Write two or three sentences here that go further than the summary above. What the work actually involves.

What usually brings people here

The situation rather than the solution. Show you have met this problem before.

What I do

The actual work. Three or four short points rather than a list of capabilities.

How it runs

The shape of the engagement, roughly how long it takes, and what you need from them.

Write two or three sentences here that go further than the summary above. What the work actually involves.

What usually brings people here

The situation rather than the solution. Show you have met this problem before.

What I do

The actual work. Three or four short points rather than a list of capabilities.

How it runs

The shape of the engagement, roughly how long it takes, and what you need from them.

Where AI genuinely changes a commercial process, and where it does not. Plus the behavioural data foundation underneath it, because without one no model has anything useful to act on.

What usually brings people here

A leadership team told AI is urgent without being told what it is urgent for. Several pilots running with no owner and no number attached to any of them. Or an expensive data project that improved the reporting and changed nothing.

The questions that separate a plan from an enthusiasm

What specifically gets cheaper, faster or better, named as a process rather than a capability. Who stops doing what, because if nobody stops the saving is theoretical. Who owns the number in twelve months. What happens when it is wrong, and who notices. What data leaves the building and on what terms. And what you are choosing not to do in order to fund it.

What I do

I start from the commercial outcome, never from the technology. We keep the two or three initiatives that survive those questions and stop the rest. Then I make sure a channel is actually wired to act on the output, which is the step that gets skipped and the reason personalisation so often turns out to be a first name in a subject line.

The foundation underneath

Identity, consent, ownership and definitions. Four organisational problems and one technical one, which is why these projects run long. Build them properly and the AI conversation becomes straightforward. Skip them and no model will rescue it.

Where this comes from

I led a cross brand behavioural data foundation in a large regulated organisation, built the business case, got agreement across several companies, and ran the first machine learning and personalisation work on top of it. I also hold an executive programme in applying AI to business, covering implementation, scaling and the ethics of it.

Write two or three sentences here that go further than the summary above. What the work actually involves.

What usually brings people here

The situation rather than the solution. Show you have met this problem before.

What I do

The actual work. Three or four short points rather than a list of capabilities.

How it runs

The shape of the engagement, roughly how long it takes, and what you need from them.

It deliberately refuses to produce a summary. Summaries are the least useful thing a model can give you about trading data.

Use it on at least twenty responses. Below that you are reading anecdotes with extra steps.

Uncomfortable and quick, which is the right combination the day before a board meeting.

The support inbox is the most under used research asset in most companies. It contains a prioritised list of your problems, written by customers, free of charge.

Reporting is a by product. The point of a data foundation is activation, which means being able to recognise the same customer across channels and do something differently as a result, in the moment, without a person in the middle.

The test that separates the two

Can you take a behaviour observed in one place and act on it in another, this week, without an export?

If a customer browses a category on a phone, does the email they get tomorrow know that? If they contact support with a problem, does the marketing automation pause? If they are already a subscriber, does the acquisition campaign stop paying to reach them?

Those three questions have caught out every data project I have reviewed, including some very expensive ones.

If a data foundation only improves reporting, it was a reporting project with a larger budget.

What makes it hard, and it is not the technology

• Identity. Recognising one person across devices, channels and a physical store is where most of the difficulty lives.
• Consent. What you may use, for what purpose, and being able to prove it. This is design work, not a checkbox.
• Ownership. When the data spans brands or business units, someone has to be allowed to decide.
• Definitions. Two teams counting a visit differently will produce two truths and one long argument.
• Governance of tagging, which is dull, ongoing, and the reason foundations decay within a year.

Notice that four of those five are organisational. The technical build is the part with a plan and a vendor. The rest is why these projects run long.

Privacy as a design input

Building consent and purpose limitation in from the start is cheaper than retrofitting it, and it produces a better product. A team that knows what it may use stops arguing about what it might use and starts building.

It also survives regulatory change, which on a five year horizon is not optional.

And then AI has something to work with

Every conversation about personalisation and machine learning eventually arrives here. Models are not the constraint. The constraint is whether there is a trustworthy, consented, unified view of behaviour for a model to act on, and whether any channel is wired to receive the output.

Build that and the AI conversation becomes straightforward. Skip it and no model will save the project.

That is not a criticism of the teams. Real personalisation requires a unified, trustworthy, consented view of behaviour, and a channel wired to act on it. Without both, what is left is decoration.

The three levels

• Cosmetic. Name, location, last purchase referenced. Cheap, mildly effective, occasionally embarrassing when the data is stale.
• Segmented. Genuinely different content for genuinely different groups. This is where most of the value sits and where most businesses stop short.
• Individual. Decided per person, in the moment, by a model. Powerful, and only possible on top of the second level being solid.

Almost everyone wants to jump from the first to the third, because the third is what gets discussed at conferences. The second is where the money is and it is unglamorous, because it looks like maintaining segments and writing more variants.

Everyone wants to jump straight to individual personalisation. Almost all the value is in the level they skipped.

When personalisation actively hurts

Recommending what someone just bought. Continuing to sell to an existing subscriber. Cheerful marketing arriving the day after a complaint. Each of those tells the customer that the business does not know them, which is worse than not personalising at all, because effort was clearly expended.

Suppression rules matter more than recommendation logic and get a fraction of the attention.

What I would build first

Not a model. A set of suppressions and a handful of well chosen segments, running reliably across every channel. That is unglamorous, quick, and it improves the customer experience immediately.

Then, with a foundation and a channel that can act, the model becomes worth building, and there is something for it to do.

The one measure worth watching

Not open rate or click rate, both of which improve when a message is merely more relevant. Watch retention and repeat purchase among the personalised group against a holdout. It is a slower measure and it is the one that tells you whether any of this worked.