Leadership teams can explain how they make money, but not their data architecture says Claire Thompson, Chief Data Analytics Officer for the Corporate Bank at Barclays. She joins The AI Adoption Podcast and makes the case that AI adoption is a people and culture challenge, built on strong data foundations and AI governance set at the outset. Claire explains the difference between deterministic technology, which gives the same answer a hundred times, and probabilistic generative AI, which may answer the same question differently each time and may not answer it correctly. She sets out three routes by which AI enters an organisation and the governance each one needs. You will hear her advice for a leadership team that judges its own data maturity to be low: find out who around the table understands AI, data and technology, and get curious. Her warning on timing is blunt, governance is much easier at the beginning than once the horse has bolted. Chapters 00:00 AI as a People and Culture Challenge 05:06 Transforming Processes with AI 09:47 Successful AI Adoption Strategies 11:34 Addressing Job Security Concerns with AI 12:41 Human-Agentic Teams in the Workplace 13:35 Senior Management’s Role in AI Implementation 15:24 The Data Management Challenge 16:54 Moving from Local AI Use to Strategic Implementation 18:21 Governance in AI Adoption 23:45 Improving Data Maturity in Organisations 25:37 Cultural Implications of AI Adoption Key topics • AI adoption as a people and culture challenge • Data culture and data architecture are inextricably linked • AI governance for personal productivity tools, vendor applications and bespoke models • Probabilistic versus deterministic AI explained • Moving from disconnected AI pilots to strategic AI adoption • Leadership role modelling and psychological safety in AI adoption • Human and agentic AI teams in financial services About the guest Claire Thompson is Chief Data Analytics Officer for the Corporate Bank at Barclays. She has spent her career in data and analytics across several financial services organisations, each at a different stage of data maturity. She describes part of her job as being the translator who turns technical capability into business terms, which makes her a credible voice on the gap between leadership teams and their data.

Giving AI full access to a map of critical infrastructure is, in Dino Ravnić's words, "no way to go". On this podcast, the Co-founder and CEO of GIS Cloud sets out the control layer that AI, location data and data sovereignty demand. Dino explains the shift from GIS as a specialist's desktop tool to a system a field worker, planner or budget holder can question in plain language through AI. You will see the practical gain, with hydrant repairs ranked by proximity to schools and hospitals, and the risk that comes with it when infrastructure data reaches the wrong hands. Dino describes a separate permission for AI access, zero data retention policies and the case for keeping data within the EU. The conversation ends with citizens, and the public data on schools and outages that AI could put within their reach. Chapters 00:00 The Importance of Data Control in AI 04:41 Use Cases of GIS in Organisations 07:52 Integrating AI with GIS 09:50 AI vs. Internet of Things: Understanding the Difference 11:15 Democratisation of GIS: Making Data Accessible 13:54 Enhancing Decision-Making with GIS 17:09 Geopolitical Implications of GIS Data 18:16 Governance and Sovereignty in GIS Data 20:56 Balancing Democratisation and Control 24:40 The Impact of GIS on Citizens Key topics • AI and geographic information systems (GIS) for infrastructure • Location data for prioritising utility repairs • Data sovereignty and EU data residency • AI governance and permissions on shared data • Shadow AI and unsanctioned data uploads • Zero data retention policies with AI vendors • Open government data and citizen access through AI About the guest Dino Ravnić is Co-founder and CEO of GIS Cloud. Dino has more than 15 years of experience with mapping tools, bringing location data closer to utilities, government and energy organisations. GIS Cloud lets field teams, specialists and decision makers work on the same location data from any device, with controls over whether AI may read or edit that data.

AI agents laid over fragmented processes produce digital clutter, says Ericson Chan of Zurich Insurance Group. Part Two of this AI Adoption Podcast conversation covers agentic AI governance, ROAI and the future of work. Ericson sets out the difference between a chatbot that answers and an agent that acts, and the controls Zurich puts around each agent as a result. He explains deterministic and non-deterministic orchestration in plain terms, then turns to the board: AI governed as part of strategy and capital allocation and judged as a portfolio that leaves room for moonshots. The second half covers managers leading teams where half the members never sleep, roles that are rebundled rather than removed, and the risk to the way junior staff learn judgement. Part One covers the global picture and Zurich's AI360 strategy. Chapters 00:00 Holistic Value in AI Adoption 02:40 The Rise of Agentic AI 06:36 Understanding Deterministic vs Non-Deterministic Approaches 08:14 Gradual or Radical Transformation in AI Implementation 09:48 Governance and Strategic Planning for AI 11:43 Managing AI as a Portfolio 13:14 Collaboration Between Humans and AI Agents 16:07 Cultural Shifts in Organisations with AI 18:53 AI Literacy for the Next Generation 22:22 Balancing AI Automation and Learning Opportunities Key topics • Agentic AI governance and operating envelopes • Deterministic orchestration of multi agent workflows • Return on AI (ROAI) and portfolio management • Board oversight of AI strategy and capital allocation • Human and AI agent teams • AI literacy as a workplace capability • Early careers, apprenticeship and AI automation About the guest Ericson Chan is Group Chief Information and Digital Officer at Zurich Insurance Group. At Zurich he runs agentic AI hackathons that bring together hundreds of employees from 20 to 30 countries to build prototypes, and his teams orchestrate agents through deterministic workflows. That experience of deploying agents inside a global insurer gives him a practitioner's view of what adoption demands of boards, managers and junior staff.

