Field proof
The Proposal Process, Automated
WO-002 · Proposal automation · Sits in front of the pursuit process- Situation
- An engineering consultancy that lives and dies by solicitations. Every piece of work starts as an RFP someone has to find, read, evaluate, and answer — and the entire process ran on OneNote notebooks, isolated files, and institutional memory.
- What they tried
- The AI reads the solicitation, checks the insurance, matches the staff, and scores the pursuit — then posts its scores next to the humans', not instead of them. A solicitation arrives. Someone reads it end to end and pulls out the requirements by hand. Someone checks whether the firm's insurance coverage clears the bar — by finding the requirements in the document and finding the policies in a drawer. Someone tries to remember which staff and which past projects fit. The go/no-go decision happens in a hallway conversation, undocumented, based on whoever read the RFP most recently. If it's a go, the proposal gets drafted in files nobody else can see, and whether it actually answers the solicitation is discovered by the client. None of it was one system. It was notebooks, memory, and hope — and it worked, the way workarounds always work: on the backs of the people holding it together.
- What we built
- An AI pipeline that works a solicitation the way a good proposal team would — one step at a time, each step building on the last. Reading and qualifying: upload a solicitation and the pipeline takes it from there — parses the document, validates the requirements against the firm's actual insurance coverage, matches key staff to what the RFP asks for, matches past projects to what the owner wants to see, checks the services fit, and drafts an executive summary of the pursuit. Each stage hands its findings forward to the next, the way a team compounds knowledge instead of re-reading the file. The go/no-go, scored — not outsourced: the pipeline's final step scores the pursuit across five weighted categories — personnel, owner experience, similar owners, similar projects, jurisdiction — and posts its scores in a matrix right next to the human reviewers' columns. The AI's column is read-only and always first; every reviewer scores independently alongside it. The recommendation isn't a binary: go as prime, go as JV, go as sub, no-go, delay. The AI informs the call. The humans make it. The proposal, reviewed before the client sees it: once a proposal is drafted, the team runs it against the solicitation itself. Before the AI critiques anything, it asks — three to five clarifying questions grounded in the actual documents, covering missing requirements, conflicts, amendments, and priorities. The team can answer or explicitly skip, and either way the choice is recorded; skipped questions never silently shape the result. Every run — the focus, the questions, the answers, the feedback — is preserved, so any review can be reopened and understood months later.
- How fast
- Scoped as Phase 1 in early February — and driving live pursuit decisions the same quarter.
- What changed
- A rebuilt business process. Solicitation to submission runs on one platform the whole team works from. Pursuit decisions are scored, documented, and made in minutes instead of remembered differently by everyone in the hallway. And every proposal gets checked against the solicitation before the client does the checking.
- What's next
- They're already planning what else in the firm belongs on the platform next.
What's the workflow nothing fits?
The form your team fights with. The process living in a spreadsheet. The tool you pay for and route around. Tell us about it — that's where we start.
Tell us about your workflowLive in 30 days or it's free. Cancel anytime. Keep everything you own.