All tracks

Supplier Response / RFP Automation Agent

max +6 bonus

Interface within the Supplier Portal that helps suppliers fill in responses to RFXs — both Fairmarkit-generated events and external RFXs imported from other systems. The agent helps suppliers respond to more events with higher-quality answers by learning from previously submitted responses as a knowledge base. By extending to non-Fairmarkit events, this capability can be productized as a premium supplier experience. Having more supplier activity — qualitative and quantitative, beyond just what is generated in Fairmarkit events — also provides more relevant data to further improve supplier recommendations.

Scope: Supplier Portal integration, RFX response generation, RAG over historical responses, external RFX import/parsing, answer confidence scoring, agent language/style configuration, feedback learning loop.

North Star Metric

Suggested answers are accurate and require no edits — only confirmation before being applied automatically to the line item or questionnaire.

Bonus Criteria (+6 pts)

+2

Format-independent external RFX handling

Different external RFXs come in different formats that need to be interpreted and retained during the answering process.

Verified by: Visible through the demo and in evals

+2

Response confidence level

The more data the agent has, the better the answers over time. Ability to show response certainty level at the line item level.

Verified by: Visible through the demo and in evals

+1

Agent language configuration

Ability to instruct the agent on answer style (brief vs. lengthy) and scope (e.g., skip a particular worksheet).

Verified by: Visible through the demo and in evals

+1

Answer references, reasoning & rating

Agent learns from supplier feedback (good vs. bad answer), references prior answers, and rates confidence.

Verified by: Visible through the demo and in evals

Suggested Evals

How embedded the experience is within the Fairmarkit event flow
Ease of upload and review of suggested answers
Response quality — low hallucination level given limited RAG dataset
No teams assigned to this track yet. Assignments happen after team submission closes.