Personalisation has a reputation problem. Done well it feels like good service: the right content, the right offer, at the right moment. Done badly it feels like being followed. AI raises the stakes on both sides, because it can personalise more, faster.
The simple test
Ask whether the customer would understand and accept how you know what you know. Recommending an article because they read a related one passes. Referencing something they did on a different site, or a detail they never gave you, does not, even if it is technically available.
Where to personalise first
- Content order: show the topics a returning visitor cares about first.
- Onboarding: shorten it for people who clearly know what they are doing.
- Offers: make them timely and relevant to declared interests, not inferred private ones.
- Support: recognise the customer and their history so they never repeat themselves.
Rules that keep it comfortable
Use data people gave you knowingly. Let them see and change what you hold. Explain recommendations in a sentence. And keep a non-personalised version that works well, so nobody is punished for opting out.
