A chatbot can answer a question correctly yet still leave the user doing extra work. Generic credit-card advice, for example, is less useful when it ignores a customer’s existing accounts, spending habits, and goals. Personalization adapts the response—or the path through a conversation—to what is relevant to that user. section 9.1 section 8.2.2
Think of context as information the assistant already has or learns during the interaction. It may include location, time zone, device, preferences, behavior, and previous conversations. The communication modality—such as voice or text—also shapes an appropriate interaction. Use these signals when they help answer the request, rather than making the user repeat them. section 9.1.2 section 9.0 section 8.2.2
From generic advice to useful advice
Asked how to manage expenses this month, an assistant might simply say to track spending and stay within a budget. For a frequent traveler, it could instead discuss travel budgets and rewards; for a recent graduate starting a job, it could address a new income and upcoming expenses. The question is the same, but the useful answer depends on the person’s circumstances. section 9.1.1
Personalization can change more than wording:
- Shorten a flow: use a caller’s detected phone number as a proposed account-lookup option, then ask for confirmation instead of making them supply it again. section 8.2.2
- Adapt a recommendation: remember an interest in eco-friendly products when helping a returning shopper choose a laptop. section 9.3.1
- Choose an appropriate response method: static answers preserve consistency but allow limited personalization; retrieval-augmented generation (RAG) can draw on retrieved information and user context for an adaptive response. section 5.4 section 9.3.1
Personalize only where the added relevance justifies the effort and risk. Protect location data, maintain user consent, and delete it when no longer needed. Keep simple queries quick and inexpensive; reserve more processing for complex requests that need it. A useful test is whether the interaction becomes more relevant and timely while requiring fewer repeated questions. section 9.1.2 section 9.3.2 section 9.1.1
