AI personalisation is using AI to tailor outreach messages, content, offers, or experiences to individual prospects or accounts at scale, based on data about their role, company, behaviour, or stated priorities. The goal is to produce personalisation that feels specific and relevant to each recipient while processing hundreds or thousands of contacts in the time it would take a human to personalise a handful.

The spectrum of AI personalisation runs from light to deep. Light personalisation substitutes a company name, job title, or industry into a template. This is better than no personalisation but recognisable as automated. Deep personalisation draws on specific signals, such as a recent funding announcement, a job posting indicating a team expansion, a LinkedIn post the prospect wrote, or a competitor they mentioned publicly, to produce an opener that could not have been generated without knowing something specific about that person.

The quality of AI personalisation is entirely determined by the quality of the input data. A model can only personalise based on what you give it. If your enrichment pipeline provides generic company descriptions and role titles, your personalisation will be generic despite the AI doing exactly what it was instructed to do. Investing in richer data inputs, from web scraping, intent tools, news monitoring, or social listening, directly lifts the quality of personalised outputs.

The risk unique to AI personalisation is confidently wrong personalisation. A model that references a fact it hallucinated, a company milestone it misread, or a pain it assumed from limited data can produce a first line that is worse than no personalisation at all. Factual errors in personalisation are more damaging than generic openers because they signal that the sender did not check basic facts before reaching out.

Measure the performance of AI personalisation the same way you measure any personalisation: compare open and reply rates on personalised variants versus non-personalised controls. Track specifically whether deep personalisation outperforms light personalisation enough to justify the additional data cost. For most B2B audiences, meaningful specificity at the account level outperforms name-and-industry substitution.

What separates a useful AI term from AI theater is whether it reduces manual work without creating new accuracy or compliance risk. The strongest teams define exactly where the model is allowed to help, what still needs human review, and which failure modes are unacceptable before they automate anything. It usually becomes more useful when it is defined alongside Personalisation, Prompt template, and Guardrails.