Project-billed AI for small engineering firms.

Engineering firms already bill on project budgets. Fixed-fee, T&M, not-to-exceed, percentage-of-construction — the contract structure varies, but in every case the firm and the client have agreed up front what the project costs and how reimbursables get handled. AI used to support the engineering work fits neatly into a structure that has existed since the AIA started publishing standard contracts a century ago.

What's already in the reimbursable expenses clause

Pull up the AIA B101 owner-architect agreement, the ACEC E-500 engineer-owner agreement, or whatever standard form your firm uses for civil, structural, or MEP work. Every one of them has a section labeled something like Reimbursable Expenses or Reimbursable Direct Costs. The standard list includes printing and plotting, courier and shipping, travel, permit fees, specialty software licenses procured for the project, and a catch-all for direct costs incurred in connection with the project that aren't part of the basic services.

AI fits the catch-all on its own merits, but the more direct analogue is the specialty software line. When a structural firm buys a one-off license for a specific finite-element package to model an unusual condition on a project, that license cost is reimbursable. When that same firm uses AI tokens to research a code interpretation on the same project, the accounting treatment is identical.

Where AI fits in the engineering workflow

In small civil, MEP, and structural firms, AI is now showing up in:

• Code research — interpreting IBC, IBC referenced standards, NEC, NFPA, and local amendments. Tracing requirements across multiple documents.
• Calculation verification — sanity-checking hand calcs, generating alternate solution paths, cross-referencing against worked examples.
• Drawing and spec review — comparing drawings against specifications, flagging coordination issues, drafting RFIs to architects.
• Report drafting — assembling engineering reports, drafting executive summaries, generating boilerplate sections.
• Specification production — drafting and editing MasterSpec sections, comparing manufacturer literature against spec requirements.

Each of these is tied to a specific project number in the firm's project management system. The PM already knows which job is which when the time gets posted.

The numbers for a small firm

Consider a small civil or MEP firm with four PEs, running about 8 active projects in a typical month. Across code research, report drafting, calc verification, and spec production, AI usage averages roughly $150 per project per month:

• 8 active projects × $150/project = $1,200/month
• $1,200 × 12 months = ~$14,400/year

On T&M projects with a reimbursable clause and on hourly-rate work with reimbursables called out, that $14,400 flows back through invoices alongside printing, plotting, and travel. On fixed-fee projects, it gets priced into the next proposal — visible to the principal who is approving the bid instead of invisible inside firm overhead.

What the proposal needs to say

Proposals and contracts don't need to change dramatically. The defensible practice:

• Add AI to the reimbursable expenses list in the proposal, alongside printing and travel. One line: "AI tools and computational research, billed at actual cost."
• Itemize on monthly invoices, the same way reimbursables are already itemized.
• For fixed-fee work, include an estimated AI budget in the basis-of-fee section, so the client sees that AI was scoped into the bid.

The piece that was missing

Engineering firms have always tracked time to project. The same software that posts billable hours can post AI usage — if the AI usage is attributed to a project at the moment it happens. Without that, AI lands in firm overhead and the project P&L reports are quietly wrong.

The accounting pattern is general — see the Cost Recovery overview for the full framework, or Interchange for the platform that makes per-project AI attribution operational.