The Data Playbook Podcast

Tom Delesie, data analytics manager at Atlas Copco, joins host Kris Peeters to talk about building a common data platform for a global, decentralised company. His team brought several platforms together into one Intelligence Hub, giving every team the same foundations while leaving ownership of data products with the business.

Tom explains how templated rooms give teams a shared way of working, why some work is deliberately kept out of the reusable layer, how transparent costs and open visibility encourage reuse, and why a clearly written purpose matters as much as good engineering. He also shares how the platform is getting ready for AI agents.

Resources shared by Tom:

🎥 Enabling the things that shape our lives, Atlas Copco Group: https://www.youtube.com/watch?v=sk3QfMgPyZM

🏭 Atlas Copco: https://www.atlascopco.com

💼 Jobs at Atlas Copco: https://www.jobs.atlascopcogroup.com

🌐 More at www.dataminded.com

Chapters:

0:00 Introduction

0:37 Atlas Copco and its four business areas

3:09 Tom's role: platform and use cases5:46 From five data platforms to one Intelligence Hub

7:30 Decommissioning legacy platforms

9:11 Aligning platform leaders across business areas

11:05 Balancing freedom and ownership

14:57 Transparent costs and an internal data economy

18:13 Reuse, access requests and sensitive data

22:05 Shared and private data products

24:06 The data product room

29:39 Deprecating data products

30:01 Why a clear purpose drives adoption

33:10 Operational entities and consumption rooms

36:26 Self-service and enablement teams

40:20 AI agents as new platform users

43:01 Dashboards, semantic layers and KPI definitions

47:30 Freedom first, convergence later

Guest: Tom Delesie, Data Analytics Manager, Atlas CopcoHost: Kris Peeters, CEO, Dataminded

Raghid Bsat, data platform engineer at Dataminded, joins the Data Playbook podcast to explain why a data platform migration is 75% analysis and 25% execution. With host Kris Peeters, CEO of Dataminded, he walks through the five phases in his migration guidebook: vision, governance, discovery, execution and cutover.

They cover how to win over business stakeholders who see no reason to change a working system, when to lift and shift and when to refactor, and how migrator and validator AI agents now handle much of the execution. Raghid also shares what makes the next migration easier: tests, documentation and vendor-agnostic tooling.

📘 Read the Migration Guide: https://migrations.cloudrun.datamaturitytest.com/

Chapters:0:00 Introduction1:03 Why engineers dread migrations2:47 Why companies migrate: cost, licences and sunset products4:08 The payoff: less maintenance on the new platform5:44 The migration guidebook and its five phases7:22 Tier one and tier two migrations8:02 Phase 1: vision and success criteria11:34 Phase 2: governance and stakeholder management13:06 Lift and shift or refactor18:08 Handling new feature requests during a migration20:55 Replicate the legacy bug or fix it23:42 Phase 3: discovery and the source-to-target dependency map29:22 When the source systems are deprecated too32:31 Phase 4: execution with AI agents35:54 Validation and faster feedback loops38:17 Phase 5: cutover and decommissioning41:53 How to reduce future migrations45:20 Data problems are social problems

🌐 More at ⁠⁠www.dataminded.com⁠⁠

Andrew Jones, staff data engineer at LocalStack, returns to the Data Playbook podcast to make the case that giving business teams raw access to company data was never enough. The real fix is data contracts, tooling built around how producer teams already work, and platforms with guardrails baked in rather than bolted on through review.

Jones and host Kris Peeters, CEO of Dataminded, cover a data platform that failed because it forced producers into unfamiliar tooling, and the one that replaced it by meeting those same teams where they already worked. They also dig into where self-service stops, why so many data teams hold onto being the bottleneck, and what happens to data semantics once AI agents start asking the questions people used to ask. Jones tells the story of an internal chatbot that gave him a feature-usage number he knew, on instinct, had to be wrong.

🌐 More at www.dataminded.com

Simon Harrer, CEO and co-founder of Entropy Data, returns to the Data Playbook podcast to make a case every data leader will need to reckon with soon: AI agents need access to company data, and the access controls built for humans do not hold once an agent can request, use, and reuse that access without anyone reviewing it.

Simon Harrer and host Kris Peeters, CEO of Dataminded, unpack the difference between data contracts and data products, how purpose-based access control catches an agent that drifts from the purpose it was granted, and what a semantic layer does for a company's data model, connecting an order ID in one system to the same field under a different name in another. They also get into data lineage, the return of data mesh now that agents need a foundation to work from, and the data product builder Simon Harrer's team runs on Claude Code.

The conversation draws on the BARC research spotlight Entropy Data sponsored, "A Data Marketplace Is What Your Agents Need," written by BARC analyst Florian Bigelmaier.

🌐 More at www.dataminded.com

Many organisations embraced Data Mesh. Many also discovered that organisational change is much harder than technology change.

In this episode, Kris Peeters speaks with Arif Wider about what remained valuable after the initial hype around Data Mesh settled.

They discuss why Data Products have become the practical foundation for modern data organisations, where Data Contracts fit into the picture, and how organisations can modernise legacy platforms without starting over.

The conversation also looks ahead at AI-assisted software engineering, cross-company data sharing, and the skills future data teams will need.

