Enterprise AI, from the inside

The short version

I'm Sven Romijn. I'm an enterprise AI consultant who writes practitioner guides on the AI capabilities shipping inside SAP, Microsoft, and Oracle platforms. My newsletter, Between the Hype, covers where enterprise systems and AI actually intersect — not the keynote version.

Before that, I spent years at Deloitte — first on SAP S/4HANA implementations, then increasingly in project management, executing cross-functional programmes where you're aligning finance, supply chain, and IT on a single system while nobody agrees on the chart of accounts. I also chaired the Deloitte works council, which gave me a completely different lens: how technology decisions — including AI adoption — look from the board level of a major professional services firm.

That combination is why I write the way I do. I've seen technology meet organisational reality from every angle — the implementation floor, the programme board, and the governance table. The gap between what a vendor demos and what your team actually adopts is where most AI initiatives fail, and it's the space I focus on.

How I got here

My career has been a thread of connecting technology to business outcomes, even when the path wasn't linear.

The common thread: I've always been drawn to the intersection of emerging technology and established industries. Not the frontier research — the messy, practical work of making new technology useful inside organisations that run on legacy systems and institutional habits.

Why enterprise AI, why now

Something shifted in 2025-2026 that most people haven't fully registered yet. The enterprise platforms that run the world's businesses — SAP, Microsoft, Oracle, ServiceNow — started shipping AI directly into their products. Not as add-ons. Not as pilots. As production features, bundled into existing licences.

SAP now has dozens of AI agents and 2,100+ Joule skills. Microsoft embedded Copilot across the entire Dynamics 365 suite. Oracle launched 50+ agentic applications in Fusion Cloud. And most enterprise teams don't even know these capabilities exist, let alone how to activate them.

That's the gap I'm focused on. Not "what will AI do someday" — but what can the AI you're already paying for do right now, and how do you actually turn it on without breaking things?

I started writing about this because I couldn't find anyone else covering it at the practitioner level. There's plenty of vendor marketing. Plenty of analyst reports. But almost nothing written by someone who's actually configured these systems, sat through the go-lives, and knows what it's like when the AI suggestion is wrong and your AP clerk has already posted the invoice.

What I'm working on now

My point of view

I believe most enterprises are overinvesting in building custom AI and underinvesting in activating what their vendors have already shipped. The companies getting the most value from AI right now aren't the ones with the biggest data science teams. They're the ones who figured out that their ERP already has AI capabilities they haven't turned on.

That doesn't mean custom AI is never the right call — it is, especially for cross-platform orchestration and company-specific workflows. But the default should be to start with what you have, not what you could build. Activate first, then build for the gaps.

I also believe that change management is the most underrated bottleneck in enterprise AI adoption. The technology is ready. The question is whether your people and processes are — and that's a harder problem than any model choice. Having sat on the governance side as works council chair at Deloitte, I've seen how AI decisions are evaluated at the board level of a major firm. That perspective — understanding how organisations govern and absorb technological change, not just how they implement it — informs everything I write.