The Databricks Diaries

In this episode of Databricks Diaries, Daniel Thornton sits down with Andreea Moldovan, Associate Director of Data Science at Sage Publishing, for a candid and wide-ranging conversation about what it actually takes to lead data science and AI in the real world.

Andreea's journey is anything but conventional — starting as a social scientist in Romania, she went on to complete a PhD funded by the Wellcome Trust, before moving through the Financial Conduct Authority, a specialist insurance company, and a price reporting agency, to ultimately find her home in academic publishing.

In this episode, they explore:

• Why a background in social science is one of the most underrated assets in data science and AI
• How to quantify and communicate uncertainty to non-technical stakeholders without losing their trust
• The mindset shift from individual contributor to data science leader — and why it's harder than it looks
• What "outcome-first" thinking really means in practice when building AI solutions
• How generative AI is changing the shape of data science roles — and why the fundamentals haven't changed as much as people think
• The one constant in a decade of working in data science: change

Whether you're an aspiring data leader, a seasoned practitioner, or simply curious about the human side of AI, this is a conversation packed with honest reflections, hard-won lessons, and a refreshing dose of realism about what the next few years hold.

In this episode of The Databricks Diaries, Andy Davis speaks with Sindy Yick, Head of Data Science and Machine Learning at Markerstudy Group, as part of our ongoing AI readiness series.

Sindy brings a valuable data science perspective to the conversation, exploring why AI readiness is not just about adopting the latest tools or building agents. Instead, it starts with strong data foundations, good system design and a clear understanding of where AI can genuinely add value.

The conversation also tackles one of the more difficult topics in AI adoption: the anxiety technical teams feel around AI agents and automation. Sindy shares her view on why AI should be treated as an assistant rather than a replacement, why junior talent still matters, and how organisations may need to rethink how they train and develop early-career data professionals.

In this episode, we cover:

• Why data quality and system foundations come before AI adoption
• The “garbage in, garbage out” risk when applying AI to poor-quality data
• How data science and machine learning teams are reacting to rapid AI change
• The impact of coding agents on junior data science and engineering roles
• Why AI is more likely to assist technical teams than fully replace them in the near term
• The importance of stakeholder engagement, business knowledge and industry context
• How organisations may need to rethink graduate and junior training pathways
• Why human judgement, communication and strategic thinking are becoming more valuable
• Potential AI use cases across insurance, including operations and underwriting
• The importance of guardrails when moving AI closer to customer-facing or front-end applications
• Sindy’s advice for data science leaders: stay open-minded and keep adapting

AI readiness is not just about adopting the latest model or building the next agent.

In this episode, Andy Davis speaks with Sindy Yick, Head of Data Science and Machine Learning at Markerstudy Group, about why strong data foundations, guardrails and human skills are critical to making AI work in practice.

They also explore the future of junior technical roles, the rise of coding agents, and why business knowledge, stakeholder engagement and adaptability may become the most important skills for data science teams in the years ahead.

In this episode of the Databricks Diaries, Daniel is joined by Kieran Pointon, Director of Data & Analytics at Bromford Flagship LiveWest — one of the largest housing providers in the UK with around 120,000 homes following a recent three-way merger.

Kieran brings 20+ years of experience across data engineering, BI, and enterprise architecture, including leadership roles at Halfords and Adidas. In this conversation, he opens up about the realities of merging three data teams, why he's betting on Microsoft Fabric over Databricks for his current platform, and how data science is helping real people get into work, education, and safer homes.

What's covered:

• Kieran's journey from headphones-on software engineer to data leader — and how being an introvert shaped his leadership style
• Why "listen before you speak" is his number one piece of leadership advice
• Building a high-performing data team at Halfords that could run without him
• Merging Bromford, Flagship, and LiveWest — three teams, three landscapes, three ERPs
• Why he chose Microsoft Fabric over Databricks for the new platform (and the five-year bet behind that decision)
• The "left brain vs right brain" model: where the technology approach and business approach meet in the middle
• Real data science use cases changing tenants' lives — condensation/damp/mould prediction, customer complaint modelling, and the "Pipeline of Talent" NLP model that's helped 450+ customers into work and education
• How to get exec and board buy-in for data and AI initiatives
• What makes the best analytics team — and why it's not always about the best tools or the best engineers
• Honest advice for any data leader facing a merger or acquisition

In this episode of the Databricks Diaries, Daniel is joined by James Blagg, a seasoned Chief Data Officer with over 25 years of experience leading data strategy, architecture, and modernisation programmes across financial services.

James shares hard-won lessons from a career that spans the 2008 financial crisis, the big data hype cycle, the move from Teradata appliances to cloud-native platforms, and now the AI wave. It's a candid conversation about what actually drives value in data — and the gap between vendor promises and operational reality.

What's covered:

• James's career journey from Oracle PL/SQL developer to CDO
• Leading BI teams at Lloyds through the financial crisis and the HBOS integration
• Nationwide's data simplification programme and the "almost cloud" Teradata investment
• How platform evaluation has changed — and why cost, skills, and relationships now matter more than features
• Why "it's on the roadmap" is the most dangerous phrase in vendor selection
• The MVP hypothesis approach to picking analytics use cases
• Why so many data warehouse projects are perceived as failures (and why metadata investment is finally fixing it)
• Where AI really sits on the maturity curve — and why financial services is 10–15 years away from full back-office adoption
• A practical playbook for governance, cost monitoring, and security guardrails before scaling AI on Databricks
• The truth about cloud consumption pricing and reserved pricing deals

In this episode of The Databricks Diaries, Andy Davis is joined by Claire Thompson, Chief Data Officer at Quilter, to explore what AI readiness really looks like inside a modern financial services organisation.

