You can't personalise your way into someone's life

Earlier this month our team joined Opticon London, where AI's reshaping of the marketing stack ran through almost every stage: marketers moving from prompting tools to working with persistent AI teammates, customer experiences that assemble and optimise themselves in real time, and sites being readied for the agents we explored last month. 

Yet some of the most memorable sessions pushed back, arguing that humanity in marketing matters more now, not less. Our own talk, He's just not that into you: Personalisation that earns the relationship, sat squarely in that tension. When the stack makes limitless personalisation possible, the strategic question becomes which personalisation earns its place. Impose engagement and customers retract. Having the data doesn't mean you need to do something with it. 

We built the talk around four bad behaviours most of us would recognise from dating. Here's how we'd flip each one. 

Read the room, don't play the field 

Blanket personalisation treats every customer as someone to be won over, whatever the moment. Reading the room means responding to what someone is actually trying to do right now, not what a profile predicts they might want. That begins as a strategic choice about where personalisation genuinely helps, and becomes a design brief for when it should show up and when it should step back. 

Be true to yourself 

Autonomous optimisation will happily nudge your brand towards whatever converts this week, which is often whatever converts for everyone. As we argued over the summer, your voice travels inside the context you give these systems. Engineer it in deliberately, so a thousand personalised variations still sound unmistakably like you. 

Be thoughtful and available 

Chasing a customer who has already said no rarely changes their mind. A customer who declines, pauses or goes quiet is telling you something. Respect it, and design the experience so you are easy to find when they come back, on their terms rather than yours. 

Practise restraint 

Just because you can know something about a customer doesn't mean you should show it. Personalisation that reveals too much feels less like service and more like being watched. In build terms, that means clear rules about what data can be drawn on, when and for what, agreed before the optimisation engine is switched on rather than after. 

The stack will keep getting smarter. The brands customers stay with will be the ones that know when not to use it.