What I took away from Data Innovation Summit MEA 2025

On 20 May I spent the day at the Data Innovation Summit MEA at Conrad Dubai — a room full of data and AI leaders from across the Middle East and Africa. I went as a delegate, notebook open, mostly to pressure-test what I'm building against what everyone else is wrestling with. Here's what stuck.

Conferences are easy to be cynical about. You can usually predict the buzzwords from the agenda. But the value of a good one isn't the keynotes — it's the calibration. You find out whether the problems keeping you up at night are your problems or everyone's problems. This year, almost everything pointed at the same shift.

The theme nobody said out loud: pilots are over

A year or two ago, the energy in any data event was "what could GenAI do?" This time the mood had moved on. The interesting conversations weren't about whether to use AI — that's settled — but about the unglamorous middle: getting a promising demo to survive contact with production, security, and a finance review. The gap between a working prototype and a system the business actually depends on is wide, and a lot of teams are standing in it.

The hard part isn't the model. It's the two arrows.

Foundations beat models

The most grounded sessions kept circling back to something deeply unsexy: your data foundations decide your AI ceiling. You can bolt the best model in the world onto messy, ungoverned, undocumented data and you'll get confident nonsense, faster. Several leaders said versions of the same thing — that their "AI strategy" turned out, on inspection, to be a data-quality and data-governance strategy wearing a nicer jacket.

Everyone wants an AI strategy. What most of us actually need first is a boring, well-run data foundation. The AI is the easy part.

That landed for me, because it's exactly what I see building reporting platforms: the model is rarely the bottleneck. The bottleneck is whether the numbers feeding it are clean, defined, and trusted. Get that right and AI is leverage. Skip it and AI is just a faster way to be wrong.

Governance stopped being the "no" department

The other shift I noticed: governance was framed less as a brake and more as an enabler. When a regulator, a board, or a customer can ask "why did the system decide that?", the teams who can answer move faster, not slower. Auditability, lineage, and clear ownership came up again and again — not as compliance theatre, but as the thing that lets you actually ship AI into places that matter.

My one-line takeaway The winners in the next phase won't be whoever has the cleverest model. It'll be whoever can put AI into production and defend the result — with data they trust and decisions they can explain.

The talent conversation got more honest

There was refreshingly little "AI will replace analysts" bravado. The more useful framing was that AI changes the shape of the job: less time hand-assembling outputs, more time on judgement, framing, and verification. The teams getting value weren't the ones with the most tools — they were the ones who taught people how to work with the tools, with rigour. That's the part I care about most, and it's where I'm spending my own learning time right now.

What I'm taking back to my own work

Keep investing in foundations. Every hour on data quality and clear definitions pays back double once AI is on top.
Design for the second arrow. A demo is a beginning; reliability and governance are the actual product.
Make the win explainable. If you can't say why the system did what it did, you don't have a finished system — you have a liability.

I left more convinced that the unglamorous work — pipelines, governance, definitions, teaching people — is exactly where the advantage is. The hype has moved on. The teams that quietly did the foundational work are the ones now shipping AI that holds up. That's the company I'd like to keep.