Strategic consultingAI workflow designProposal systems
Frostpine starts with how the proposal work actually moves, then builds the content system and the internal tools underneath it — deployed and running in real use, not a prototype.
Measured
All five figures60%
Cut in proposal turnaround.~4 h → ~30 min
First editorial pass on a 50-page document.Services
Four kinds of work, usually in that order. Most engagements start with the proposal process, because that is where the deadlines are, and end with tooling the team can run without me.
Map how the work actually moves, find where it stalls, and design workflows people still use after the novelty wears off. The output is an operating decision, not a tool list.
Content libraries and semantic search over the material a team already owns, so drafting starts from the firm's own past work instead of a blank page.
Small, specific systems shaped like the organization that has to run them — not another platform to administer, license, and eventually abandon.
Planning, rollout, and the unglamorous adoption work that decides whether any of it is still running a quarter later.
Evidence
A multi-agent editing system for a proposal team, built and deployed and in daily use, with semantic search running across the firm's existing project database and content library. Each figure below carries its own mark: one square is one system, one tick is one document, and both hour bars are drawn on the same eight-hour scale.
60%
Cut in proposal turnaround.
12
AI workflow skills and production systems deployed.
49
Proposal documents in the semantic search library.
~4 h → ~30 min
First editorial pass on a 50-page document.
4–6 h
Saved per medium-complexity proposal, on research and first-draft content.
The client is a Fortune 500 engineering firm — described, not named.
Pipeline
Five stages of a proposal, in the order they run. Each one carries a published before-and-after for a single task, drawn on the same eight-hour scale as the ledger above — which is why the first stage runs off the top of it. Choose a stage for the friction sitting under the figure.
Where it breaks
Client-intelligence profile
Nothing gets written until someone works out who the buyer is, what they have bought before, and what this document actually has to answer.
Where it breaks
Past-performance research
The proof exists somewhere — a project record from a few years back, filed under a name nobody uses now. Finding it is the work.
Where it breaks
Relevant-experience first draft
The blank page is rarely blank. It is the last proposal, pasted in and half-edited, carrying somebody else's scope through one more cycle.
Where it breaks
Boilerplate adaptation
Reusable text stays reusable right up to the moment it has to match this scope, this reviewer, and this page limit.
Where it breaks
First editorial pass, 50-page document
Voice, consistency, compliance, and formatting, checked across the whole document at once, always against the clock.
Each figure is a published before-and-after for the task named with it; the stage names are generic. The intake bar runs past the top of the scale because that stage is measured in days, not hours.
Approach
Three stages, in sequence. Nothing gets built in stage one, and nothing is considered finished at the end of stage three. The window below is the one the whole engagement is planned around.
Stage 1
Map how a proposal actually gets made here: the pressure points, the reuse, the approvals, the person who quietly holds it together. Tooling decisions come after that, or they come apart.
Stage 2
What survives is a system the team can read, trust, and repair — grounded in documents they already recognize, not in a tool they were talked into.
Stage 3
The work is not done at launch. It is done when the team stops asking whether to go back to the old way, and the system keeps running without me standing next to it.
On the AI
It is used to assemble and adapt what a team already owns. It is not used to invent facts.
Contact
Describe the proposal work as it runs today and where it breaks. You will get an answer about whether a system is the right thing to build, and what the first four to eight weeks would have to cover.