The Capacity Assumption
Your firm rolled out an AI tool. Within a month, someone in leadership said some version of this: "Now that we have AI, we should be able to handle more pursuits."
This is happening across every industry, but it hits AEC proposal teams especially hard. The assumption shows up in engineering firms the same way it shows up in software companies: leadership sees an AI tool on a screen and concludes that the team now has infinite capacity.
The assumption is intuitive. AI automates tasks. Fewer manual hours per task means more tasks per person. So give the proposal coordinator three more RFPs to chase this quarter.
The problem is that the assumption is wrong — not because AI is useless, but because it solves a different problem than the one creating the bottleneck.
What the Data Actually Shows
Three data points frame the reality:
- Only 27% of AEC professionals use AI in any part of their workflow (ASCE, December 2025). The other 73% have not started.
- 67% of AEC organizations say less than half their proposal process is AI-powered (a 2026 industry survey of nearly 300 proposal professionals).
- 88% of proposal teams report high stress and 63% regularly work overtime — even at firms that have adopted AI tools.
That last point is the key one. Firms are adopting AI and their proposal teams are still burning out. The tool is not the problem. The expectation mismatch is.
Where AI Actually Helps in Proposal Work
AI is good at generating and transforming text. That makes it useful for a specific slice of the proposal process:
What AI does well
- Drafting boilerplate narrative. Management approach sections, safety program descriptions, QA/QC methodology — the sections that follow a predictable structure and draw from standard language. AI generates solid first drafts that a human then edits for accuracy and pursuit-specific details.
- Summarizing RFP requirements. Upload a 60-page solicitation and ask what the evaluation criteria are, what the page limits are, and what the key personnel requirements look like. AI extracts this in minutes instead of the hour it takes to read and annotate manually.
- Writing first-draft technical approaches. For pursuits where your firm has done similar work, AI can draft a technical narrative that the PM then refines. It is not the final version. It is a starting point that eliminates the blank-page problem.
- Answering questions about your own data. If your project history and staff qualifications are stored in a structured system, AI can query it: "Which of our engineers have LEED AP certification?" or "Which projects involved stormwater management in the last five years?"
What AI does not do
- Format resumes for a specific pursuit. The same engineer's education, certifications, and project history needs to appear in different layouts for different clients — SF330 Section E, branded one-page resumes, SOQ resume formats. This is not a text generation problem. It is a data formatting problem.
- Coordinate 11 contributors across a 3-week timeline. The average AEC submittal involves 11 or more contributors (industry benchmark data). Getting inputs from project managers, technical leads, subconsultants, and principals on time and in the right format is a workflow problem. AI does not chase down late submittals.
- Make go/no-go decisions. Whether to pursue an opportunity requires judgment about win probability, team availability, strategic fit, and client history. AI can surface relevant data, but the decision framework is a human process.
- Ensure compliance. Checking that every Section E resume matches the org chart in Section D, that project dates fall within the recency window, that amendment acknowledgments are included — this is detail work that requires domain knowledge and attention to the specific solicitation's requirements.
The Real Breakdown of Proposal Hours
Here is where proposal time actually goes for a typical AEC submittal:
| Task | % of Total Effort | AI Impact |
|---|---|---|
| Writing technical narrative | 15-20% | High — good first drafts |
| Reformatting staff resumes per pursuit | 20-25% | None — this is a data/formatting problem |
| Tailoring project experience sheets | 15-20% | None — same structural issue |
| Coordinating across contributors | 15-20% | None — workflow problem |
| Assembly, formatting, QC | 15-20% | Minimal — layout compliance is manual |
| Go/no-go evaluation and strategy | 5-10% | None — judgment call |
AI tools help with roughly 15-20% of the total effort. That is real and valuable. But it means that even a perfect AI drafting tool leaves 80% of the workload untouched.
When leadership sees "AI" and assumes 2x capacity, they are applying the 15-20% improvement to 100% of the work. The math does not work.
Why the Expectations Keep Growing Anyway
Three dynamics drive the capacity inflation:
1. Visible speed on one task gets generalized to all tasks
Leadership sees the proposal coordinator produce a management approach section in two hours instead of six. They conclude the whole process is faster. They do not see the eight hours that still go into resume formatting, the five hours on project sheet tailoring, or the ten hours coordinating inputs from a dozen people.
2. AI adoption announcements create pressure
When a firm announces it has adopted AI tools, there is implicit pressure to show ROI — usually measured in pursuit volume. "We invested in AI, so we should be able to pursue 40 opportunities this year instead of 30." This ignores the question of whether pursuing 40 opportunities at lower quality produces better results than pursuing 30 at higher quality. Firms with disciplined go/no-go processes report 15-25% higher win rates (APMP).
