Response time was a commercial constraint.
RFQs arrived as unstructured requests. Sales teams needed to identify parts, check availability, and apply pricing context before responding. In-stock quotes took 18 hours and out-of-stock requests took 15 days.
An RFQ is a moment in a relationship.
A request can be a follow-up, come from a sister entity, or depend on a prior pricing agreement. Treating every message as standalone loses that context. The data model needed to carry the relationship alongside the extracted request.
The decision
- Chose
- Automate routine quoting with commercial guardrails.
- Over
- Keep quoting fully manual.
- Evidence
- In-stock quotes took 18 hours; out-of-stock quotes took 15 days.
- Trade-off
- Reserve autonomy for requests that satisfy both confidence and commercial controls.
- Cost
- Model account context and escalation paths before expanding automation.
What is being requested?
- Part number
- Availability
- Requested quote
What makes this quote appropriate?
- Prior quotes
- Sister entities
- Pricing tier
- Relationship history
- Revenue risk
- Lead time
- Alternates
- NPI vs repeat
Automation with an explicit boundary.
The workflow connected intake, extraction, inventory matching, and quote-to-order routing. Confidence tiers and price floors defined the boundary of routine automation, with escalation and human override available for commercial judgment.
Account context included prior quotes, sister entities, pricing tier, relationship history, revenue risk, lead time, alternates, and new-product versus repeat-order intent.
System Details
- Intake
- Intent classification and part number extraction from RFQs.
- Matching
- Inventory lookup and acceptable alternates.
- Context
- Prior quotes, account relationships, pricing tiers, and order intent.
- Routing
- Model confidence and commercial risk determine the review path.
- Controls
- Price floors, escalation paths, and human override on every quote.
- Integration
- Quote-to-order workflow.
Faster quotes, with the source of each result clear.
In-stock quoting moved from 18 hours to under 1 hour. Out-of-stock turnaround moved from 15 days to under 3 days across a workflow handling 8K RFQs/month.
Separately, shipment tracking and forecasting reduced logistics costs by 22%. That result belongs to the logistics work, not the RFQ workflow.
Permission matters as much as prediction.
An accurate extraction does not make a quote commercially appropriate. Trust came from carrying account context into the decision and making the limits of automation explicit.