DataFramed

AI agents are spreading across the data and AI industry, promising to automate everything from research to outreach. At the same time, teams are learning that these tools can hallucinate, leak data, or act in surprising ways. In day-to-day work, the challenge is deciding which tasks to hand off, what data to share, and how to keep the output trustworthy. Do your agents actually add value, or just add noise? Are they running in a secured, ring-fenced environment? How do you balance playful experimentation with critical checking when an agent confidently gets a key fact wrong?
Danielle leads go-to-market strategy at WNS, Capgemini's AI transformation services arm. Previously, Danielle was Chief Data Officer at American Express and Albertsons. She also write The Remix substack on technology trends, and is an Editorial Board Member for CDO Magazine.
In the episode, Richie and Danielle explore AI agents at work, experimentation with guardrails, data privacy, access, tone controls, OpenClaw automation wins and failures, token costs, tying AI plans to P&L strategy, shifts in careers and hiring, how data teams handle unstructured data governance, and much more.
Links Mentioned in the Show:
WNS
Connect with Danielle
AI-Native Course: Intro to AI for Work
Catch Danielle speaking at RADAR—April 1
Related Episode: AI Agents Are the New Shadow IT (And Your Governance Isn’t Ready) with Stijn Christiaens, CEO at Collibra
Explore AI-Native Learning on DataCamp

New to DataCamp?
Learn on the go using the DataCamp mobile app

Empower your business with world-class data and AI skills with DataCamp for business

World models are emerging as the next step after large language models, pushing AI from book knowledge toward systems that can simulate the physical and social world. Instead of just generating text or short videos, the goal is steerable simulation with long-horizon consistency and planning. For practitioners, this raises practical choices: what data and representations do you need, and when do you mix symbolic reasoning with generative models? How do you test whether a model can follow actions over minutes, not seconds? And where do you start—robotics, driving safety, or synthetic data generation?
Professor Eric Xing is President of Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) and a world-leading computer scientist whose work spans statistical machine learning, distributed systems, computational biology, and healthcare AI. A fellow of AAAI, IEEE, and the American Statistical Association, he has authored over 400 research papers cited more than 44,000 times.Before MBZUAI, Eric was a Professor of Computer Science at Carnegie Mellon University, where he also founded the Center for Machine Learning and Health. He is the founder and chief scientist of Petuum Inc., recognized as a World Economic Forum Technology Pioneer, and has held visiting roles at Stanford and Facebook. He holds PhDs in both Molecular Biology and Computer Science.
In the episode, Richie and Eric explore world models as simulators for action, the jump from book intelligence to physical and social skills, why long-horizon planning is still hard, architectures, robots, data generation, open K2 Think LLMs, virtual-cell biology, and much more.
Links Mentioned in the Show:
MBZUAI
Pan World Model
Connect with Eric
AI-Native Course: Intro to AI for Work
Related Episode: Developing Better Predictive Models with Graph Transformers with Jure Leskovec, Pioneer of Graph Transformers, Professor at Stanford
Explore AI-Native Learning on DataCamp

New to DataCamp?
Learn on the go using the DataCamp mobile app

Empower your business with world-class data and AI skills with DataCamp for business

Across the AI industry, high-stakes tools are being deployed in places where errors can harm people: sepsis alerts in hospitals, identity checks, welfare fraud detection, immigration enforcement, and recommendation systems that shape life outcomes. The pattern is familiar: scale and speed go up, while human review becomes rushed, shallow, or punished for disagreeing. In daily work, that can look like a nurse forced to act on false alarms, or a team using an LLM summary in ways the designers never planned. When should you slow down deployment? How do you detect new “wild” use cases early? And what does responsible tracking and oversight look like under real pressure?
Atay Kozlovski is a Postdoctoral Researcher at the University of Zurich’s Center for Ethics. He holds a PhD in Philosophy from the University of Zurich, an MA in PPE from the University of Bern, and a BA from Tel Aviv University. His current research focuses on normative ethics, hard choices, and the ethics of AI.
In the episode, Richie and Atay explore why AI failures keep happening, from automation bias to opaque targeting and hiring models. They unpack “meaningful human control,” accountability, and design in healthcare, government, and warfare. You’ll also hear about deepfakes, consent, digital twins, and AI-driven civic engagement, and much more.
Links Mentioned in the Show:
“Lavender” IDF recommendation system
Amnesty International reports on AI/automation in welfare systems
“Meaningful Human Control” (MHC) framework
Connect with Atay
AI-Native Course: Intro to AI for Work
Related Episode: Harnessing AI to Help Humanity with Sandy Pentland, HAI Fellow at Stanford
Explore AI-Native Learning on DataCamp