AI is global, and the US, China and Europe are each building it for a different objective. On this AI Adoption Podcast episode, Ericson Chan of Zurich Insurance Group sets out a global AI strategy that holds across three regulatory regimes. Part One of a two-part conversation. Ericson has worked in the US, China and Europe, and he explains the consequence of their different priorities for an insurer operating in each. He sets out Zurich's STAR principles, the cultural roots of customer trust in AI, and the reason local teams, not a central function, must own AI implementation. The episode closes with Zurich's AI360 strategy and its three pillars. Key topics • Global AI strategy for multinational organisations • AI regulation in the EU, US and China • AI governance principles: Security, Transparency, Accountability, Reliability • Cultural attitudes to AI and customer trust • Centralised strategy with local AI implementation • End to end AI in insurance underwriting and claims • Small language models and agentic AI in financial services About the guest Ericson Chan is Group Chief Information and Digital Officer at Zurich Insurance Group and has been with the group for six years. He grew up in Hong Kong and spent a decade in the United States and a decade in China before moving to Europe. That experience across the three regions shaping AI gives him a view of global AI strategy from inside each system. Part Two, on adoption and implementation in detail, follows next week. Chapters 00:00 The Importance of Transparency in Leadership 03:16 Global Perspectives on AI Development 07:53 AI as a Threat and Opportunity 08:39 Challenges and Opportunities in Scaling AI 12:39 Cultural Influences on AI Adoption 14:44 The Need for a Unified AI Strategy 17:56 Coordination in Decentralised Implementation 19:23 Leadership Priorities in AI Adoption 23:57 Transforming Business with AI 26:52 Zurich's AI Strategy: AI 360

Proving AI works in one police force is not the hard bit. Dylan Alldridge of PoliceAI joins The AI Adoption Podcast on scaling AI in policing across 43 independent forces, public trust in police AI, and giving officers time back for victims. The Home Secretary has asked PoliceAI to make policing in England and Wales "the most AI enabled police force in the world", yet it can mandate very little of the 43 forces it serves. Dylan sets out the approach: pathfinder forces that trial and evidence a tool, a first wave of AI aimed at case files, disclosure and transcription, and a public AI registry that every force will maintain. He is open about the Home Office target of around six million hours, and about the risk that rising token costs and vendor lock in could make AI unaffordable for the public sector. You will come away with a clear view of adopting AI across organisations that cannot be ordered to comply. Key topics • AI in policing across 43 forces in England and Wales • Scaling AI pilots through pathfinder forces • AI for case files, disclosure schedules and transcription • Public trust and the police AI registry • Designing AI tools with the workforce, and job security • Vendor lock in and token cost risk in public sector AI • The Police Reform White Paper productivity targets About the guest Dylan Alldridge is Deputy Director for Strategy, Oversight and Coordination at PoliceAI, the National Centre for the Responsible Adoption of AI in Policing. His career spans policing and government, at the intersection of operational policing and capability, and he is a former intelligence analyst who has studied the role academically. His remit is the hardest part of AI adoption in policing: coordinating 43 independent forces towards consistent, trusted use of AI. Chapters 00:00 Introducing PoliceAI and the case for a national centre 05:13 Scaling AI across 43 independent forces 07:03 Pathfinder forces and the unglamorous paperwork 10:00 Culture, ease of use and the six million hours target 14:05 Roles, jobs and time back for officers 16:39 Data, governance and consistent standards 21:51 Public trust and the AI registry 25:02 The impact on victims and society 27:12 Acting early and the risk of unaffordable AI 31:18 The victim comes first

NatWest holds every AI system to the EU AI Act standard worldwide, even where local law does not demand it. Paul Dongha, Head of Responsible AI and AI Strategy at NatWest Group, on this AI Adoption Podcast episode speaks about AI governance, agentic AI and customer trust in banking. Episode summary Paul Dongha runs responsible AI at one of the UK's largest banks, and he treats governance as an operational discipline, not a policy document. In this episode you will learn why NatWest applies a single EU AI Act standard across every market, how an AI ethics panel settles hard trade-offs through reasoned debate, and why literacy training reaches branch staff who never use AI directly. Paul also explains risk tiering, guardrails and the extraterritorial reach of the EU AI Act. Key topics • AI governance frameworks in financial services • Agentic AI oversight and human-in-the-loop control • EU AI Act risk tiers and extraterritorial compliance • Risk tiering and risk assessment for AI systems • AI literacy and persona-based training • AI ethics panels and councils in banking • Customer trust as the goal of responsible AI About the guest Dr Paul Dongha is Head of Responsible AI and AI Strategy at NatWest Group, where he leads the bank's approach to deploying AI safely and setting its AI strategy. He holds a PhD in agentic AI and has co-chaired an AI ethics board in banking. His focus is translating AI governance into practices that protect customers and the trust they place in their bank. Chapters 00:00 The dual nature of AI: risks and benefits 02:00 Responsible AI in banking 03:09 Governance of agentic AI in financial services 08:29 Aligning with the EU AI Act 12:06 Risk assessment and categorisation in AI 15:16 Cross-jurisdictional governance challenges 17:30 The importance of trust in AI governance 20:19 AI literacy: building a knowledgeable workforce 22:33 Persona-based training for AI governance 24:46 The role of AI ethics councils 29:11 Balancing business needs and ethical governance

