Anthropic just raised $1.5B with Blackstone, Hellman & Friedman, and Goldman Sachs. OpenAI raised $4B with TPG, Bain, and Advent the same day. Both labs are building delivery organizations explicitly modeled on Palantir's forward-deployed engineer playbook.

In this episode of Humans In The Loop, Kasey Roh and Galina Fendikevich unpack the following question: are FDEs the right model for enterprise AI adoption?

00:00 Trivia: Anthropic vs OpenAI on the secondary market
04:00 The news: $5.5B raised to fund delivery organizations
06:30 Why Wall Street is funding deployment, not just models
08:30 "Having the model alone doesn't change your workflows" — Goldman's quiet admission
12:00 Why enterprise environments are a rat's nest (and why APIs alone don't ship)
15:00 95% of enterprise AI pilots fail. MIT NANDA on the gap between API and production.
17:30 What Palantir's 20-year playbook actually proves
19:00 The 6,000-company ceiling: why this only works at the top
22:30 FDE economics: $350K to $500K+ salaries and what they mean for customers
24:00 The permanent retainer: why FDE deployments don't end
26:30 Lock-in dressed as transformation
28:30 The contrarian take: this isn't bullish, it's the opposite
30:00 The mid-market opportunity nobody is solving yet

🗣️ Follow us on YouTube: youtube.com/channel/UCJbkvAHHTRlo2EXkGyUrA-g/
🎧 Also on Spotify and Apple Podcasts: 
- Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo
- Apple Podcast: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#enterpriseai #humansintheloop #humanintheloop #generativeai #aiadoption

Token Maxing: The New KPI or the Next Enterprise AI Budget Trap?

Token economics is reshaping how enterprises think about AI spending. In this episode of Humans In The Loop, Kasey Roh and Galina Fendikevich unpack Jensen Huang's token maxing framework, Uber's blown 2026 AI budget, Meta's failed leaderboard experiment, and why only 28% of finance leaders can measure AI ROI.

00:00 Token Maxing Intro
01:09 Jensen Token Trivia
03:22 Budget Reality Check
06:24 Token Economics Framework
09:57 Tokens Per Employee Debate
12:06 Meta Leaderboard Backfire
17:11 Uber Budget Blowout
21:49 CFO Visibility And ROI Gap
27:10 Governance And Spend Consolidation
30:24 Open Source Comeback
35:44 Platform Over Single Model
37:54 Wrap Up And Where To Listen

🎧 Also on Spotify and Apple Podcasts: 
- Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo
- Apple Podcast: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#enterpriseai #humansintheloop #humanintheloop #generativeai #aiadoption

Enterprise AI Reality Check: Only 2 Insurers in Fortune 500 Show Quantified ROI (Plus Maersk, HPE & Tiffany Success Stories)

In the first episode of Humans in the Loop, Kasey Roh and Galina Fendikevich introduce the show’s focus on the gap between AI hype and real enterprise adoption, especially in regulated industries like insurance. Kasey shares analysis of Q4 earnings-call CEO remarks across ~60 companies, tracking AI mentions, quantified ROI, and 2026 plans; among Fortune 500 insurance carriers, only Chubb and Travelers cited quantified AI ROI, and Travelers mentioned AI about 45 times. They discuss why insurance may be more conservative and argue for small, tangible AI deployments rather than endless pilots. They highlight success cases: Maersk pivoting from a $100M blockchain effort to AI route optimization saving $375–$500M annually in fuel, HPE building a finance-ops AI agent (“Alfred”) with Deloitte, and Tiffany using genAI to create 700+ monograms for engraving.

00:00 Welcome to the Show
00:15 Meet Galina
01:04 Why Humans in the Loop
02:20 Meet Kasey
03:15 Earnings Call AI Study
04:38 Insurance ROI Reality Check
07:20 AI Mentions and Metrics
11:33 2026 Plans and Budgets
15:30 Why Insurance Moves Slow
16:09 Start Small with AI
18:43 Maersk Fuel Savings Win
21:48 HPE Alfred Finance Agent
24:48 Tiffany Engraving with AI
27:10 Wrap Up and Thanks

🎧 Prefer audio? 
- Listen on Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo | 
- Apple Podcasts: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#HumansInTheLoop #EnterpriseAI #AIGovernance #HumanInTheLoop

Can AI Rewrite the Past? Mainframes, COBOL Modernization, and Real-World AI Wins

Kasey and Galina discuss insights from the NAMIC mutual insurance conference in Chicago and why AI adoption can feel like an uphill battle outside the tech bubble. We dig into mainframes—over 90% of global credit card transactions and about 68% of IT workloads run on them at roughly 6% of global IT cost—and why hybrid architectures can enable AI without migrating off legacy systems. We unpack Anthropic’s claim that Claude can help modernize COBOL and the market reaction that coincided with a $30B IBM market-cap drop, while emphasizing the real challenges of testing and validation amid an aging COBOL workforce. We also cover Marriott’s mixed results with guest-facing AI versus “invisible AI” in core operations, and Hallmark’s approach of keeping humans in the loop while using AI for supply chain, personalization, and workflow efficiency.