New to DataCamp?
Learn on the go using the DataCamp mobile app

Empower your business with world-class data and AI skills with DataCamp for business

Data governance has been around long enough to develop playbooks, but AI governance is evolving in real time. Industry trends like LLMs, agents, and emerging “swarms” are changing what oversight even means, from data lineage to agent-to-agent provenance.
For working teams, the questions are immediate: who leads—legal, security, IT, data, or a new AI role? How do you set standards so engineers aren’t using a different tool for every task? What maturity framework should you measure against, and how often should you reassess as technology shifts? How do you help teams move fast without breaking trust?
Stijn is a data governance veteran and one of the leading thinkers in the space. He runs data strategy, data infrastructure, and product evangelism at the data and AI governance company Collibra. Since founding Collibra 18 years ago, Stijn has held several executive positions, including COO and CTO.
In the episode, Richie and Stijn explore AI governance failures and wins, risks from agents that can act on systems, creating visibility with an agent registry, how AI governance differs from data governance, ownership across legal, security, IT, and data teams, EU AI Act risk tiers, and much more.
Links Mentioned in the Show:
Collibra
Connect with Stijn
AI-Native Course: Intro to AI for Work
Related Episode: The New Paradigm for Enterprise AI Governance with Blake Brannon, Chief Innovation Officer at OneTrust
Explore AI-Native Learning on DataCamp

New to DataCamp?
Learn on the go using the DataCamp mobile app

Empower your business with world-class data and AI skills with DataCamp for business

Automation is moving from APIs to full “computer use,” where agents click through screens like a human. That power is transforming evidence collection, access reviews, and repetitive security tasks, but it also raises new risk. In everyday workflows, the safest gains often start with read-only actions, sandboxes, and clear opt-in for anything that writes changes. Do your tools know when an access request is an anomaly? Can you keep humans in the loop with fast review-and-approve steps? And if an agent can browse your systems, how do you stop data from walking out the door before customers or attackers notice?
Jeremy Epling is Chief Product Officer at Vanta, where he leads product strategy and execution for the company’s trust management platform. He focuses on helping organizations automate security and compliance, enabling them to build and scale with confidence.
Previously, he was VP of Product at GitHub, overseeing Actions, Codespaces, npm, and Packages—core components of the modern developer workflow used by millions worldwide. Before GitHub, Jeremy spent more than 16 years at Microsoft, leading product teams across Azure DevOps Pipelines and Repos, OneDrive, Outlook, Windows, and Internet Explorer. His work has centered on developer platforms, cloud infrastructure, and productivity tools at global scale.
In the episode, Richie and Jeremy Epling explore AI-driven security risks, vendor data use and trade-secret leakage, governance and access controls, compliance beyond audits, how agents automate security questionnaires and vendor reviews, how to ship faster safely, human-in-the-loop design, and “computer use” automation, and much more.
Links Mentioned in the Show:
Vanta
Vanta State of Trust Report
Connect with Jeremy
AI-Native Course: Intro to AI for Work
Related Episode: Governing Pandora's Box: Managing AI Risks with Andrea Bonime-Blanc, CEO at GEC Risk Advisory
Explore AI-Native Learning on DataCamp

New to DataCamp?
Learn on the go using the DataCamp mobile app
Empower your business with world-class data and AI skills with DataCamp for business