Saying “we use AI” can mean three different things. A pattern-matcher that scores your leads, an explorer that groups your customers, or a gambler that spends your budget. Each learns in a different way, sits in a different system, and fails in a different way. A marketing leader does not need to build any of them. They need to be able to say which one just made a decision.
You just said “we use AI” in a marketing meeting. That could mean three completely different approaches.
One scores your leads. One groups your customers. One spends your budget.
They’re all machine learning, but each learns in a different way, runs on a different system, and is governed differently.
You do not need to build any of them as a marketer. But you should be able to tell which one you have. That is the whole of AI literacy for a marketing leader.
The pattern-matcher
The job: lead scoring, churn and propensity, send-time optimisation.
MarTech system: predictive scoring inside your marketing-automation platforms, and propensity models in a customer data platform.
How it learns: from labelled, structured history. You show it thousands of past cases with the answer attached (“this lead converted, this one did not”), and it learns to guess the answer on new ones.
What to watch: it only knows the world you labelled for it. You have to catch the drift and refresh the data in time.
The explorer
The job: audience segmentation, look-alike and affinity discovery, anomaly spotting.
MarTech system: the segmentation and audience engines inside your customer-data and analytics platforms.
How it learns: from unstructured, unlabelled data. You hand it a pile of customer data and it proposes the groups.
What to watch: it will always find groups, even when some of them are not statistically valid. A human has to review them before your team builds a campaign.
The gambler
The job: bid and budget optimisation, next-best-action, dynamic pricing.
MarTech system: automated bidding in ad platforms, and journey-orchestration and on-site optimisation engines.
How it learns: by trial and reward. It tries something, sees the score, and does more of whatever scored well.
What to watch: it does exactly what you tell it. Set the reward to clicks and it will buy you the cheapest, emptiest clicks on earth.
How to design your AI adoption plan
- Step 1. Pick use cases. Tag every candidate by the learning approach. “Predict who will churn” is a pattern-matcher. “Find segments we have not named” is an explorer. “Decide where the next pound goes” is a gambler. The tag tells you the data you will need to train the model, and the risk you take on when you deploy it.
- Step 2. Get the data ready. The pattern-matcher needs labelled history. The explorer needs clean, rich data and a human to verify its output. The gambler needs room to experiment before it goes live.
- Step 3. Choose tools. When a vendor says “AI-powered”, ask which of the three it is. The answer tells you the failure mode you are signing up for, and what your team has to check after go-live.
- Step 4. Set up governance principles for different risk types. The pattern-matcher drifts as its labels age. The explorer hands you groups that may be just noise. The gambler games the reward you set. Each needs its own control. A single blanket “AI policy” will not be enough.
- Step 5. Upskill and define owners. Literacy is not everyone learning to code or building the model. It is everyone able to say which AI just made a decision. Make sure you have owners for each model during training, deployment and maintenance.
Share this with one marketer who still thinks “AI” is one thing.
Meltem Günyüzlü Ateş, CFCIM, CAIP is an AI-first marketing leader, advisor and educator in the AI era, and a member of European Women on Boards. She leads marketing operations across 60+ markets at the British Council and writes the LinkedIn newsletter Marketing AI, without the hype. If you are doing this rewiring inside a global marketing function and want a second pair of eyes on the operating model, that is the work she does.