Every civil engineer running a consulting firm has read a contract from a developer, a school district, a housing authority, or a city that had no business being signed as written. The indemnification was broad-form. The IP clause handed ownership of the drawings to the client. The liability wasn’t capped at all, or it was capped at total project cost instead of the contract scope value. The standard of care said "warrants."
You caught it because you’ve been doing this long enough. Or your broker flagged it when you asked about insurance. Or you paid an attorney to read it, which is the right call on the ones that matter.
What’s harder to justify is spending two to three hours of a principal’s time on the pattern-matching pass before you even know whether there’s something worth an attorney’s hour. Most of that review is not judgment. It’s recognition: clause by clause, against a known checklist, compared to industry standards you’ve seen a hundred times. Recognition is exactly what an agent does well.
I run a civil engineering firm, Calichi. We receive contracts on most projects, from a range of client types, across four offices. So a few years into watching principals do the same checklist work on every agreement that came in, I built an agent to do the production part. Here’s the honest version of what that workflow is, what it used to take us, and what it takes now.
What a contract review actually involves
A client contract for civil engineering services comes in a few shapes. It’s a direct agreement between the firm and an owner: a developer, a school district, a municipality, a housing authority. Or it’s a subconsultant agreement where the firm is sub to an architect or a prime consultant, and the prime’s agreement with the owner flows down. These are structurally different reviews.
On a subconsultant agreement, many of the problematic terms came from the owner’s contract. The prime didn’t write them and can’t change them. Negotiation energy has to go to what the prime actually wrote, not what they inherited. Confusing those two categories wastes time and relationships.
Either way, the review follows a pattern. You read the indemnification clause: is it fault-based, or does it say "regardless of negligence"? Is there a duty to defend? Is it one-way? Most civil engineers know that broad-form indemnification is void in California under Civil Code 2782.8. What they don’t always know is that every other state has its own version of that rule, with different scope, and you have to look it up for the project’s state rather than assume California knowledge transfers everywhere.
You read the IP clause. Does the client get a limited license, or does the contract say documents "become property of Client"? Work-for-hire language. IP assignment clauses. These matter because your instruments of service have value beyond the current project, and an IP clause that doesn’t protect that is a problem you won’t notice until it’s too late.
You check for a limitation of liability. If there’s none, you’re exposed to whatever a court decides a design error cost. A cap pegged at the contract value is the industry standard. Anything uncapped is a flag. Anything capped at total project cost on a project where your scope is a fraction of that total is also a flag, for a different reason.
You check the standard of care. "Professional engineering standard" is what it should say. "Warrants," "guarantees," "highest standard," or "error-free" are words that put you on the hook for perfection instead of reasonable professional judgment.
You read the insurance requirements. Does the client want to be named as additional insured on your errors and omissions policy? That can’t be satisfied. The insured-vs-insured exclusion voids E&O coverage for exactly that scenario. Does the contract require you to maintain coverage "in perpetuity"? A claims-made E&O policy can’t satisfy a perpetuity requirement.
Then you compare what you found against EJCDC E-500, AIA B101, and ACEC guidelines, which tell you what a balanced engineering services agreement actually looks like. You document findings. You draft redlines. You separate items that can be negotiated from items that can’t. And then you write the negotiation letter.
For a typical client agreement, that’s two to three hours of work before you have anything the principal can act on.
What it used to take us
At Calichi, contract review ran two to three hours of a principal’s time on a typical client agreement. That’s our own measured number, from production.
The time wasn’t in the hard part. The hard part is judgment: is this clause actually a problem given how we’d be performing on this project, or is it standard boilerplate that no one enforces as written? Should we push back on the liability cap, or accept it given the client relationship and the scope? Those calls belong to the principal.
The two to three hours was mostly the pattern-matching. Reading every clause against a mental checklist. Looking up the relevant state statute. Pulling the EJCDC reference language to compare. Drafting the redline. Building the negotiation letter from the findings. Same shape every time, done by the most expensive person in the room.
How the agent does it now
The engineer emails the contract PDF to the agent, along with the project state and the contract scope value, and the agent runs a structured three-pass review.
Pass one is CRITICAL issues: clauses that are deal-breakers, legally unenforceable, or uninsurable. That includes broad-form indemnification, IP assignment and work-for-hire language, a missing limitation of liability, the absence of a mutual consequential damages waiver, standard of care that uses warranty or guarantee language, an additional insured requirement on the E&O policy, a perpetuity coverage requirement on a claims-made policy, non-compete clauses, liquidated damages tied to the design services, and termination-without-payment terms. Each finding gets the specific contract language quoted, the legal or coverage basis stated, and a recommended action.
Pass two is WARNING issues: terms that are unfavorable and worth negotiating but not reasons to walk on their own. One-way indemnification. Elevated standard of care language that doesn’t reach the warranty level but still says "utmost care" or "best efforts." Insurance limits above the firm’s baseline. Pay-when-paid clauses with no time limit. Retainage on professional services, which is a construction concept that doesn’t belong in a design agreement. Venue clauses that pick a jurisdiction with no connection to the project or the firm. Provisions where only the client can recover litigation costs.