00:00 Welcome Back Intro
00:14 Chicago Conference Takeaways
01:37 Can AI Rewrite Legacy
02:53 Mainframe Trivia Shockers
04:57 Why Mainframes Still Win
09:27 Hybrid AI On Mainframe
12:00 Claude Modernizes COBOL
13:32 Talent Gap And Validation
17:26 IBM Watson And Urgency
19:23 Marriott AI Concierge Fail
22:44 Invisible AI Backend ROI
25:42 Hallmark Human In Loop
31:18 Trust But Verify Wrap Up
33:44 Closing And Call To Action

🎧 Also on Spotify and Apple Podcasts: 
- Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo
- Apple Podcast: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#enterpriseai #humansintheloop #humanintheloop #aigovernance #generativeai

When Elon's $1B automation turned into a tent full of humans

Kasey and Galina discuss how businesses often try to force operations to fit AI, instead of mapping workflows in detail so AI can adapt to the business and reveal high-ROI opportunities. Kasey shares her 2018 Tesla experience during Model 3 “Production Hell,” when Elon Musk pushed near-total factory automation to hit 5,000 cars/week but production lagged, costs soared (about $100M/week), and we resorted to manual work and a tent-based assembly line. We explain the turnaround: breaking work into granular steps, using real-time KPI dashboards, keeping humans in the loop, iterating, and automating only where it fit—paralleling today’s AI POCs that fail in production without integration and workflow understanding.

00:00 AI Adoption Reality Check
02:37 Meet Kasey and Tesla Roots
04:43 Production Hell and Funding Chaos
07:27 The Fully Automated Factory Bet
13:16 Why Automation Broke Down
18:16 The Tent Line Human Pivot
22:22 Workflow Mapping Turnaround
29:43 Leadership Lessons for GenAI
37:55 Avoid FOMO and Cautionary Tales
40:32 Final Takeaways and Wrap Up

#enterpriseai #humanintheloop #humansintheloop #generativeai #aigovernance #tesla

What Happens When You Lay Off 30,000 SWEs and Let AI Drive?

Kasey and Galina unpack the gap between the Silicon Valley AI frenzy and slow enterprise adoption, starting with observations from New York Build Week where construction is only beginning to explore AI. We discuss trivia and stats: Google and Microsoft publicly report over 30% of internal code is written by AI, while Meta predicts 50% within a year. We review layoffs and a Forrester finding that about 55% of leaders who cut jobs due to AI expect to rehire, often at lower pay or offshore, alongside data showing software engineering job postings rising. We contrast PWC’s AI rollout—governed tooling, peer-led “prompting parties,” embedded change agents, and function-specific use cases—with Amazon’s layoffs and mandate to use its internal coding tool, plus reported outages and concerns about AI-generated errors. We close by questioning training pipelines, advocating apprenticeships, and emphasizing human judgment, curiosity, and creativity in an AI-assisted future.

00:00 AI Hype vs Reality
00:28 Construction Industry Check-In
01:15 Live Trivia Setup
02:51 How Much Code Is AI
05:38 Layoffs and Rehiring Regrets
08:58 Why Engineer Jobs Rise
12:03 PWC AI Rollout Playbook
14:36 Prompting Parties and Change Agents
17:33 Upskilling vs Layoffs Tradeoffs
22:56 Amazon Mandates and Outages
28:57 Leadership Lessons for Adoption
30:05 Apprenticeships and Creativity Future
32:26 Curiosity and Closing Thoughts

🎧 Also on Spotify and Apple Podcasts: 
- Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo
- Apple Podcast: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#enterpriseai #humansintheloop #humanintheloop #aigovernance #generativeai

You Vetted the Vendor. Did You Vet the Model?

Kasey and Galina discuss a major blind spot in enterprise AI procurement: security questionnaires heavily vet vendors but often fail to ask which foundation model underlies the product. After consolidating five enterprise questionnaires into 130 unique questions, we found they focus on data protection, controls, and governance yet omit direct model provenance. We highlight how startups can host an open-source model in US cloud regions, fine-tune it, and still pass most reviews without disclosing the base model, complicating accountability when failures occur. We also share market signals that many US startups using open-source stacks rely on Chinese-origin models and note growing Chinese model usage. The episode argues model choice affects delegated decision-making and bias, and suggests adding at least one question: name and version of the foundation model.

00:00 AI Between Hype and Reality
00:35 Security Questionnaire Pain
01:18 Missing Model Vetting
03:36 AI Collapses the Stack
04:50 Open Source Trivia Shock
08:42 Chinese Models Rising Fast
13:06 Why Model Origin Matters
19:26 Questionnaire Blind Spots
21:28 Simulation Passing the Audit
24:59 Accountability When AI Fails
27:04 Provenance vs Trade Secrets
32:09 Bias Example Nuclear Game
38:13 Practical Fix One Question
41:44 Wrap Up and Call to Action

🎧 Also on Spotify and Apple Podcasts: 
- Spotify: https://open.spotify.com/show/5CAwQDcofIzMM9p5GIChOo
- Apple Podcast: https://podcasts.apple.com/us/podcast/humans-in-the-loop/id1892864937
💼 Connect with Us:
- Kasey: https://www.linkedin.com/in/kaseyroh/
- Galina: https://www.linkedin.com/in/gfendi/

#enterpriseai #humansintheloop #humanintheloop #generativeai #aiadoption