Everyone is selling you an AI agent. Almost nobody mentions why it won't work yet.

There is a lot of noise right now about agentic AI — systems that do not just recommend but act, running review responses, listings, posting, email sequences, campaign optimisation on their own. The pitch is everywhere, and unlike a lot of AI hype, this one is real. It is genuinely moving into live workflows this year.

So here is the part the vendors tend to skip. The biggest thing standing between you and a useful AI agent is not the agent. It is your data.

Only about a third of marketers say they can actually unify their customer data — and that, not model quality, is the ceiling on what agentic AI can do. An agent is only as good as what it can see. Point a sophisticated autonomous system at fragmented, contradictory, half-tagged data spread across six tools that do not talk to each other, and it will confidently do the wrong thing, faster than a human ever could.

This is the unsexy truth we keep running into. The brands that will get real value from AI agents in the next year are not the ones buying the most advanced tool. They are the ones who did the boring work first — got their customer data into one trustworthy place, cleaned up the tagging, made sure "a customer" means the same thing in every system.

Buying an agent before fixing your data is like hiring a brilliant analyst and handing them a filing cabinet someone emptied onto the floor.

Fix the plumbing first. The agent can wait two months. A wrong decision made autonomously at scale cannot be easily undone.