3. The bottleneck is invisible to leadership
The hours spent reformatting resumes, tailoring project sheets, and chasing contributor inputs are not visible to principals and BD directors. They see the final deliverable — a polished submittal — and assume the process behind it is efficient. They do not see the proposal coordinator working until midnight reformatting the same engineer's resume for the fourth time this month because each client wants a different layout.
What Actually Increases Proposal Capacity
If you want to handle more pursuits without burning out your team, the solution is not "add AI to the existing process." The solution is to fix the process that consumes the most hours.
Fix the data problem, not the writing problem
The 40-50% of proposal time that goes to resume formatting and project sheet tailoring is not a writing problem. It is a data problem. The same information — education, certifications, project history, role descriptions — gets reformatted from scratch for every pursuit because it lives in Word documents rather than structured data.
When staff qualifications are stored as structured data, generating a tailored resume becomes a configuration task, not a reconstruction task. Select the relevant projects, choose the format, generate the output. What currently takes 30-60 minutes per resume takes minutes.
Reduce pursuit volume, increase win rate
Industry benchmark data shows that 44% of AEC firms cannot complete 20-29% of incoming RFPs due to capacity constraints. These firms are saying yes to more pursuits than they can handle, producing lower-quality submittals across the board.
A better approach: use a go/no-go framework to say no to marginal opportunities, invest the saved hours in higher-quality submittals for better-fit pursuits, and win at a higher rate. Fewer pursuits, better submittals, more wins.
Automate the formatting layer, not just the drafting layer
The early AI adopters seeing real results are not just using AI to write narrative. They are using structured data systems to eliminate the formatting layer entirely. A Bluebeam report from October 2025 found that 68% of early adopters saved at least $50,000 and 46% saved 500-1,000 hours — primarily by eliminating repetitive formatting, search, and assembly tasks.
The savings come from solving the right problem. Not "write faster." Instead: "stop rebuilding the same information from scratch for every pursuit."
How to Have the Capacity Conversation With Leadership
If your principal or BD director is pushing for more pursuits because "we have AI now," here is how to reframe it:
Step 1: Track your hours for five submittals. Break them down by task category using the table above. Show where the time actually goes.
Step 2: Show the gap. AI helps with 15-20% of the effort. The other 80% is formatting, coordination, and assembly. Increasing pursuit volume without fixing those tasks means the 80% gets worse.
Step 3: Propose the right investment. Instead of "more pursuits with AI," propose "same pursuits, less time per pursuit, using a system that manages the data layer." The time saved goes into quality, not volume — until your team actually has capacity to add pursuits without overtime.
Step 4: Anchor to win rate. The goal is not more proposals. The goal is more wins. If adding three pursuits per quarter drops your win rate because quality suffers, the net result is the same number of wins with more burned-out staff.
Frequently Asked Questions
Can AI write a complete AEC proposal?
No. AI can generate first-draft narrative sections — technical approaches, management plans, past performance summaries — but it cannot assemble a complete submittal. A typical proposal requires formatted resumes, tailored project sheets, org charts, compliance documentation, and coordinated input from 10+ contributors. AI handles the text generation layer, which represents roughly 15-20% of the total effort. The formatting, coordination, and assembly work remains manual unless a firm has moved to structured data systems.
How much time does AI actually save on proposals?
For the drafting portions of a proposal — management approach, technical narrative, boilerplate sections — AI typically saves 40-60% of the writing time by producing usable first drafts. For a full submittal, that translates to roughly 3-5 hours saved on a proposal that takes 30-60 total hours. The more significant time savings come from structured data systems that eliminate resume and project sheet reformatting, which can save 10-20 hours per submittal.
Why are AEC firms slower to adopt AI than other industries?
The ASCE survey identified several factors: concerns about accuracy in technical documents, lack of internal AI expertise, compliance requirements for government submittals, and a culture of conservative technology adoption. The 27% adoption rate reflects an industry that is cautious by nature — which is appropriate when accuracy matters, but creates a competitive gap as early adopters gain efficiency.
Should we stop using AI for proposals?
No. AI is genuinely useful for drafting and summarizing. The point is not to abandon AI — the point is to set realistic expectations about what it does and does not automate. Use AI for the tasks it handles well (narrative drafting, requirement extraction, data querying). Fix the underlying data and process problems that consume the other 80% of proposal effort separately.
What tools actually help with proposal formatting and assembly?
Purpose-built proposal management tools that store staff qualifications and project data as structured information, then generate formatted outputs per pursuit. This is different from AI writing tools. The value is not in generating text — it is in generating the right format from a single source of data. RFPM.ai does this for AEC firms: structured staff profiles generate tailored resumes in any format, and project data generates experience sheets per client requirements.