Claire shares her perspective on the pace of change in data and AI, why strong foundations still matter, and why organisations cannot afford to wait for everything to be perfect before starting. The conversation covers how to build momentum through practical use cases, engage senior stakeholders, create safe environments for experimentation, and make governance a helpful enabler rather than a blocker.

Andy and Claire also discuss the importance of continuous learning within data teams, how to prioritise AI opportunities, and why technical teams need to get better at telling the story of the value they create.

This episode is for data, technology and business leaders who want a grounded view of AI adoption beyond the hype, with practical lessons on moving from early experimentation to scalable, valuable outcomes.

In this episode of the Databricks Diaries, host Andy is joined by Gordon Logan, Director of Data Services at PRS for Music, for a deep dive into the unique challenges of preparing a century-old organisation for the AI era. Managing data for over 180,000 members—including composers, songwriters, and publishers—Gordon shares how his team navigates astronomical data volumes driven by the modern shift to streaming and digital consumption. He discusses the critical transition from isolated experiments to a structured, momentum-driven AI strategy backed by executive leadership.

Gordon explores the essential components of building an AI-ready culture, prioritising data literacy and "AI efficacy" to ensure the workforce uses powerful tools like Copilot ethically and responsibly. The conversation highlights a disciplined, business-led approach to use cases, moving away from "technology vanity exercises" and focusing instead on measurable outcomes like accurate royalty distribution and improved member services. Listeners will learn how PRS for Music employs a cross-functional AI team to stress-test hypotheses, ensuring that every technological advancement facilitates tangible business value while protecting the sensitive data of the UK’s creative community.

In this episode of The Databricks Diaries, host Andy sits down with Steve Green, a senior data leader who has navigated the high-stakes intersection of data and AI from both the regulatory halls of the FCA and the front lines of private banking. Steve strips away the persistent AI hype to offer a masterclass on building a truly "AI-ready" business within the rigorous constraints of the financial sector.

Steve explains why the "garbage in, garbage out" reality remains the biggest hurdle to success and how a horizontal architectural view must prioritise business outcomes over technology for its own sake. The conversation reveals how to align legal, compliance, and IT through ethical control frameworks that maintain trust in generative AI outputs while ensuring foundational data hygiene remains a non-negotiable priority. By moving beyond the simple automation of annoying tasks, Steve challenges leaders to re-engineer end-to-end business processes to capture true value.

In this episode of The Databricks Diaries, I sit down with Tom Heath, a data and AI leader who has spent the last 20+ years working across data, semantic technologies, machine learning, and organisational transformation.

Tom shares a really thoughtful perspective on how the conversation has shifted — from having to convince people AI mattered, to helping organisations work out what to actually do with it.

We talk about what it takes to lead AI inside large, complex businesses, why so many projects still fail, and why technical capability on its own is never enough.

This is a grounded conversation on strategy, leadership, operating models, and the real work required to turn AI into something meaningful.

1. Why the public arrival of ChatGPT changed executive conversations overnight
2. The link between data structure, semantics, and the future of intelligent agents
3. Why AI success depends on more than just technical execution
4. The difference between efficiency, quality, and innovation as AI value drivers
5. Why some AI projects succeed technically but still fail commercially
6. The importance of problem definition before touching the technology
7. What leaders should focus on when building AI capability inside large organisations

In this episode of The Databricks Diaries, I sit down with Chris Hounslow, Head of Data Science & AI at Lebara, to talk about what it actually takes to turn AI from hype into real business impact.

Chris shares the journey behind Lebara’s transformation — from building solid data foundations to rolling out production-grade AI tools across a lean team… and winning industry recognition along the way.

We get into the realities of operating with limited resource, choosing the right use cases, and why most AI projects fail after the proof of concept stage.

If you’re trying to scale AI in a real business (not just experiment with it), this one’s worth your time.

1. Why the shift from “data science” to “AI” is often just rebranding, not reinvention
2. How to run a high-performing, lean AI team (5 people, 80% success rate)
3. Why being problem-led (not tech-led) is the key to choosing the right use cases
4. The importance of end-to-end ownership (not just building models)
5. Why most AI projects fail after the POC stage
6. How to build trust and adoption with non-technical stakeholders
7. The role of data foundations as the real competitive advantage
8. Why testing and validation are the biggest blockers to scaling AI

In this episode of the Databricks Diaries, Andy sits down with Mike Leverington, Director of Data Analytics at Skyscanner, to discuss the practical realities of AI readiness in a rapidly evolving tech landscape. Mike shares why successful AI integration begins with a cultural mindset rather than just technical infrastructure, urging leaders to "lean in" and embrace experimentation even amidst market volatility and uncertainty.

The conversation explores the threefold impact of Generative AI: driving internal productivity, revolutionising the customer experience, and disrupting entire industries. Mike provides actionable advice for data leaders on securing board buy-in, leveraging "small lab" innovation teams, and why the current hype cycle presents a unique opportunity to finally master self-serve analytics. From role-modelling AI use in daily admin tasks to "unlearning" how to Google, this episode is a masterclass in driving the last mile of the data value chain