BlogProduct design

Product design

Design AI Search Around Finding and Verifying Information

Create an AI search experience with useful results, visible sources, manageable filters, and clear ways to refine a request when the first answer misses.

People search because they need information for another task. A search experience should help them find the right material, judge whether it applies, and continue their work. AI can offer a more flexible way to express a request, but the design still needs to support verification, navigation, and refinement.

Choose the right result for the task

Some queries need a document list; others benefit from a short synthesis with links. Do not assume every request should produce a conversational answer. A person looking for a specific form may prefer the form itself, with its owner and revision date visible.

In a hypothetical equipment portal, a request about installation could return the applicable guide, relevant section, and product revision. If the product cannot determine the revision, ask for it or present clearly distinguished options. The result should help users choose the right source rather than merely display plausible text.

Keep scope and provenance understandable

Show the collection being searched and any filters that affect the result. Users should be able to tell whether they are searching public guides, internal procedures, or account-specific records. Make it easy to change scope without losing the original request.

Place source links near the information they support. Useful context can include a document title, section, date, and owner when available. Avoid overwhelming the result with metadata, but expose enough for people to decide whether the information is appropriate for the task they are performing.

Design the second attempt as carefully as the first

A first query may be vague or use different terminology from the documents. Let users refine it through a follow-up question, a filter, or a suggested term. Preserve the current context so refinement feels like progress instead of starting from nothing.

Handle empty or weak results with concrete next steps. Explain the searched scope and offer a route to browse the relevant library or contact its owner. Review unsuccessful queries for content gaps as well as search problems. Better information architecture may be the most useful response to a repeated failure.