The Data Storytellers Podcast

Returning a year after his first appearance, Nurtekin Savas draws on his expanded risk-infrastructure remit at PayPal to examine the design of agentic workflows, enterprise foundation models, and the data and organizational conditions needed to put them to use.

Nurtekin Savas, Head of Global Risk & Servicing Infrastructure, Platforms, Data & AI (VP) at PayPal, joins Laszlo Mucsi to examine how enterprise AI can become part of the way decisions are made. His emphasis is on deliberate design: choosing valuable workflows, matching models and tools to each step, and setting boundaries on what agents can do.

We also explore how enterprise foundation models can help establish a shared customer representation that supports many decisions through task-specific prediction heads. Throughout the conversation, Nurtekin explains the potential to simplify pipelines and speed deployment, alongside the talent, integrated data, and changes in data-science work needed to make that architecture useful.

The conversation covers:
- Agentic workflows with step-specific models, tools, and risk controls
- Agentic operating systems for customer intent, routing, and permissions
- Shared customer representations, prediction heads, and simpler model pipelines
- The talent, data integration, and role changes required for adoption

This episode is part of Enterprise AI: State of Play, a practitioner-led interview series on what enterprise AI looks like on the front lines.

Connect with us:
Website: https://thedatastorytellers.com/
LinkedIn: https://www.linkedin.com/company/the-data-storytellers
Apple Podcasts: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476
Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa
YouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:
00:00 Enterprise AI: A Year in Review
08:39 AGI, Model Progress and Token ROI
14:08 From Models to Enterprise Intelligence
20:53 Agents, Workflows and the Agentic OS
28:46 Data and AI Platforms at PayPal
32:19 How Enterprise Foundation Models Work
42:34 Adoption, Explainability and Culture
54:18 Benevolent Bandwidth: AI for Good

Episode description

Taylor Black returns for a mid-2026 scorecard on the predictions he and Laszlo Mucsi made in their first conversation. They examine the shift from model breakthroughs to agentic systems that perform real work: coding tools finding product-market fit, autonomous software, small teams with outsized leverage, and the application layers where enterprise value is accumulating.

Enterprise AI is moving beyond the contest to build the most capable model. The advantage now lies in the systems around it: the workflows, data, evaluations, deployment infrastructure, and human judgment that turn model capability into business outcomes. Agents can execute, but they do not carry trust, liability, or responsibility; AI-native operators must redesign the work rather than simply automate the old process. The central business question is no longer who has access to the best model, but who can turnmodel capability into proprietary workflows, faster execution, and measurable operating leverage.

Connect with us:

Website:⁠ https://thedatastorytellers.com/⁠

LinkedIn:⁠ https://www.linkedin.com/company/the-data-storytellers⁠

Apple Podcast:⁠ https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476⁠

Spotify:⁠ https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa⁠

YouTube:⁠ https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:

00:00: Introduction and revisiting the 2025 AI predictions

01:39: Taylor’s work at Microsoft, the Vatican, and stepping away to reflect

04:48: Recapping part one and setting up the 2026 AI prediction scorecard

07:32: Has AI started delivering real economic value?

11:18: AI coding, developer productivity, and the rise of “dark factories”

14:24: The one-person billion-dollar company and the limits of AI employees

19:30: Why AI’s value is shifting from models to applications and workflows

24:11: Is the AI bubble disappearing, or just changing shape?

29:52: AI, employment growth, data centers, and US reindustrialization

33:23: Open-source models and the changing enterprise AI landscape

34:19: The Vatican, the Catholic Church, and the moral questions around AI

40:33: Pope Leo’s perspective on AI, power, relationships, and human flourishing

46:10: How do we measure human flourishing, and when should kids use AI?

In this episode of The Data Storytellers Podcast, we speak with Albert Mangahas, a data and analytics leader now at Roo. Albert shares how his path from industrial engineering and SQL led him into data leadership roles across PwC, Green Dot, Facebook, Turo, and now Roo.

We explore how trust in data became a foundation of Albert’s career, how marketplaces like Turo balance supply, demand, safety, and user experience, and what it takes to scale data teams through rapid growth and crisis. Albert also discusses his time at Facebook, how Turo navigated the pandemic, and why the real opportunity with AI may be less about replacing people and more about scaling better business insight.

