Ask yourself where your last set of drawings went after it left the project folder. If the honest answer includes a browser tab that nobody in leadership has ever seen, you’ve met shadow AI.
Foundations First is our series on AI readiness for AEC firms, and this post starts with the rules. AI has arrived in AEC, and the firms that gain the most from it start with the groundwork every project needs.
A clear AI policy for architecture firms and sensible access controls protect your drawings and IP before your team’s habits write the rules for you.
What’s at Stake in AEC Firms
Your drawings are your intellectual property. Models and details represent years of design thinking, and clients pay for the exclusive use of that work. But once a drawing set lands in a public AI tool, you lose control over where it goes next.
Client contracts add another layer. Many owners write confidentiality clauses into their agreements, and some send security questionnaires before they award work. One pasted spec can put you in breach of terms you signed in good faith.
Reputation carries a lot of weight in AEC. Your firm grows through referrals, and owners and contractors tend to know each other well. When a story about leaked project data gets out, it spreads quickly, and the trust it damages can take years to win back.
A leak would put several things at risk for your firm, and the list below shows where:
- Drawings, models, and details that represent your firm’s intellectual property
- Confidentiality clauses and security obligations written into client contracts
- Sensitive project data, including site plans and pricing details
- Your reputation with the owners and partners who refer work to you
Shadow AI: Why Banning It Fails and Ignoring It Costs More
Shadow AI means staff are using free, fast AI tools that the firm hasn’t reviewed or approved, usually because the tools help. It might be something as simple as a project architect drafting a meeting summary or an estimator tidying a scope description, just to win back some hours in a busy week.
AI use at work is already mainstream. Gallup’s May 2026 survey of US employees found that 15% use AI daily in their role, while 30% use it a few times a week or more. Some of that use will sit outside anything your firm has approved.
A blanket ban pushes the habit out of sight. People keep using the tools on personal phones and home laptops, and the firm loses all visibility. But ignoring the issue is worse, because the firm carries the risk without any say in how it’s managed.
The scenario unfolds on a busy Friday, with a bid deadline close. All it takes is a team member pasting part of a client brief into a free tool to speed up the response, and while nothing looks wrong at the time, the brief now sits on a server the firm doesn’t control.
The better path gives people clear rules and a safe place to work. When staff know what’s allowed, most will follow the guidance gladly. That’s the foundation for everything else in this series.
What a Sensible AI Policy Covers
A good AI policy for architecture firms fits on two pages that your team will read and remember. Short, plain wording gives people a clear answer when they’re busy, which is when shadow AI tends to creep in.
Cover these four points, and protecting drawings and IP gets much easier:
- Approved tools. Name the AI tools your team can use for firm work, and say which ones are off limits.
- What data stays out. Spell out what never goes into an AI tool, including unreleased drawings, client contracts, personal data, and anything covered by a confidentiality clause.
- Who decides. Give one person or a small group the authority to approve new tools and answer questions as they come up.
- How often it’s reviewed. AI changes quickly, so set a review date every six months to keep the policy current.
Write the policy in plain language and draw on examples from your own projects, so every rule feels familiar. Ask a few project leads to read the draft before you share it. They’ll spot gaps you’d miss, and they’ll be more willing to back a policy they helped shape. Then add it to onboarding so every new hire sees it in their first week.
Access Comes First
A point that often surprises many principals is that an AI policy only works when there are access controls to back it up. If anyone in the firm can open any project folder, any connected AI tool can open it too. AI assistants work with the files a person can reach, so loose permissions mean loose AI.
Start with the folders your firm uses regularly:
- Set permissions by project, so only the project team sees live drawings and models
- Set permissions by role, so finance and HR files stay with the people who need them
- Review who has access when a project closes or a team member leaves
- Require multifactor authentication (MFA) on every account
This is where cybersecurity and governance meet AI. Tidying permissions is routine housekeeping that pays off from day one, and it makes every later step easier. Firms that sort out access first tend to find the rest of the journey much smoother.
Safe Tools, Used Well
Once the rules and permissions are in place, give your team something good to use. Gartner’s 2026 survey of 12,004 employees and managers across 40 countries found that 88% of employees whose employer provides AI tools also use personal AI tools for business tasks. Approved tools work best when a clear policy and some training sit around them.
Microsoft 365 is a strong starting point for many AEC firms because most people work in it every day. Its AI features operate inside the permissions and protections you’ve set, which is why the access work above matters so much. With the right controls in place, those features can help with meeting minutes and RFI logs while your project data stays inside your own environments.
The right setup usually includes:
- Sensitivity labels on confidential project files
- Data loss prevention rules that flag sensitive content before it leaves the firm
- Conditional access, so only trusted people on trusted devices get in
- A short training session so staff know how to use the tools well
Build Your Foundations First
Book your free AI Readiness Foundations Consultation with Scott Davison, Chief Revenue Officer at Plexus.
This short, practical conversation looks at the data, security, governance, apps, and people foundations that AI depends on. You’ll receive a written readout within 48 hours, so you can see where you stand and what to tackle first.
FAQs
- What should an AI policy for architecture firms include?
A good AI policy for architecture firms names the approved tools, lists the data that must stay out of AI tools, says who approves new tools, and sets a regular review date. It should also reference your client confidentiality obligations and use plain language. - What is shadow AI, and why does it matter for AEC firms?
Shadow AI is the use of AI tools that the firm hasn’t approved or reviewed. It matters because staff may paste specs or client emails into public tools, which puts client confidentiality and your IP at risk. - How do we protect drawings and IP when staff use AI?
Set permissions by project and role, and require MFA on every account. Add sensitivity labels to confidential files and give staff approved tools so they have a safe option. - Should we ban AI tools at our firm?
A blanket ban usually pushes AI use onto personal devices, where you can’t see it. A clear policy paired with approved tools works better, because people know what’s allowed and have a safe place to work. - Do we need an AI policy if we have a small team?
A firm of 20 people carries the same client confidentiality obligations as a firm of 200. A short policy and sensible permissions take little time to put in place and protect the work your reputation rests on.