Fix · The second line
Back-office AI,
when the bottleneck moves.
Fix sales follow-up and the bottleneck moves: to quoting, to paperwork, to the owner's inbox. That's when back-office AI pays. Before then, it mostly trims costs.
Our rule: back-office AI is worth building when the back office is the bottleneck: quotes go out slowly, documents pile up, or the owner's day goes to repeat messages. Then we build it like everything else, with tests from real cases, a named owner and a decision ledger you keep.
01 / What we build
The work
behind the sale.
- Quotes
Quoting and proposals
Drafts from the job details and your pricing rules, checked by a person before they go out.
- Documents
Documents and invoices
Attachments read, matched to the right deal or customer, filed and entered once.
- Editorial
Editorial desks
Copy-editing and rewriting assistants built on your style guide, with an evaluation harness that tests them on real articles.
- Inbox
The owner's inbox
Repeat questions drafted, meetings turned into tasks, so the owner's hours go to decisions.
02 / An example
Evals before
go-live.
For a Canadian travel trade publisher we built four editorial assistants (copy editor, rewriter, press-release desk, interview desk) on the publication's style guide, and an evaluation harness that tests them on real articles.
- 01
Why the harness matters
Editorial quality is easy to claim and hard to see. Tests on real copy show whether a change helped.
- 02
Why the tests belong to the client
When the style guide, prompts and test cases are kept by the client, a better model can be tried on their own history before anyone switches.
Questions
Straight
answers.
Why do you lead with sales and not the back office?
In owner-led companies the bottleneck is most often revenue follow-up. AI off the bottleneck can trim costs, but it doesn't make the business produce more. When the bottleneck moves to the back office, so do we.
How much of a leader's time goes to messages?
In Porter and Nohria's study of 60,000 CEO hours, electronic communication took about a quarter of working time. In an owner-led firm, that time comes straight out of the bottleneck.
What stops back-office AI making mistakes?
Tests built from real cases before go-live, a person approving anything that leaves the building, and a log of every decision and override.
Next step
Is the back office
your bottleneck?
Book a 30-minute call. If it isn't yet, we'll tell you what is.