Connect with us:

Website: https://thedatastorytellers.com/

LinkedIn: https://www.linkedin.com/company/the-data-storytellers

Apple Podcast: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476

Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa

YouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:

00:00: Introduction and Albert’s first time in Vancouver

01:50: From computer science to industrial engineering

03:50: The SQL class that sparked Albert’s interest in data

06:04: PwC, auditing, and building trust in data

13:36: Moving into retail fintech at Green Dot

17:28: Mobile banking, underbanked customers, and Instant Pay

20:11: From BI reporting to analytics, finance, and FP&A

21:57: Joining Turo and taking on analytics, FP&A, and fundraising

26:43: The Turo business model and the case for peer-to-peer car sharing

34:39: How two-sided marketplaces balance supply and demand

40:14: Trust, safety, insurance, and managing marketplace risk

45:53: Facebook, Cambridge Analytica, and data transparency

48:42: Leadership, influence, and measuring complex platform problems

57:46: Why Albert returned to Turo

58:57: COVID, travel shutdowns, and wartime data leadership

1:06:15: Profitability, supply growth, and marketplace resilience

1:09:07: AI, trust, and the future of business decision-making

In this episode of The Data Storytellers Podcast, we speak with Dejan Mitkovski, a data and technology leader whose career has spanned Microsoft, Amazon, Google, FanDuel, and BetMGM. Dejan reflects on the last 25 years of digital transformation, from the early days of relational databases and BI tools to the rise of big data, data science, machine learning, and AI.

We explore how data leadership has evolved from reporting to prediction, why legacy organizations struggle to adopt new technologies, and how leaders can prove value through small, practical tests. Dejan also discusses the difference between traditional machine learning and AI, the importance of data quality and governance, and why curiosity, discomfort, and people leadership remain central to building successful data organizations.

Connect with us:

Website: https://thedatastorytellers.com/LinkedIn: https://www.linkedin.com/company/the-data-storytellersApple Podcast: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xaYouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:

00:00: Introduction and Dejan’s return to Vancouver

03:43: The 25-year arc of data, digital transformation, and AI

04:26: Early influences, science, and choosing a different career path

09:40: First exposure to analytics and the move into Microsoft

12:20: Relational databases, Hadoop, and the early digital era

16:07: Stitching data together, personalization, and the rise of BI

20:38: Big data, data science, and the shift from reporting to prediction

29:35: Amazon, mentorship, forecasting, and thinking like an owner

37:02: Google, legacy industries, and the challenge of changing how companies work

42:35: Proving value through small tests and cultural adoption

50:45: Machine learning, measurement, and what makes AI different

59:41: FanDuel, data products, governance, and scaling a data organization

1:10:51: ChatGPT, generative AI, and the limits of messy data

1:18:18: BetMGM, AI enablement, and lessons for data leaders

1:21:58: Curiosity, discomfort, and the human side of leadership

Enterprise AI adoption is accelerating, but most organizations are still unprepared for the security, governance, and operational challenges that come with deploying generative AI at scale.

In this episode of The Data Storytellers Podcast, Laszlo sits down with Engin Bozdag and Stefano Bennati, authors of the book AI Governance, to explore why many GenAI initiatives fail once they move beyond the demo phase. They break down the biggest risks organizations face today, from prompt injection and data leakage to vendor over-reliance and weak governance processes. The conversation also introduces their six-level AI governance framework and explains how leaders can move from high-level principles to practical controls that actually protect enterprise systems and data.

AI Governance book: https://www.manning.com/books/ai-governance

45% off the book with no exp date or format restrictions: TDSTbozdag

Engin's affiliate tracking link (specifically for this book): https://hubs.la/Q03_5wmG0

Stefano's affiliate tracking link (specifically for this book): https://hubs.la/Q03_5vLc0

Connect with us:

Website: https://thedatastorytellers.com/

LinkedIn: https://www.linkedin.com/company/the-data-storytellers

Apple Podcast: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476

Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa

YouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:

00:00 – Introductions and the enterprise AI adoption moment
05:00 – Why GenAI deployments fail in production
12:00 – The “demo in production” problem
20:20 – The gap between AI governance principles and implementation
27:20 – AI adoption models and governance trade-offs
36:30 – Build vs buy in enterprise AI systems
40:40 – The six-level GenAI governance framework
52:20 – Why traditional security is not enough for AI
1:00:30 – Privacy risks in generative AI systems
1:10:00 – What trustworthy AI actually means
1:15:30 – A practical checklist for AI leaders deploying GenAI

What separates analytics teams that get a seat at the table from those stuck in the reporting queue?

