State Street Requested this
Enhancement description
When an SSIM user selects the AI Draft option intended to retrieve direct Content Library responses, Responsive should semantically identify and return relevant approved library answers based on the meaning and intent of the question, while honoring the user’s selected Collection, Tag, product, or other content filters.
Current behavior / problem
Users frequently receive no direct Content Library result even where they expect approved content to exist. State Street described the experience as returning “nil” results and not functioning like the semantic search they expected. This forces RFP users to manually locate, source, and rewrite answers, eliminating much of the expected efficiency from AI Draft.
Required product behavior
- AI Draft should surface multiple responses when searching Content Library first, rather than just one exact match, since semantic search can find relevant content even without verbatim keyword overlap
- Semantic search does not appear to be the search logic used in this feature. It currently seems to use Auto Respond / keyword matching logic.
- User-selected content filters must be respected while still allowing semantic matching within the eligible content population.
- When no result is returned, the behavior should reliably indicate that no sufficiently relevant eligible content was found
- Results should be ranked by semantic relevance and remain traceable to their underlying Content Library source.
- The experience should be reliable enough for standard-response workflows where teams expect approved library content to be reused before generating a net-new answer.
Business impact
This is SSIM’s highest-priority AI Draft concern. On a recent one-day-turnaround request containing 99 questions, the team reported that the direct-library option produced no usable response and the generated option also required extensive manual work. The team had to manually source and change responses, undermining the expected value of semantic retrieval.
