AI in Marketing
Readiness Framework.
Test whether ideas are ready for AI, one at a time.
Work through the steps in order.
You can see which AI ideas are worth the marketing team’s time. An idea has to change the work, what customers see, or how they buy, and it has to beat what the business already spends and already gets. You come out knowing what to test and what to drop.
- 01Problem
- 02Baseline
- 03Inputs
- 04Test
- 05Organisation
Let’s get stuck in.
Problem
Choose a customer problem worth solving.
Name one customer and one problem. Then say whether AI would improve the work, the marketing that reaches them, or their experience.
- Who is the customer, and what are they trying to do?
- Which of those three would actually change?
- What would that person notice if it worked?
- If the only gain is the same marketing, done faster, is this still worth a pilot?
Write down
Customer
Who they are, and the job they are trying to do.
The change
The work, the marketing, the customer’s experience, or more than one.
Why it is worth it
The difference they would notice.
Stop if you cannot name the customer and the change they would notice. There is nothing to pilot yet.
Baseline
Establish the baseline.
Record the current cost, effort, quality and business outcome before anyone describes the future.
- What does this cost now, in money and in hours?
- How good is the current result, and who is judging that?
- Which business outcome are you hoping will move?
- What will the test be compared with?
Write down
Cost and effort
What you spend now, including the checking.
Quality
How you can tell whether the work is good.
Business outcome
The result you will watch, in the unit you already use.
Stop if cost, effort, quality and the business outcome are still unrecorded. An announcement is not a measurement.
Inputs
Prepare the inputs.
The test needs facts someone can stand behind: product information, definitions, permissions and evidence, with an owner.
- Which facts does this depend on?
- Which of those facts are missing, conflicting or out of date?
- What customer information is involved, and who is allowed to use it?
- Who owns the important information, and when did they last check it?
Write down
What the test needs
Product facts, definitions, evidence and permissions.
What is not ready
Anything missing, conflicting or stale.
Owner
A person, not a shared inbox.
Stop if the important information has no owner. A pilot that starts with a scavenger hunt will not survive the week.
Test
Run a bounded test.
Decide what the system may do, who checks the work, and what would make you stop or expand it.
- What may the system do?
- What must a person still decide?
- Who checks the work before a customer can see it?
- What result would stop it, and what result would justify expanding it?
Write down
Bounds
What it may do, and what stays with a person.
Checker
Who reviews the work, and when.
Stop or expand
The evidence for each, and who is allowed to call it.
Stop if nobody has agreed what the system may do, who checks it, and what would end or expand the test.
Organisation
Change the organisation around what works.
Update responsibilities, training and agency relationships only after the test has shown something worth keeping.
- Whose job changes if this stays?
- What should people learn, and what should they stop doing?
- What happens to the agency or the supplier?
- Where do the method, the sources and the check live?
Write down
Responsibilities
What moves, and to whom.
Training or suppliers
What people need, and what the agency relationship becomes.
Where it lives
Somewhere other than one person’s head and a prompt.
Stop if the useful result still depends on one person and an undocumented prompt. It is not finished.
Leave with this.
Five lines. Empty is allowed. Stopping at the first one you cannot answer is a result.
Problem
The customer, where AI would change something, and why that change is worth it.
Baseline
Cost, effort, quality and the business outcome today.
Inputs
What the test needs, what is missing, and who owns it.
Test
What it may do, who checks it, and the rule for stopping or expanding.
Organisation
What changes in the team, the training or the agency, and where the method lives.
Bring it to a
real problem.
One customer problem, a baseline, and a test small enough to stop. I use this framework with marketing teams.
Talk it through How AI Is Changing Marketing