Anthony Jackel, Senior Director of Business Intelligence & Analytics at Ferrara, has spent his career answering that question, from Kraft to one of America's largest candy companies.

In this conversation, we explore:

• What "data-driven enterprise" actually means
• The shift from data providers to decision enablers
• Why user-centered design matters as much for dashboards as it does for iPhones
• How to build analytics products that drive action, not just curiosity
• The mindset shift that earns analytics a seat at the table

Connect with us:

Website: https://thedatastorytellers.com/

LinkedIn: https://www.linkedin.com/company/the-data-storytellers

Apple Podcast: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476

Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa

YouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters:

00:00 – Introduction to Anthony Jackel and his role at Ferrara
03:35 – Building Ferrara’s analytics team from scratch
07:42 – The problem with dashboards and low-value data work
11:50 – Connecting analytics to business decisions and revenue
16:25 – The importance of commercial empathy for data leaders
20:18 – Getting out of the service provider mindset
25:04 – Translating technical insights for non-technical audiences
28:45 – Why building trust is more important than being right
33:29 – Lessons from finance that shaped his analytics approach
37:12 – Coaching and developing high-performing analytics talent
41:05 – The challenge of balancing curiosity with execution
45:18 – Internal marketing and the power of repeatable wins
49:56 – Final advice for future analytics leaders

📍 Chapter Timestamps (Finalized for 50:13 runtime)

What happens when a technology starts to look, sound, and reason like us?

Taylor Black, Director of the AI Ventures Ecosystem in the CTO’s Office at Microsoft, has spent his career at the intersection of data, venture building, and emerging technology. From bootstrapping a SaaS company to helping shape early stage AI innovation inside one of the world’s largest technology organizations, Taylor brings a rare long term perspective on what AI is, and what it is not.

In this conversation, we explore how today’s AI moment compares to past technological revolutions, why most current AI applications are still low hanging fruit, and where the real economic and organizational disruption is likely to come from next.

We discuss the limits of automation, the rise of agentic systems, the moral responsibility of those building AI, and why the hardest work ahead is not technical, but cultural.

This is a wide ranging discussion on AI, human work, governance, and the difference between making hard things easier versus making impossible things merely difficult.

Connect with us:

Website: https://thedatastorytellers.com/
LinkedIn: https://www.linkedin.com/company/the-data-storytellers
Apple Podcasts: https://podcasts.apple.com/gb/podcast/the-data-storytellers-podcast/id1493766476
Spotify: https://open.spotify.com/show/2N0vZtHZHgod4Tll2LX2xa
YouTube: https://www.youtube.com/channel/UCz9e56lhYUfORiOHMiLlPmA

Chapters

00:00 – Introducing Taylor Black and his role at Microsoft
04:00 – From bootstrapped SaaS founder to data driven decision making
10:30 – Learning to translate technical insight into business value
16:30 – Venture studios, zero to one work, and entering Microsoft
23:30 – How ChatGPT changed perception inside large organizations
28:00 – The steam engine analogy and why infrastructure matters
33:30 – AI, work, and why modern jobs already felt broken
40:00 – Presence, remote work, and the limits of digital collaboration
46:00 – AI optimism, AI fear, and misunderstanding the technology
52:00 – What large language models actually are and are not
58:30 – Governance, moral responsibility, and who shapes AI
1:05:30 – Agentic systems, job disruption, and what comes next

In this episode of The Data Storytellers Podcast, we sit down with Ylan Kazi, Chief Data & AI Officer at Blue Cross Blue Shield of North Dakota, to cut through the noise around enterprise AI. Ylan explains why early machine learning hype never matched operational reality, how LLMs changed the game once they became accessible to anyone, and why most AI failures stem from cultural friction rather than technical limitations. He breaks down the gap between expectations and actual ROI, the overlooked complexity behind agentic AI, and the fundamental difference between inserting AI into old workflows and redesigning processes to be AI native.

