Most marketing teams spent last year doing AI. Licenses rolled out. Workshops run. A proud slide in the board deck showing adoption up and to the right. Twelve months on, I can tell you what most of that bought: faster first drafts, prettier decks, a team a bit more productive at the work it was already doing. What it did not buy, for most people, is a single decision made differently. The machine got faster, but it did not get smarter.

AI and agentic are not the same thing. AI makes your existing team faster at the work they already do, while agentic changes what the work is. One is a productivity story and the other is an operating model story. Conflating them is the most common mistake I see in my peer group, and it is costing people a year they cannot afford to lose.

I saw this early, and I moved. Not because I got it all right, but because the sooner you understand that readiness is about foundations, not features, the sooner you stop spending against the wrong problem.

Here is what I would tell any CMO still working it out:

It is a mindset shift first. After 20 years in this industry, the hardest thing for me to accept was that a lot of what I have learned — how campaigns get built, how teams are structured, how decisions get made — I would have to unlearn and relearn. That is uncomfortable at any stage of a career, and more so the further in you are. But it is the work.

Context matters more than the agents. The hard part is not the AI but the context that an agent needs to make a good decision. For me that meant ICP signals, deal stage, what sales is hearing on calls, which segments have converted and which have burned budget. Most of that lives in senior people's heads. An agent without context is an over-confident intern!

Interest is not the same as buy-in. Interest means it gets discussed. Buy-in means budget, roles redrawn, and someone senior owning the outcome when it gets hard. If you cannot name the exec who loses sleep if this fails, you do not have buy-in.

Hire people who already know what good looks like. You cannot figure this out in real time while running the function. The greatest value comes from people who have lived a version of this elsewhere and can tell you, within a week, which of your assumptions are wrong.

Measure the commercial outcome, not the adoption. Seats filled, prompts run, hours saved… all vanity. You must measure decisions made differently, cycle times compressed, and revenue you can defend to the CFO.

Tinkering is not transformation. The foundations take longer than you think, and there is less time to build them than most people realize.