What was breaking before?
The problem was time and consistency. Every proposal at this B2B services firm was assembled by hand: hours per document, formatting that shifted from author to author, and turnaround slow enough for prospects to cool while the draft sat in progress. A proposal is mostly repeatable structure. Scope, pricing, terms and company credentials change little between deals, yet each writer rebuilt them from an old file, hunting for the latest version of the pricing table or the current logo.
Copy and paste introduced errors. Two proposals sent in the same week could look like two different companies. The senior people who should have been selling were formatting tables instead. None of this was a skill problem. The team knew exactly what a good proposal said. The mechanics of producing one, layout, boilerplate, version control, simply consumed hours the business could not bill for.
What did we build?
We built an AI workflow that drafts branded proposals and quotes, with a human approving every document before it sends. The system has three parts. First, an intake step collects the deal details: client name, scope, line items and pricing, pulled from the enquiry or entered through a short form. Second, an AI drafting engine assembles the document from an approved library of sections, writing the client specific parts and slotting standard terms, credentials and pricing tables into a locked brand template, correct fonts, correct logo, correct layout, every time.
Third, an approval gate. The finished draft lands with a named person as a ready to review document, never as a sent email. That person edits, approves or rejects. Only an explicit approval releases the proposal to the client. The firm's existing tools stay in place; the workflow reads from them and writes the final document back where the team already works.
Client identity and specific commercial metrics stay under NDA. What we publish is the shape of the system and the outcomes the client has approved. On a call we walk through live systems, not slides.
How does the system decide what to do?
The system decides by rules, and every judgement call escalates to a person. Standard requests follow the happy path: recognised service, known pricing structure, complete details, so the workflow drafts immediately from the approved section library. Anything outside those rules stops and asks. Missing scope information triggers a request for the gap instead of a guess. Unusual pricing, non standard terms or a request the library does not cover gets flagged for a human before drafting continues.
The AI never invents a discount, a delivery date or a legal clause; those values come from the approved library or from the person handling the deal. The final gate is absolute: no document reaches a client without explicit human approval, and every draft, edit and approval is logged. The result is a system that behaves like a fast, careful assistant, not an autonomous sender. People stay accountable for what goes out.
What changed for the team?
Two things changed: drafts are ready in minutes instead of hours, and every document that leaves the firm is consistent and on brand. The minutes versus hours shift is the visible one. A request that used to sit in a queue while someone found time to write now returns a reviewable draft almost immediately, so the conversation with the prospect keeps its momentum.
The consistency shift is quieter but just as valuable. Formatting no longer depends on who wrote the document or which old file was copied. The template is locked, the boilerplate is current, and the brand looks the same in every proposal and every quote, every time. The team's role moved up a level: less typing and formatting, more reviewing, tailoring and deciding. The approval step keeps a person on every document, so quality control did not loosen. It became the whole job instead of the afterthought.
How would this look in your business?
The same shape fits any business that sends documents built from repeatable parts: proposals, quotes, service agreements, statements of work, onboarding packs. The pattern stays the same. Deal details in, an approved library of sections and a locked brand template in the middle, a human approval gate before anything sends. The tools change per business. The drafting can start from a CRM stage change, a form, an email or a WhatsApp message.
The output can be a PDF, a document link or a line item quote in the accounting system. The approval can happen in email, in the CRM or in a chat channel, wherever the responsible person already lives. Tell us which document eats the most hours in a week and how it gets produced today. We will map the current steps, show which ones a workflow can absorb, and give an honest answer on whether the build is worth it.
Related capabilities. The same parts, your business.
Keep reading. Pages close to this one.
Tell us what runs slow. We build what fixes it.
Send one message describing the document your team dreads producing. We reply with an honest read on what a drafting workflow can fix, what it will take, and whether it is worth building at all.
