Personalise on data people knowingly gave you, towards an outcome that's good for them as well as you. Use AI to segment, score and nurture, judge it on qualified leads and lifetime value rather than volume, and ask whether the customer would be comfortable seeing exactly how the message reached them.
There are two kinds of personalisation, and customers can tell them apart instantly.
The first feels like the business paid attention. The recommendation makes sense. The follow-up arrives when you’d actually want it. The message reads as if someone who understood your problem wrote it. The second feels like being followed around a shop by someone reading your diary. Same technology, opposite effect.
AI has made both far easier to do at scale. Which means it’s now just as easy to build an engagement programme that quietly annoys thousands of people as one that helps them. The tool doesn’t know the difference. That judgement is still yours.
How do you set up AI personalisation that works?
Start with the data you’re allowed to use and the outcome you want, not with the tool. Personalisation works when you match the message, offer and moment to what a customer actually needs, using data they knowingly gave you, towards an outcome that’s good for them as well as you. It fails when it’s built to extract rather than to help, or when it runs on data people didn’t realise they were handing over. Get the foundation right and the technology is the straightforward part.
Two things have to be true before any of it works. The proposition has to be clear, so the system knows what it’s matching people to. And the data has to be clean and lawfully held, so the matching is based on something real. AI personalisation on a vague proposition and messy data just produces confident nonsense, faster.
Segment, then score, then nurture
For lead generation specifically, AI earns its place in three jobs that used to eat a team’s week.
Segmentation is the first. Instead of a handful of broad audiences built on hunches, a model can cluster customers by how they actually behave and what they’re worth, and pick out who’s genuinely worth reaching. That’s more useful than another campaign aimed at everyone, and cheaper.
Lead scoring is the second. Lead scoring means ranking your leads by how likely they are to become customers, so sales spends time on the ones that matter. AI does this by learning from who converted before, rather than from whoever shouted loudest in a meeting. It’s one of the clearest wins available, because it settles the oldest argument in the building: sales and marketing disagreeing about lead quality. Score the leads on evidence and the argument gets shorter.
Nurture is the third. For anything sold over a long cycle, and most considered purchases are, AI can help time and shape the follow-up so it supports the customer’s decision instead of pestering them towards it. Conversational triage can qualify and route enquiries without a person, as long as it hands over cleanly the moment it’s out of its depth. The measure of good here is qualified leads, retention and lifetime value, not how many emails you managed to send.
The line between persuasion and manipulation
Here’s where marketing has to be honest with itself. The same personalisation that helps can tip into something worse, and the more powerful the targeting, the easier it is to cross without noticing.
Hyper-personalised, emotionally targeted messaging can move from persuasion into exploitation, especially with people who are vulnerable, in debt, unwell or simply having a bad week. Just because data can be collected and used doesn’t mean it should be. There’s a line where personalisation stops feeling like service and starts feeling like surveillance, and crossing it costs trust that’s very expensive to win back.
You can stay the right side of it with a few habits. Personalise on data people knowingly gave you, for purposes they’d recognise. Be able to explain, in a sentence a customer would accept, why you’re using what you’re using. Give people a real way to see and change what you hold about them. And apply more caution, not less, the more sensitive the moment or the audience. Compliance with the law is the floor. Whether a customer would feel comfortable if they saw exactly how the message reached them is the better test.
The point isn’t to be timid. Sephora built one of the most-copied personalisation stories in retail by solving a genuine customer problem, letting people try thousands of shades on their own camera before buying, then tying that to recommendations. It worked because it helped people buy with confidence, not because it squeezed them. That’s the model. Solve a real problem for the customer, at scale, with AI doing the heavy lifting.
What to do this week
Look at one live engagement or lead-gen programme and read it as a customer would. What data is it using, and would the customer be comfortable if you told them? Does the timing help them or suit you? At what point does a real person take over, and does that handover actually work? And is the whole thing aimed at a qualified, valuable customer, or just at volume?
If the honest answer is that it pesters more than it helps, that’s fixable, and fixing it usually improves the numbers as well as the relationship. Personalisation done right is quieter and more effective at the same time.
Is your personalisation busy without being useful?
If your CRM is full but nobody trusts what’s in it, or your personalisation is busy without being useful, tell me what’s getting stuck and we’ll work it out.