Topics include:

• Data Mesh in practice
• Data Products
• Data Contracts
• Data Architecture
• Legacy data warehouses
• AI agents in software engineering
• The future of data engineering

🌐 More at https://www.dataminded.com/

Connect with Arif Wider

• Website: http://arifwider.com
• ResearchGate: https://www.researchgate.net/profile/Arif-Wider
• Book - Data Mesh in Action: https://www.oreilly.com/library/view/data-mesh-in/9781098108502/

Very few companies manage 10,000 of anything today, and Eric Broda thinks that's the real challenge behind AI agents at scale. Broda runs Broda Group Software, helping banks and insurers build agent ecosystems he calls the agentic mesh.

With host Kris Peeters, CEO of Dataminded, Broda explains why agents need a data mesh foundation, why one client tracks R&D tax credit exposure by watching code commits in real time, and why his team moved from fine-grained knowledge graphs to coarse markdown files after RAG broke down at scale.

Find Eric Broda's work at agenticmesh.substack.com and on LinkedIn and YouTube under The Agentic Mesh Podcast.

🌐 More at www.dataminded.com

Modern data platforms are becoming more powerful. They are also becoming more complex.

In this episode of The Data Playbook, Kris Peeters talks with Wannes Rosiers, Head of IT & Development at WEAREONE.world, the company behind Tomorrowland, about why many organisations solve familiar data problems with increasingly complicated technology.

They explore what actually matters when building data platforms: delivering quickly, keeping systems maintainable, bringing business knowledge closer to engineering teams, and preparing for AI without rebuilding the entire technology stack.

The conversation covers:

• Data platform architecture
• Data Mesh
• Data engineering
• AI adoption
• Platform simplicity
• Data leadership
• Enterprise architecture

🌐 More at https://www.dataminded.com/

Building advanced analytics is rarely a technology challenge alone. Success depends on understanding the organisation, delivering value incrementally and enabling teams to solve real operational problems.

In this episode, Kris Peeters speaks with Max Hubeau, Manager of the Advanced Analytics Team at NMBS/SNCB.

They discuss how a central analytics team works alongside business teams, how data science supports train punctuality, passenger counting and energy forecasting, and why AI becomes valuable when it fits existing operations instead of disrupting them.

The conversation also explores data platforms, AI-assisted development, and practical lessons for CIOs, CDOs, Heads of Data and analytics leaders building data capabilities inside large organisations.

🌐 More at https://www.dataminded.com/

Follow Dataminded for more conversations with data leaders shaping the future of Data & AI.

Agentic AI is everywhere. Production-ready AI is not.

In this episode, Kris Peeters talks with Dataminded Data Engineer Jesus Garcia about the gap between AI hype and enterprise reality.

Drawing on lessons from building a large-scale Agentic AI platform, they discuss security, governance, user adoption, change management, AI skills, and the future role of engineers in an AI-first world.

For data leaders, CIOs, CDOs and technology executives, this conversation provides practical insights into what it takes to deploy AI systems that people trust and actually use.

🌐 More at https://www.dataminded.com/

Topics covered: Agentic AI, Enterprise AI, RAG, AI Security, AI Governance, Change Management, Data Leadership, AI Adoption, Data Engineering.

Chapters:

00:00 Agentic AI in Production: Introduction

01:34 Enterprise Knowledge Management with AI Agents

05:19 RAG, Vector Search and Agentic AI Explained

07:36 Deploying Agentic AI at Enterprise Scale

12:14 AI Governance, Security and Access Control

17:06 What Data Leaders Can Learn from AI Security Incidents

25:34 Will AI Replace Software Engineers?

40:20 Why Coding Is Easier for AI Than Business Decisions

43:36 Real Productivity Gains from AI Skills and Automation

49:35 The Limits of Large Language Models

56:44 Change Management: The Real Challenge of AI Adoption

Agentic AI is changing how we build software, manage data, conduct research, and learn new skills.

But as AI takes over more cognitive tasks, a fundamental question emerges: what capabilities do humans still need to develop themselves?

In this episode, Kris Peeters sits down with Frank Neven, Professor of Computer Science at Hasselt University and Vice Director of the Data Science Institute, to discuss the future of data engineering, AI-assisted learning, database systems, and human-AI collaboration.

The conversation explores:

• Why understanding remains essential in an AI-driven world
• How universities are adapting to AI-powered education
• What the latest database research tells us about the future of data platforms
• The rise of agentic coding and AI-native software development
• How AI is transforming scientific research
• Why structured knowledge and semantic data are becoming more valuable

This episode is particularly relevant for data leaders, CIOs, CDOs, architects, and engineering teams navigating the rapid evolution of AI.

🌐 More at https://www.dataminded.com/

#DataEngineering #AgenticAI #DataLeadership #ArtificialIntelligence

Chapters:

00:00 From Theory to Data Engineering

03:20 AI and Healthcare Data Integration

06:00 The Data Science Institute in Practice

19:00 Why Computer Science Education Matters

25:40 AI in Education: Opportunity and Risk

35:00 You Can Outsource Reasoning, Not Understanding

38:45 Agentic AI and the Future of Databases

45:50 How AI Is Changing Research

51:45 Personal AI Systems and Knowledge Management

59:30 Staying Open-Minded in Technology