We also explore the economic and societal contours of the current AI cycle, including energy constraints, the cultural backlash against AI generated content, and why exponential progress looks slow until it doesn’t. Ylan shares what the coming micro bubbles might look like, how labor markets are shifting, and why new technology forces each of us to examine the meaning of human work. We end with a look at his AI Edge newsletter and what he is tracking as this transformation accelerates.

Chapters:

00:00 Introductions and Ylan’s early AI skepticism
04:30 Traditional AI vs generative AI and why prompting skill matters
08:00 Why enterprise ROI lags and why most failures are cultural
12:00 The economics of LLMs, energy constraints, and infrastructure realities
18:00 Agentic AI, undocumented processes, and why value requires redesign
25:00 Automation, talent pipelines, and shifts in labor markets
34:00 Cultural reactions to AI content and the value of human creation
45:00 The case for micro bubbles and faster boom bust cycles
57:00 Ylan’s newsletter, what he is writing about, and closing thoughts

In this episode of The Data Storytellers Podcast, we sit down with Ron Hetrick, Principal Economist at Lightcast, to unpack the deep structural issues driving the U.S. labor shortage. Ron explains why this is not a short-term blip but a long-term demographic shift that will redefine how companies hire, grow, and operate. From fertility rate collapse and immigration policy to flawed assumptions about productivity, he helps us understand the real economic forces reshaping the talent landscape.

We explore why AI and automation are not the silver bullets many executives hope for, and what leaders should focus on instead. Ron also shares what economists missed about remote work, the dangers of oversimplified labor data, and how companies can rethink workforce strategy in an age of persistent scarcity.

Chapters:

00:00 – Introductions and Ron’s career path from TV to economics
03:12 – What’s really driving the U.S. labor shortage
07:21 – Why this isn’t a temporary workforce dip
11:44 – The demographic math no one is talking about
15:32 – Immigration, fertility, and the myth of the "lazy worker"
20:10 – Why unemployment stats don’t tell the full story
25:33 – The silent shift from job scarcity to talent scarcity
30:48 – What economists missed about remote work and productivity
36:15 – The false hope of automation as a labor solution
41:09 – How companies are overestimating what AI can replace
46:27 – The hard ceiling of productivity without people
51:12 – Forecasting the long-term consequences of talent shortages
56:33 – What future-ready organizations are doing differently
1:01:44 – Redesigning workforce strategy for a shrinking labor pool
1:06:28 – Advice for leaders on adapting to the new labor reality
1:10:13 – Final thoughts on demographic truth vs tech hype

On this episode of The Data Storytellers Podcast, Rich Heimann, author of Doing AI, GenAI Revealed, and the newly released Sutskever's List, joins us for a deep and provocative conversation on why most enterprise AI initiatives fail.

Rich unpacks how internal power structures, belief systems, and corporate incentives shape what AI gets funded, how it's framed, and why so many efforts fall short. We explore the gap between organizational theater and technical progress, and why understanding human systems matters more than mastering the technical stack.

If you're an AI or data leader struggling with implementation, this episode offers a fresh lens that might explain why the problem isn't your model.

Want more? Rich’s new book Sutskever’s List: Foundational Ideas of Modern AI is now available — and you can get it HERE with 45% off.

Chapters:

00:00 – Introductions and Rich’s background
02:44 – The real reasons AI adoption fails in enterprise settings
07:12 – Power, status, and the illusion of AI progress
12:25 – Why many AI initiatives are more about performance than impact
17:31 – Data science vs business theater: who’s really calling the shots?
21:50 – Cultural friction: how belief systems shape technical outcomes
26:43 – Rich on the danger of over-rationalizing human systems
32:17 – Lessons from behavioral economics and AI ethics
37:09 – What leaders pretend to want vs what they actually support
42:03 – Redefining AI success: who gets to decide what matters?
47:58 – Stories vs truth: how narratives become strategy
53:11 – The objectivity myth and why some orgs cling to it
59:04 – LLMs, RAGs, and the current moment of hype
1:04:50 – Why technologists need to understand organizational design
1:09:44 – Advice for AI leaders trying to make change stick
1:14:02 – Final reflections on trust, transformation, and learning