Proposals
After working through proposal workflows with AEC operators, the pattern is consistent: principals rewrite boilerplate, engineers hunt for past project narratives, and deadlines compress review time. Here is what that costs, and what one firm changed.
Alice Wong
May 14, 2026 · 8 min read
It is Friday afternoon. A qualifications-based RFP for a mid-size municipal project is due Tuesday. The proposal coordinator has the shell built, forms filled, org chart roughed in, layout ready. What she does not have is the part that wins: the technical approach, the right project references, and a principal's signature judgment on the win themes.
So the same thing happens that happens every cycle. A principal who bills at a premium spends his weekend rewriting boilerplate he has rewritten thirty times. Two project engineers stop fee-earning work to dig through old folders looking for the right past-project narrative. Review gets compressed into Monday. The proposal goes out at 4:58 Tuesday, good enough, and everyone exhales until the next one lands.
This is the RFP trap, and almost every AEC firm is in it. The bottleneck is not the proposal team. It is the handful of senior people the proposal cannot move without.
Across the industry, subject-matter-expert delays are the single most-cited proposal challenge, named by 48% of respondents. In AEC that pain is sharper than anywhere else, because the experts a proposal depends on are the same people whose hours are supposed to be billable. Every hour a project engineer spends reviewing a draft is an hour not booked to a job. The proposal eats fee-earning capacity twice: once in the time spent, and again in the work that time displaced.
And it is not one or two people. A third of AEC firms regularly coordinate 11 to 15 contributors per RFP; a quarter pull in 16 to 20; and 60% report that 50 or more people touch their proposal process. That is a lot of senior calendars to align against a hard deadline.
For the first time in five years, "bandwidth" has overtaken SME collaboration as the number-one challenge for proposal teams, a twenty-point surge. The read is simple: firms are submitting more, the market is busier, and headcount has not kept pace. More pursuits, same senior people, same number of hours in a weekend.
Put numbers on it. A typical proposal runs about 32 hours of effort and 9.3 days of elapsed time. Most teams need six to ten days to turn one around. Load senior contributor time at the going $150-$200 an hour and a single proposal carries four figures of cost before you count the billable work it pushed aside.
Now the part that hurts most. Most AEC firms admit they cannot respond to 10-19% of the RFPs that come in, not because the work is wrong for them, but because there is no senior bandwidth to write the response in time. Typical pursuit values run $1M to $10M each. Every skipped RFP is a real number walking out the door, and the reason it walked is almost always the same: the people who had to write it were already underwater.
At a ~40% industry win rate (closer to 37% in North America), the math of the trap is brutal. You spend your scarcest, most expensive hours on a process where the base rate says most submissions lose. And you skip the ones you might have won because there was nobody left to write them.
Every firm has tried to solve this with a boilerplate library. It helps at the margins and then stalls, for one reason: assembling boilerplate still requires senior judgment. Which three of forty past projects are the right references for this client? Which version of the firm's QA narrative fits this delivery method? What is the win theme that ties it together? A template can hold the words. It cannot make the call. So the principal is back in the document anyway.
The other half of the problem is retrieval. The best project narratives live in people's heads and in the closing memos nobody re-reads. When an engineer "hunts for past project narratives," he is doing manual search across a knowledge base that was never built to be searched. That is hours of senior time spent finding things, before a word of new writing happens.
The useful framing is not "AI writes the proposal." It is AI takes the senior person out of the parts that never needed senior judgment in the first place. A capable proposal agent today will:
A handful of platforms are now AEC-aware: ContraVault and AutogenAI on the complex-RFP and capture side, Joist AI built specifically for AEC marketing teams, QorusDocs for proposal content. The category matters less than the shift it enables: the agent handles retrieval, requirement-shredding, formatting, and the first draft. The principal moves from author to editor.
That is the whole win. When the machine produces a compliant, reference-backed first draft in minutes, the 32 hours collapse toward the few that genuinely need a senior brain: the win theme, the scoping judgment, the relationship read. Cut the senior portion by 75% and you have saved $3,600-$4,800 per proposal in displaced time, and you have freed the bandwidth to respond to the RFPs you used to skip.
Take a representative mid-size civil firm, call it forty people, a dozen pursuits a month, two principals who personally touch every proposal. Their old cycle looked like the Friday-night draft above: principals authoring boilerplate, engineers excavating narratives, review crushed into the last day.
They changed one thing first, and it was not the writing. They built the knowledge layer, past projects, approach narratives, and resumes in one searchable place, and pointed an AI drafting step at it. The first draft of every proposal now arrives reference-backed and compliance-checked before a human opens it. The principals stopped writing and started editing. The engineers stopped hunting. Review moved earlier in the week because there was a real draft on Monday instead of Tuesday.
The pursuits per cycle they could actually staff went up, because the senior hours each one demanded went down. They did not hire a proposal team. They stopped spending their most expensive people on their least leveraged work.
Two things keep this grounded. First, an AI draft is not a submit-ready proposal. The win still comes from human win themes, a real understanding of the client, and a principal willing to make a sharp call about approach. AI gets you to a strong first draft fast; it does not replace the judgment that closes.
Second, none of this works without the knowledge layer underneath it. If your past projects are scattered across drives and inboxes, the AI has nothing good to retrieve and will draft generic mush. The firms that win with this build the searchable record first, then let the drafting sit on top. Order matters.
The RFP trap is not really about proposals. It is about where a firm spends the only resource it cannot scale: senior attention. Right now most firms spend it on retrieval, formatting, and boilerplate, then wonder why the people who scope the work and win the clients are always underwater.
The technology to take those hours back is here and, increasingly, AEC-specific. The firms that pull out of the trap will not be the ones with the biggest proposal teams. They will be the ones who decided their principals should only ever touch the 20% of a proposal that no one else can.
Related posts
See all posts →AI Adoption
AEC firms that pair AI agents with loop engineers, humans who verify and approve agent work, win more pursuits in less time. Here's how the role works.
AI Adoption
Microsoft Copilot, Adobe AI, and project management integrations sit unused inside most AEC software stacks. Here is an honest audit of where the actual value lives, and why adoption stalls without a connected workflow layer.
Knowledge Management
Most AEC firms have decades of project data locked in PDFs, shared drives, and binders. The cost is not just retrieval time. It is the senior hours spent answering questions that the archive already contains.