When a customer asks for “the seal from the last shutdown,” an experienced rep may remember the equipment, the kit supplied and the engineering decision behind it. The rest of the team has to find that information before preparing the next quote.
AI can retrieve those earlier records and use them to suggest the right part. The useful work is connecting the customer’s shorthand to the decision that made the earlier quote correct, while checking what has changed since then. Here is how to make that history usable without turning old discounts and one-off substitutions into permanent rules.
Use the old quote to identify the part, then check today’s terms
Historical records answer what happened before. Your current catalog, agreements and approvals determine what you can offer now. A previous quotation can identify a part without authorizing the old price.
ERP systems also distinguish quote records from later transactions. In Sage X3, quote revisions and the conditions for order conversion affect which record represents the agreement. Preserve those differences when importing history.
Store the customer wording with the part and supporting record
Store the source quotation and the line that supports a match. For Excel RFQs, keep the original sheet and row beside the resolved value. Add the customer account and, where relevant, the equipment or project. Keep the decision date and any validity limit.
Store a record like this:
| Field | Example |
|---|---|
| Customer | Account 1048, Manchester site |
| Customer wording | “Shutdown seal kit” |
| Resolved item | Kit K-17, revision B |
| Evidence | Quote Q-318, line 4; engineering clarification dated May 6 |
| Limit | Pump P-204 with housing revision C |
| Commercial rule | Recheck current price; previous discount was project-specific |
Leave missing values open and route them to the account owner.
Reuse the match without copying the whole quote
Suppose a buyer requests the same kit for another plant. The old quotation may identify a candidate, but the new plant may use a different equipment revision. Present the evidence and ask for the missing detail before treating the match as confirmed.
The same applies to prices. An Odoo pricelist can have validity conditions; a historical rate does not override them. In NetSuite quote pricing, trace the current rule before carrying an earlier rate forward.
For a seal kit, the earlier installation is only part of the evidence. Parker’s O-Ring Handbook covers elastomer selection and operating applications as well as dimensions. A kit that fitted the housing can still be unsuitable after the customer changes the process fluid or operating temperature.
Store the application conditions behind an approved substitution with the historical match. If the buyer says “same kit, new cleaning process,” retrieve the earlier material decision and route the changed conditions to technical review. The rep can ask about the actual change instead of repeating the old choice or starting the investigation again.
Record which customer or application a correction applies to
When a rep corrects a match, record what changed and where the correction applies. “F-10 means filter X for customer A” is different from “replace F-10 with filter X everywhere.”
Have the reviewer distinguish a standing rule from a one-time exception. A changed request can need renewed approval even when the earlier quote was valid. Retain the prior decision and its replacement so the team can investigate a later mistake. The agent should show which rule or source it used; a generic statement that it learned from feedback is insufficient for review.
Test whether the agent uses the correction on later requests
Reserve some RFQs that were not used in setup. After one correction, test a new request from the same account and similar wording from another account. Record correct reuse, missed reuse and incorrect transfer separately.
Agree on pilot acceptance criteria with the commercial owner before running those requests. Include a repeat request with changed operating conditions, such as the seal-kit example above. The correct outcome may be a new technical question, even when the earlier item match was accurate. Keep checking these cases after launch as customers change equipment and applications.
Make your best reps’ knowledge available across the team
Choose Bourne when your quoting process depends on knowledge scattered across past quotes and customer conversations. We can build account memory that connects a request to the supporting record, then gives the reviewer the current terms and the reason for a proposed match.
A new rep can see why the senior rep chose a particular seal kit, and the team can correct the account record once rather than explain the same exception in every handoff. Build your quoting workflow with Bourne.
Frequently asked questions
Can AI learn from our old quotes without training a model?
Yes. A workflow can retrieve relevant records and use them as context without changing the model’s weights. The supplier should explain how it stores, retrieves and corrects that information.
Should we upload every historical quote at the start?
Begin with a bounded set that includes reliable completed work and representative exceptions. Check account identifiers, revisions and access permissions first. More records can add contradictory or expired information as well as useful context.
How do we stop the agent repeating an expired discount?
Keep the historical price as evidence and check authorization against current terms. Include an expired agreement in the evaluation and confirm that the agent requests approval rather than silently copying it.
What if two reps handled the same exception differently?
Surface the conflict to the person who owns the policy. Do not make the most recent or most common action an automatic rule unless the business approves that interpretation.
Further reading
Parker: O-Ring Handbook
Elastomer properties, application guidance and seal dimensions.
Arda Bulut
Arda Bulut is the co-founder and CTO of Bourne and HockeyStack. He leads engineering at Bourne, building the platform people use to create AI products, agents and automations.