the software moat (SaaS companies can no longer rely on their code as a competitive advantage), human-centric product philosophy (building things that add value without taking value from other humans), the innovation process (every major Famous Labs breakthrough has come from getting smart people together in a literal hotel room), Heisenberg as "Cursor for chemists" (vertical AI for small molecule drug discovery using chemistry-specific foundation models), why vertical AI is the next major evolution (every profession needs its own cursor equivalent), the SaaS-pocalypse pushback (nobody is going to vibe code their ERP because the real moats are compliance, testing, integration, and business logic), the hybrid workforce concept (every knowledge worker must be AI-enabled or competitors will eat your lunch), game theory dynamics (AI lets competitors enter your territory the way calorie-dense potatoes enabled New Zealand's territorial unification), and practical strategic and tactical advice for executives.
Timestamps:03:20 - The taste shift: judgment matters more than coding ability05:15 - How hiring has changed: "What have you built?" replaces technical interviews06:14 - The junior engineer pipeline and the education system10:05 - Layoff narratives are overblown: AI is just the latest excuse to right-size11:55 - The Industrial Revolution parallel15:11 - The software moat is gone15:54 - Human-centric products16:30 - AI as coworker17:55 - Context windows and training19:33 - The ideation to output pipeline20:16 - Innovation workflow25:16 - Vertical AI: every profession needs its own cursor equivalent27:51 - Specialized models for specific professions30:36 - The SaaS-pocalypse pushback: nobody is vibe coding their ERP34:25 - SaaS companies should use AI to make their tools 10x better37:32 - The hybrid workforce38:44 - Every knowledge worker must be AI-enabled43:11 - The New Zealand potatoes analogy: AI enables territorial expansion45:04 - Systematically enable each role with AI46:12 - Recommendation: Nassim Taleb and Antifragile

In this episode of the Disambiguation podcast, host Michael Fauscette talks with Calvin Cooper, Co-Founder and COO of Neurometric AI, about why the dominant narrative of scaling ever-larger frontier models is giving way to a more practical reality: smaller, specialized models fine-tuned for specific tasks that are faster, cheaper, and more accurate for the vast majority of enterprise AI workloads.

Calvin started his career in early-stage venture capital at NCT Ventures in the Midwest, then founded Rove, a consumer fintech company he took public via a Nasdaq direct listing. Now he and Rob May have co-founded Neurometric AI, which builds task-specific small language model infrastructure. They went full time in August 2025, at a time when the dominant narrative was still "scale compute, scale larger models, AGI," because they were seeing something very different in the research and in practical enterprise deployments.

The conversation covers the surgeon analogy (why you do not hire a surgeon to schedule an email), how their leaderboard proved that no single model is universally best and that inference time tactics can be as impactful as model choice, the AT&T case study (scaling from 8 billion to 27 billion tokens per day while cutting costs by 90%), how 24/7 AI agent runtimes turned subscription costs into six-figure monthly inference bills, why 75% of enterprise AI tasks do not need a frontier model, their marketplace of 115+ task-specific models under 20 billion parameters with fixed monthly pricing per endpoint, the Coding Swarm (orchestrating task-specific SLMs across the development lifecycle), why AI coding agents prove that AI expands jobs rather than replacing them, the four-stage enterprise AI maturity model, why calling a bubble is intellectually lazy (railroads had a bubble too), GPU underutilization and the case for both scaling capacity and improving efficiency, edge compute as the next frontier, and practical advice for enterprises on multi-model orchestration.

Timestamps:
00:00 - Introduction
00:44 - Calvin's background: VC at NCT Ventures, founding Rove, Nasdaq exit
01:37 - Following curiosity: why inference is the largest market opportunity of our lifetime
03:47 - The surgeon analogy: why frontier models are overkill for most tasks
04:58 - Smaller specialized models are faster, cheaper, and more accurate
06:03 - Ship fast: the leaderboard as first proof point
06:26 - No universal good model: different models perform differently at different tasks
07:26 - Early adopter customers and the enterprise journey
07:57 - Real example: Llama model at 4x cost and latency improvement
10:20 - AT&T: 8 billion to 27 billion tokens per day, cut costs 90%
11:30 - The 24/7 agent runtime problem: from subscription to $100K/month bills
13:09 - Multi-model orchestration as the natural next step
14:05 - SaaS pricing disruption and the need for cost predictability
14:53 - 115+ task-specific models under 20 billion parameters
15:06 - Fixed monthly pricing per endpoint with frontier fallback
18:01 - 75% of enterprise tasks do not need a frontier model
18:57 - The Coding Swarm: task-specific SLMs for the development lifecycle
20:34 - AI and jobs: coding agents expanded demand for developers
23:09 - Stage 4 maturity: from monolithic AI to dynamic resource matching
23:31 - First KPI is learning, not ROI
28:16 - Infrastructure: existing GPUs are underutilized
31:14 - Efficiency is not just cost: latency, privacy, compliance
32:11 - Privacy and compliance reasons for multi-model architecture
33:09 - No one God model: the future is less Mission Impossible, more Tron
34:17 - VC perspective shaping the Neurometric business model
37:08 - Practical advice: cut your inference bill by 80-90%
39:28 - Wrap-up

Guest: Calvin Cooper, Co-Founder & COO, Neurometric AI
Host: Michael Fauscette, CEO & Chief Analyst, Arion Research

Subscribe and turn on notifications so you never miss an episode.