A database holds ‘what happened’, never the decisions behind it. Dhiraj Rajaram, founder and CEO of Mu Sigma, joins The AI Adoption Podcast on enterprise AI, decision sciences and agentic AI. Dhiraj explains why enterprise AI has not matched consumer AI, and what leaders can do about it. He introduces the field of context, built from action, decision and perception traces, and shows that databases capture only the first. He argues that the big D that matters are decisions, not data, and that problems should be seen as a connected constellation rather than separate stars. He closes with a physics analogy for context that should reframe leaders thinking about the enterprise AI adoption. Key topics ● Enterprise AI versus consumer AI ● Field of context in enterprise AI ● Action, decision and perception traces ● Decision sciences and the big D ● Problems as constellations, not stars ● Agentic AI readiness for boards ● Systems thinking for AI transformation About the guest Dhiraj Rajaram is the founder and CEO of Mu Sigma. The firm helps large Fortune 500 companies through decision sciences and AI, with the goal of preparing them for a world of algorithms. The name pairs mu, the expectations people hold, with sigma, the variance from those expectations, the two forces Dhiraj sees behind every decision. Chapters ● 00:00 Challenges in building enterprise-level AI business cases ● 04:45 Complexity in large organisations and its impact on AI ● 07:03 Problems as constellations, not silos ● 10:42 From databases to a field of context ● 12:05 Making decisions more visible with AI ● 18:24 Unpacking action, perception, and decision traces ● 20:57 The three I's: intention, intelligence, initiative ● 24:31 Real-world example: problem redefinition in organisations ● 32:28 Key takeaways and future steps for enterprise AI

AI could erase the competitive edge that made your organisation win. On The AI Adoption Podcast, Vilas Dhar, President of the Patrick J. McGovern Foundation, explains why leaders, not tools, decide the outcome of enterprise AI adoption and AI governance. Vilas observes that many firms ask the wrong first question about AI, starting with the technology rather than the future they want it to build. He argues for a strategy-first approach and says that AI checklists mislead boards. The reflex to cut headcount with AI misses the opportunities to prepare organisations for fundamental changes in economic structures that AI brings about. He points to a blood-supply organisation that rebuilt its operations around AI demand projections and overturned over 25 years of assumptions and practices. His challenge to leaders is to prepare the organisation to sustain their competitive position once AI gives everyone richer information. Key topics • Strategy-first AI adoption versus vendor-led tooling • Why AI readiness checklists measure the wrong thing • Reimagining supply chains with AI demand projection • Red teaming AI decisions in the boardroom • AI, competitive advantage and sector convergence • Public markets, private capital and quarterly-earnings pressure • Raising the level of human decision-making in agentic processes About the guest Vilas Dhar is President and Trustee of the Patrick J. McGovern Foundation. He has spent his career at the intersection of building technology at scale and the rule of law. He leads the Foundation as a civil society voice arguing for a more just and humane use of AI.

Human in the loop is the wrong way to think about AI, argues Kerri O'Neill, Chief People Officer at Ipsos and its Global AI Workforce Transformation Lead. A podcast conversation on AI adoption, organisational culture and agentic AI at work. In this episode of The AI Adoption Podcast, Kerri O'Neill makes the case that AI adoption is decided in the cultural layer, not the technology. I discussed a range of topics including shared purpose between people and AI agents, the case for shifting mindsets rather than values, and the fate of human connection as agents take on more of the daily work. Kerri sets out the "augmented Ipsos" vision and the principles of trust, truth and transparency behind it. Viewers get a leadership frame for adoption that puts culture first. Key topics ● Culture as the deciding factor in AI adoption ● Shared purpose between people and AI agents ● Mindsets versus values in culture change ● Agentic AI and organisational design ● Certainty, guardrails and the limits of agents ● Human connection, dialogue and workplace wellbeing ● The augmented Ipsos vision and AI adoption principles About the guest Kerri O'Neill is Chief People Officer at Ipsos UK and Ireland and its Global AI Workforce Transformation Lead, a remit spanning 20,000 people across 93 countries. She is the author, with Alex Bailey, of Rapid Reculturing, published by Wiley. Her book places organisational culture at the centre of how companies adopt AI, which makes her a credible voice on adoption as a leadership challenge.