Article
AI as a project cost code — cost recovery for construction firms.
Construction is one of the easier industries to reframe AI cost recovery in, because the accounting infrastructure already exists. A general contractor running cost-plus, GMP, or T&M contracts is already passing direct costs through to the owner with markup. CSI MasterFormat and CSI UniFormat already provide a coding structure for every conceivable line item. The only question is which cost code AI lives under.
Cost-plus contracts already do this
Under a standard AIA A102 or A103 cost-plus arrangement, the owner reimburses the GC for "Cost of the Work" plus a fee. Cost of the Work includes labor, materials, equipment rental, subcontractor invoices, permit fees, blueprint reproduction, project-specific software, and a long tail of direct costs incurred for the benefit of the project. The contract is explicit: anything spent on the project is reimbursable; anything spent on running the firm is not.
AI used to draft an RFI response for the Riverside job is a direct project cost. AI used to draft the firm's holiday card is not. The distinction is the same one the accounting team already applies to printer toner, fuel for the company truck, and the BIM software license that gets allocated across active jobs.
Where AI shows up on construction projects
Across a typical mid-size project, AI is now showing up in:
- Bid preparation — analyzing solicitation documents, drafting narratives, generating schedule logic, scoping subs.
- RFI drafting and response — both creating RFIs and assisting design teams in turning around answers faster.
- Submittal review — comparing submittals against spec sections, flagging deviations, drafting transmittals.
- Change orders and claims — assembling backup, drafting the narrative, summarizing the impact.
- Schedule analysis — interpreting CPM updates, drafting recovery schedules, narrative writeups.
- Punch lists and closeout — generating O&M summaries, warranty letters, training narratives.
Each of these activities maps to a specific phase of the project and a specific cost code in the GC's accounting system. Attribution is conceptually trivial — the project manager already knows which job the work is for.
The numbers for a small GC
Consider a small general contractor with about 5 active projects in any given month. Across bid prep, RFIs, submittals, and change-order documentation, AI usage averages roughly $250 per project per month:
- 5 active projects × $250/project = $1,250/month
- $1,250 × 12 months = ~$15,000/year
On cost-plus and GMP contracts, that $15,000 is fully reimbursable as part of Cost of the Work, plus whatever fee or markup the contract authorizes. On lump-sum contracts, it stays in firm overhead — but at least it's now visible and can be priced into the next bid. Either way, the spend stops being invisible.
What changes in the contract
Most owners are not surprised by AI showing up on a pay application. The conversation is conceptually identical to the one that happens when a GC adds a new specialty software subscription or upgrades the project management platform. The defensible practice:
- Name AI explicitly in the schedule of values or in a general-conditions line item. Owners on cost-plus contracts are paying for general conditions anyway; AI fits cleanly under that umbrella.
- Code AI usage to a specific cost code — many GCs are using something like 01-50-00 Temporary Facilities and Controls or a new 01-31-00 Project Coordination subcode.
- Itemize on the pay app with backup that shows usage attributable to the project (the same way you'd back up a copy of a sub invoice).
The piece that was missing
The accounting model already works. What hasn't existed is per-project, per-user attribution at the moment of AI use. Generic AI subscriptions show up as one bill at month-end with no project allocation. You can't put it in the right cost code if you don't know which project it was spent on.
That's the operational gap. The contract framework is already in your standard AIA forms.
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.