Self-service is only as good as the number behind it
Automation does not create pricing problems. It removes the last person who was quietly absorbing them.
As long as a person writes every quote, a wrong price gets caught by judgement — someone looks at it and frowns. The moment a system quotes on its own, that check disappears and the number becomes a promise the business has to keep.
This is the part most teams discover after they launch, which is the expensive order to discover it in.
Find the second author
The usual root cause is not a bad formula. It is that the same business fact is computed in more than one place: the website, the quoting tool, the ERP. The copies do not disagree on day one. They drift — when one is updated and the others are not — and the business finds out through a customer.
So before opening anything to self-service, audit for second authors. For every number your business states out loud — cost, margin floor, lead time, availability — there should be exactly one place it is computed, and everywhere else should be a reader, never an author. Where you find a second author, you have a customer-facing error with a date on it you cannot see yet.
Why this is not a pricing project
Consolidating the engine looks like housekeeping and gets deprioritised accordingly. It is not housekeeping. It is the floor under every automation you build afterwards, because an assistant drafting a quote is only ever as trustworthy as the figure it is handed.
People treat data architecture as a separate, duller topic from AI adoption. It is not a separate topic. It is the precondition.
These notes come out of the Weekly Guidance emails. Level 1 is €499 a month: one strategy email a week, a monthly AI opportunity scan and a consolidated monthly report.
Start Level 1