AI-First Means Data-First
AI does not create advantage on its own — data does. Why outdated data infrastructure is the biggest barrier to AI adoption, and what the companies pulling ahead do differently.
"AI first" sounds bold — but without "data first," it's just noise.
I was recently asked by CIO Online to share my perspective on how data systems are shaping the future of AI. The takeaway is simple: AI doesn't create advantage on its own — data does.
Models are commoditizing. Data access isn't.
In the AI-native economy, success won't come from who has access to the best models. Those are increasingly commoditized — every competitor can call the same APIs you can. Durable advantage comes from who can:
- Access the right data — the operational data that actually describes your business, not the fraction that happens to live in a warehouse
- Trust the quality of that data — because an agent acting on wrong data is worse than no agent at all
- Move it quickly across the organization — at the speed autonomous systems operate, not at the speed of a nightly batch job
- Govern it responsibly — with access controls and audit trails designed for software that makes decisions, not just reads reports
The real barrier to adoption
Outdated data infrastructure is becoming the biggest barrier to AI adoption. Legacy systems weren't built for real-time processing, unstructured data, or autonomous decision-making — and it shows the moment you try to put an agent in production. The pilot works in the demo environment; then it meets the ERP that updates overnight, the CRM with five definitions of "customer," and the document store nobody has governed since 2014.
I've spent two decades inside these systems — CRM, CPQ, ERP, billing — at Oracle, GE, F5, Fastly, and now leading agentic AI at enterprise scale. The pattern is consistent: the AI initiatives that stall don't stall on model quality. They stall on data the model can't reach, can't trust, or isn't allowed to touch.
What the leaders do differently
The companies pulling ahead are rethinking data as a strategic asset, not a byproduct. They're investing in modern, unified data platforms that make AI actually work at scale — before they scale the AI itself. That sequencing matters. It's unglamorous work, and it's exactly why their agents ship to production while everyone else's stay in pilot.
Bottom line: if you want to be AI-first, you have to be data-first.
Originally shared on LinkedIn, following a commentary request from CIO Online.