Readiness is about your accounts, not the AI
Most "are you ready for AI" pieces talk about culture, buy-in, and change management. Useful, but it skips the part that actually breaks. When automation goes wrong, it is rarely because the model was dumb. It is because it was pointed at an account that could not tell it what success meant, could not stop it from going too far, and could not show anyone what it changed.
So the real question is narrower and more honest. Can your accounts give an automated system a clean signal to act on, a hard boundary it cannot cross, and a record of everything it did? If yes, you are ready to start small. If no, automation will move faster than your ability to catch it, which is the opposite of what you want.
The four conditions below are the ones that matter. You do not need all four at full strength. You do need to know where each one stands.
Condition one: tracking you would bet a budget on
Automation optimizes toward whatever you measure. If your conversion tracking double-counts, fires on the wrong page, or attributes revenue to the last click and nothing else, the AI will learn the wrong lesson and learn it quickly. Bad data does not make automation cautious. It makes it confidently wrong.
You are ready on this front when you trust your numbers enough to make a real spend decision from them without cross-checking three other tools first. You are not ready when "the data is a bit messy" is a sentence your team says out loud. Fix the tracking before you automate against it, not after.
Condition two: limits you can state in one sentence
This is where bounded autonomy lives. Every safe automated change follows the same shape — a trigger it watches for, an action it takes through a real connection, and an impact it measures afterward. That is the Trigger, Action, Impact model, and it only works when the action half has a ceiling.
Before you turn anything on, you should be able to answer these without hesitating:
- Spend caps — what is the most this can move in a day, and on what?
- Change ceilings — how big a single adjustment is it allowed to make on its own?
- Exclusions — which campaigns, audiences, or assets are off-limits entirely?
- Approval thresholds — above what size does a human have to sign off first?
If those answers exist, you have the raw material for guardrail-driven automation. If they do not, you are not asking the AI to be careful. You are asking it to guess where your limits are, which is exactly the black-box behavior worth avoiding.
Condition three: access that is read-only first and reversible always
A ready account connects on terms you control. The honest version of safe access is not "the tool never touches my account" — it touches it, that is the point. The honest version is read-only and auditable to start, with connections you can revoke in one click and changes you can undo.
Ask how a tool connects before you ask what it can do. Can it observe and recommend before it acts? Is every change logged with a clear before-and-after? Can you pull the access tomorrow without breaking your account? If the answer to any of those is fuzzy, that is a readiness gap on the vendor side, not yours — and it is a good reason to slow down.
Condition four: a definition of "good" written as a number
An automated system needs a target to steer toward, and "more leads" is not a target. The fourth condition is the one teams skip most: a clear, agreed cost-per-result, return target, or efficiency number that everyone would name the same way if you asked them separately. Without it, automation will optimize toward the metric that is easiest to move, not the one that pays the bills.
You are ready here when the goal is specific enough to argue about. You are not ready when "good" means a different thing to the person buying the media than it does to the person reading the board deck. Decide what success looks like in a number first, then let automation chase it — not the other way around.
Signs you are not ready yet (and that is fine)
Some accounts should wait, and waiting is a legitimate answer. You are probably not ready if your tracking is in mid-migration, if nobody can name your real cost-per-result target, if "who has access to what" is a mystery, or if a single bad week would put the business at risk. Automating on top of any of those does not de-risk your marketing. It scales the fragility.
Not-ready is a to-do list, not a verdict. Tighten the tracking. Write down the caps. Map the access. Decide what "good" looks like in a number. Most accounts can clear those in a focused stretch of work, and then automation becomes an accelerator instead of a gamble. The point of checking first is to spend that effort where it actually moves your readiness, not everywhere at once.
Find out where you actually stand
The Readiness Score walks through 13 questions across tracking, limits, and access, then tells you which conditions you have met and which to fix first. Four minutes, no login.
Get your free Readiness Score →Keep reading
- AI marketing guardrails: what to set before you automate — turn condition two into concrete limits.
- Add AI to your martech stack — where automation fits once you are ready.
- The execution depth spectrum — how far to let automation go, from recommend-only to act.