Pass three is a completeness check. Is the scope of services defined? Are there payment terms and an invoicing schedule? Is there a late payment provision? Is there an additional services mechanism so scope additions don’t become unpaid work? Are dispute resolution provisions in place? Missing provisions are flagged as items to request before signing, not afterthoughts.
The agent auto-resolves the governing state from the project address or the named jurisdiction, then applies that state’s specific anti-indemnity statute to the indemnification review. For California projects, that’s Civil Code 2782.8. For Texas projects, it’s Chapter 130 of the Civil Practice and Remedies Code. For Illinois projects, it’s the Construction Contract Indemnification for Negligence Act. The agent researches the equivalent provision for whichever state the project lands in, rather than applying one state’s framework everywhere.
Every finding is tagged as either FLOW-DOWN or NEGOTIABLE. On a subconsultant agreement, the agent identifies which terms came from the owner’s agreement and which terms the prime consultant wrote. Negotiation requests go only to NEGOTIABLE items. You don’t spend time asking a prime to change language they didn’t write and can’t modify.
The agent runs a full insurance comparison: each coverage type the contract requires, compared against the firm’s baseline limits. Gaps are flagged. Additional insured on E&O is flagged CRITICAL. Perpetuity requirements on any claims-made policy are flagged CRITICAL. Limits above the baseline are flagged WARNING with a note on what obtaining the required coverage would involve.
The liability exposure analysis proportions the risk to the civil construction value, not the total project cost. On a large institutional project where civil site work is a fraction of total construction, a design error in the grading or utility plan does not expose the firm to losses measured against the entire project construction cost. The agent frames the liability cap analysis against what a CDG-scope error could actually cost, not the headline number.
The output package includes a branded report with every CRITICAL and WARNING finding documented, the specific contract language quoted, the EJCDC, AIA, or ACEC standard cited for comparison, and the recommended redline. It produces a redlined contract PDF. It produces a negotiation letter containing only the NEGOTIABLE items, framed as a professional request referencing industry standards as the basis for each ask.
At Calichi, that two-to-three hour task comes back in about 20 minutes of engineer time. That’s our own production-measured number. The principal’s 20 minutes is almost entirely review and judgment, not clause scanning.
The QC gate
The agent doesn’t deliver its output to the principal directly. Every package goes through a validation pass first. The pass checks that all required sections are present, that no placeholders remain unfilled, that the delivered files are complete, and that the audit trail of external references is intact: the EJCDC clause text cited, the AIA section quoted, the statute text pulled and sourced.
The point is the same as in every other agent workflow here. The first pass isn’t always perfect. Neither is the principal reviewing it at the end of a long week. A deterministic check that runs the same way every time catches the gaps before they reach the person whose job is to decide whether to sign.
What stays human
The agent is decision support for the principal. It is not legal advice.
It surfaces risk flags, compares against industry standards and state law, and produces the redlines and the negotiation letter. What it doesn’t do is tell you whether to sign.
That call belongs to the principal. The client relationship matters. The project context matters. The nature of the scope matters. On some projects you accept contract language that the review flags as unfavorable because the relationship and the work justify it. On others you walk. The agent gives you the information to make that call in 20 minutes instead of two to three hours.
For agreements with significant exposure, an attorney still reviews the final version before execution. What the agent changes is what the attorney is actually reviewing: after the pattern-matching pass has surfaced the substantive issues, the attorney works on those rather than doing the checklist pass themselves.
Every deliverable is reviewed by the principal before it goes anywhere. The engineer owns the judgment on each finding, the negotiation strategy, and the final decision to execute.
Where it applies
Any contract a consulting engineering firm receives from a client. Developer agreements for multifamily and commercial projects. School district and public agency professional services agreements. Housing authority contracts. City and county agreements. Subconsultant agreements where the firm is sub to an architect or a prime engineer.
The agent is built for the contracts that come in most often: standard forms modified by the client’s counsel to favor the client, where the same three or four clauses are problems in every version and still take two to three hours to document each time.
It’s also useful on master service agreements that govern multiple projects. A single review of the master terms, documented with the specific findings and what was negotiated, gives you a defensible record before the first task order is executed.
We run it on every contract at Calichi before any principal touches it.
What this means for your firm
I’m not going to tell you it’ll save your firm two to three hours per contract review. That’s what we measured at Calichi, on our contracts, with our standards. Your number depends on your volume, your current process, and how your principals work today. The honest way to find it is to measure it on your real agreements rather than trust a projection from someone else’s firm.
That’s what the first two gates of how I deploy are for. An AI Readiness Audit to find where the highest-value automation sits across your firm, then a Strategic AI Discovery that maps your actual workflows and produces the hours-saved picture by workflow. Every gate is a stop or go. You only go further once you’ve watched the last step work on your own projects.
I built this inside my own firm before offering it to anyone. Contract review was one of the clearer cases: a task that recurred on nearly every project, ran two to three hours every time, and was almost entirely pattern recognition rather than engineering judgment.