Our pricing rules change every quarter. How can a forward-deployed engineer update an internal quoting tool without breaking existing quotes?
Separate pricing rules from the rest of the quoting workflow, and make each rule change a controlled release. Before building, the engineer should work with the person responsible for pricing to document effective dates, exceptions, approval authority, and what happens to quotes already sent.
Do not assume that a new rule should recalculate every old quote. Agree on the business policy first. The tool may need to retain the rule version and inputs used for each quote, so staff can explain an amount later. Whether an existing quote remains valid or must be revised is a leadership decision, not one the AI should make on its own.
Create a set of test cases from real quoting situations, with sensitive details removed where appropriate. Include normal cases, boundary amounts, exceptions, and quotes created near a pricing change. Before release, compare the new results with the expected results and have the pricing owner review differences.
- Give rule changes an effective date and an identifiable version.
- Restrict who can propose and approve a change.
- Test changes away from the live quoting process before publishing them.
- Keep a practical rollback path if a released rule produces incorrect quotes.
Review rule changes as part of your existing operating rhythm. If a release exposes a recurring issue, bring it to the appropriate EOS Issues List rather than repeatedly patching individual quotes.
Category: Forward-Deployed Engineering