Civil Engineering AI

AI for the work you sign and seal. It shows its work.

CivitasOPS is AI built for land surveyors and civil engineers, and it runs in the browser. It does the reading, the checking and the arithmetic that stand between a signed contract and the first billable hour. The AI reads the words; the numbers are computed from the text as it was read. Every finding shows you the line it came from, and anything it could not read cleanly is flagged rather than filled in.

Book a demo How our AI works

Every finding quotes its page Your own AI provider, your own key Built for a licensed professional's review

Three plans, one way of working

Every plan reads the same way and holds the AI to the same rules. What changes is how many people sign in, whose key the reading is on, and which jobs are open to your firm.

Standard

For a sole practitioner who already has an AI vendor account.

1 named seat

Documents a month: Monthly allowance
Chat questions a month: Monthly allowance
Reading on the CivitasOPS key: Your firm's own key

Chat: included
Land Survey: included
Field to finish: included
Permitting: not included
Reviews: not included
Plan QC: not included
RFP: not included
Construction Report: not included
Utility conflicts: not included
Project Management Dashboard: not included
SF-330: not included
Geotechnical: not included

No vendor account needed 02

Team

For a working team. Nobody at the firm has to open a vendor account.

5 named seats

Documents a month: Larger allowance
Chat questions a month: Larger allowance
Reading on the CivitasOPS key: CivitasOPS key included

Chat: included
Land Survey: included
Field to finish: included
Permitting: included
Reviews: included
Plan QC: included
RFP: included
Construction Report: included
Utility conflicts: included
Project Management Dashboard: not included
SF-330: not included
Geotechnical: not included

Premium

For a firm that reads every day. Everything, with the most headroom.

10 named seats

Documents a month: Largest allowance
Chat questions a month: Largest allowance
Reading on the CivitasOPS key: CivitasOPS key, higher allowance

Chat: included
Land Survey: included
Field to finish: included
Permitting: included
Reviews: included
Plan QC: included
RFP: included
Construction Report: included
Utility conflicts: included
Project Management Dashboard: included
SF-330: included
Geotechnical: included

We do not publish a price yet. Sign up for the plan that fits, and a member of our team will reach back within a working day with a quote and a 30-day pilot on a project you have already worked. Billed annually or monthly; the detail is on the product page.

Every project starts with a week nobody bills for

The pile of reading, checking and arithmetic a client never thinks they are paying for.

It lands on your most expensive person

It takes judgement, so it goes to whoever has the most of it — the person you had other plans for this week, doing work that cannot be handed to anyone junior.

It is slow in a way nobody can invoice

Careful reading, careful checking, and arithmetic that has to be right — priced into a fee that quietly assumed it would go faster than it ever does.

And when it goes wrong, it is silent

A mistake made in this week does not announce itself. It turns up later, in front of a client or a court, and costs a multiple of what the week did.

We built an AI to take that week apart — and built it so the person who signs can check every line of it. What a firm can hold us to is on the product page; the jobs themselves we show on a demo.

AI you can put a seal behind

Most AI asks you to trust it. Ours is built so you do not have to. Three rules, and the product is what follows from them.

Rule one

The AI reads. It never computes.

A language model locates and transcribes the words of your document — the calls, the clauses, the comments. Every bearing, distance, curve, closure, area and coordinate is then arithmetic on that text as it was read. The model produces no number; the reading is what you check, and it is shown to you.

Rule two

It quotes, or it is discarded.

Everything the AI reports has to point at the passage it came from. A finding it cannot quote is thrown away rather than shown. The licensed professional who reviews it is a glance from the page, not a leap of faith.

Rule three

It flags. It never fills in.

A reading the AI could not make cleanly is marked as uncertain and left for a person. There is no guess dressed up as an answer, which is the failure that costs a licence.

The whole of it, including what the AI is not asked to do, is on How our AI works.

Your AI provider, your key, your terms

We do not lock your documents to one AI vendor, and we do not read them on ours.

Your provider

Anthropic, OpenAI or Google

A firm administrator connects the firm's own account with any of the three, and switches whenever the firm chooses. Your documents are read under your agreement with that provider — or, on the Team and Premium plans, on the CivitasOPS key if your firm prefers.

Your files

Kept for your firm

Each firm's documents and projects are kept apart from every other firm's. Everyone at your firm sees the firm's work; nobody at another firm can. We do not sell your data, and nobody trains on it.

Your kit

Nothing to install

It runs in the browser your office already uses. No server to stand up, nothing to roll out to each computer, nothing new to buy. Sign in and bring a file.

How a job goes

Three steps, all of them in a browser tab.

Bring the file

Drop a deed, a review letter, a plan set or a solicitation into the project from your browser. It is read once, and kept with the project from then on.

Check the AI's work

Every measurement is arithmetic on the text as it was read, every finding points back at the page it came from, and what could not be read cleanly is flagged — so a licensed professional can review it and sign with a straight face.

Hand it over

Export it, print it, send it. The project is already where everyone else at the firm can find it.

Our AI reads your documents; it does not do your arithmetic, and it does not do your judgement. Every measurement we hand back is computed from the words of your document as they were read, every finding quotes the line it came from, and what we could not read cleanly we flag rather than fill in. It is research support for a licensed professional's review, not a survey. That is the whole pitch, and everything else is downstream of it.

— The CivitasOPS team

From the blog: AI in licensed practice

All posts →

14 Aug 2026

The worst kind of wrong

An answer with one thing missing looks exactly like a complete one. That is the failure we spend most of our care on.

See it on a parcel you already know

Bring one you have worked and can argue with. Half an hour, no slides, and you keep the output — and you will see exactly where the AI was right, where it flagged, and why.

Book a demo

Join us

3 open positions. We are a small team building applied AI for people who measure the ground.

See careers