Where does your team stand on AI?
Find your team's AI maturity level, what unlocks the next one, and the workflow worth starting with. One department, your own estimates.
FreeNo loginArchitects, engineers, contractors, owners, firm leaders
Your level and your next move
AI changes one department at a time, so you answer for one. You get your maturity level, what unlocks the next one, and the workflow to start with.
Six levels. Each one has to be unlocked.
Every level says what has to be true before you use it, what you gain, what breaks, and who signs. Nobody in AEC runs Level 6 yet.
One decision and nine steps
The process we run inside firms and teach in Mastering AI in AEC. Policy and education come before tools: clarity speeds adoption, fear stalls it. Each step produces one artifact. Each card says what happened when we ran it.
How AI fits a workflow: the Fit Ladder
Every workflow sits on one rung. The rung says what the human still does and how much checking the output needs. Examples for Architecture.
| Rung | Example | Best for |
|---|
Verify, then sign: what the person who sends it checks, Architecture
If the assistant cannot say where a claim came from, the claim does not ship.
Write the policy first
Clients ask for it before they sign. Your people are already using AI. One page fixes both, and it is the step firms skip on the way to buying licences.
Approved tools, on which accounts
Business or enterprise accounts with model training off. Personal free accounts are not used for client work. Anything not on the list needs approval before first use.
What never goes in
Client-confidential material, anything under NDA, government or restricted project data, personal data of staff or clients, unpublished financials. Structure without identity when the work is sensitive.
Verify, then sign
AI output is a draft. Any number, standard reference or contractual language is checked against the source by a qualified person before it leaves the firm. The person who sends it owns it.
Disclosure
Where AI materially produced client-facing work, you disclose when asked and always answer honestly. AI-generated imagery in presentations is labelled.
Who approves new tools
One named person, after a sandbox test on non-client material, with a response time you write down. Not a committee.
Review
Quarterly. Tools change monthly; a policy that stands still is already wrong. Name the next review date today.
The eight questions (answer them out loud)
- Which AI tools do your people already use for work, officially or not?
- Which of those will the firm pay for and stand behind, on proper business accounts?
- What data never leaves your network?
- Who checks AI-assisted work before it goes to a client, and do they know it is their job?
- Your position on AI-generated imagery in client presentations: allowed, labelled, or banned?
- Who is the one person who approves a new tool, and how fast must they answer?
- What happens when someone breaks the policy the first time? Retraining beats punishment; scared people hide their AI use.
- When will you review this policy next? Name the date now.
Get the Policy Kit
Template, our filled example, the eight questions and the drafting prompt. One PDF. Firms over fifty people add the long form: tool register, data classes, sign-off table, incident handling. Not legal advice; your counsel reviews before adoption.
The process does not change with size. The artifacts do.
What changes is who owns each step, which assistant tier is safe, how learning is organised, and how long rollout takes. Inside every size, it still moves one department at a time.
Working with firms from five to five thousand people, this is what we see
Firms want to start at the demo. The demo is Step 6. The firms that move start with a sentence about why, and a page about what is allowed.
Publishing a one-page policy increases AI use. People were holding back because nobody had said what was allowed.
AI reveals fragmentation; it does not fix it. The stack map finds tools nobody uses and workflows nobody wrote down.
The wins that stick are boring. Search, QA/QC, takeoffs, proposals, meeting notes. Design automation comes later, if at all.
One owner beats a committee. A single gatekeeper through whom every request routes is the near-death pattern.
Deadlines are the benchmark. Tools that survive a live project are real; tools that survive a demo are a demo.
The first return is cognitive: less rework, faster decisions, fewer interruptions. The counted savings come after, and they are labelled projected until measured.
Meaningful adoption takes one to three years. Department by department works. Everyone at once does not.
Do one thing this week
Where this comes from
Everything on this page comes from implementation work inside architecture, engineering, interiors, modular and construction firms, and from teaching it to more than a thousand professionals.
Know where you stand. Know what is next.
Find your level. Start the right workflow. Write the policy this week.