Architecture, interior design and A/E practices

AI for architecture firms: beyond renderings

Your designers already use AI to make images. The bigger return sits in the rest of the practice: RFP responses, fee build-ups, office masters, QA reviews and submittals, where hours are lost and margin is won or leaked.

Start where fees are won and lost: proposals, fee build-ups and project burn.
Keep the architect of record in charge: every AI output checked, cited and owned.
Measure proposals out, hours returned and turnaround times, not logins.

Book a strategy call Have our engineers review your idea →

The landscape

Where AI stands in architecture

Most architecture firms now have AI somewhere in the office. RIBA's 2025 survey found 59% of UK practices using it, up from 41% a year earlier. Deltek's 2026 study of US and Canadian A&E firms puts adoption at 70%. In Canada, Statistics Canada reports that about a third of professional, scientific and technical services businesses used AI to deliver services in the past year.

Use is not the same as value. Only 38% of the A&E firms in Deltek's study could point to a measurable positive business impact. The AIA found only 8% of US architecture firms had implemented AI solutions in their practice, and only 6% of professionals used AI regularly. Only 15% of the practices RIBA surveyed have an AI policy.

Look at where it gets used. Among RIBA's AI-using practices, 70% use it for early design visuals and 89% for report writing, but only 29% use it for fee calculation and 28% for project cost management, the work that decides margin. The AIA saw the same gap: architects rate their firms least efficient at updating office masters, cost estimating and writing complex specifications, yet few have pointed AI at those tasks.

The pressure behind all this is real. 69% of RIBA respondents say current fee levels are unsustainable given project complexity. Deltek reports A&E utilization just under 60% and staff turnover above 13%. That's the context for your AI plan: fewer hours to sell, more complexity per project, and people leaving faster than their knowledge gets written down.

59%

of UK architecture practices use AI, up from 41% in 2024

RIBA AI Report, 2025 ↗

15%

of practices have an AI policy

RIBA AI Report, 2025 ↗

38%

of A&E firms report a measurable positive business impact from AI, while 70% use it

Deltek Clarity A&E Study, 2026 ↗

32.4%

of Canadian professional, scientific and technical services businesses used AI to deliver services in the last 12 months

Statistics Canada, 2026 ↗

Why generic advice fails

What makes architecture different

Generic AI advice assumes a sales floor, a call centre and lots of clean data. An architecture practice runs on something else, and the plan has to fit it.

You win work on qualifications

Public and institutional clients often use qualifications-based selection, so every RFP is scored on comparable projects, team résumés and approach. That content already exists, spread across old submissions, InDesign files and the memory of whoever wrote them. AI that helps here has to know your projects and your people. Generic marketing copy loses points.

The licensed professional owns the result

The OAA has said the member supervising the practice stays fully accountable for AI-assisted work, as if they had done it personally. The RAIC said the same in 2026: AI doesn't replace the architect's duty of care. So a code analysis, a spec section or a CA response from AI is a draft for a licensed person to verify, and the tool has to show where every answer came from.

Fees are fixed, hours aren't

Most of your work is a percentage or fixed fee by phase. Every extra hour in construction documents or contract administration comes straight out of margin, and scope agreed in a meeting becomes unbilled additional services. AI that shortens fee build-ups and shows burn early is worth more to you than AI that writes faster emails.

Your knowledge lives in office masters and old decisions

Master specs, standard details, QA checklists and lessons learned are the practice's real asset. In the AIA's research, 51% of architects called their firm inefficient at updating product libraries and office masters, yet only 4% had tried AI on it. When a project architect leaves, the reasons behind a hundred decisions leave with them.

Codes change by edition and jurisdiction

Ontario's 2024 Building Code took effect on January 1, 2025, alongside national energy codes and municipal requirements. A chatbot that answers "what applies here" without citing the edition and section is a liability, not a shortcut. Code research needs cited sources and a named reviewer, every time.

Where to start

Where AI pays off first

Start with work that is repetitive, document-heavy and close to fees. Each item below maps to a build we've delivered. Unless we say so, it wasn't built for an architecture firm, but it's the same work your practice does every week. See the full set in the Use Case Explorer.

Director of Marketing or Business Development

RFP responses drafted inside Microsoft 365, from your own past submissions

Marketing teams lose days rebuilding the same project descriptions, résumés and approach sections. This build puts drafting and people search inside the apps your team already has open, so a response starts from your firm's best past content, in your format. Answering more RFPs with the same team means fewer good pursuits passed over for lack of time.

A build we’ve deliveredProposal writing inside the tools you already useHundreds of hours a year of archive search returned · First release, then four extensions

See the build →

Principal leading pursuits

Which past projects answer this RFP's evaluation criteria?

Every RFP asks for comparable projects and the people who delivered them. The tool reads the selection criteria, ranks your past projects against them and produces project sheets ready to send. You can also ask in plain language, for example which project architect and interior designer have delivered a long-term care renovation together.

A build we’ve deliveredWhich of our past projects fits this bid?Asking around the office, to a search · About 3 months

See the build →

Business development lead

Public RFPs found and ranked while there's still time to respond

Municipal, school board, health and federal opportunities sit across many portals, and someone checks them when they have time. Automatic discovery ranks each new opportunity by fit with your sectors, in English and French, so go/no-go calls happen early and from one list.

A build we’ve deliveredFinding the opportunities you were missingBid sites watched now and then, to all the time · About 3 months

See the build →

Principals and Director of Operations

Fee build-ups from your own history of hours

Principals still build fee proposals by typing hours into an old spreadsheet, phase by phase. The same build we delivered for estimators predicts hours per task from past jobs, with a confidence rating and version history on every estimate. Your fees become consistent, easier to defend to a client and grounded in what similar projects actually took.

A build we’ve deliveredEstimates without starting from a blank spreadsheet15+ senior hours a week returned · About 4 months

See the build →

Controller or Director of Finance

Fee burn and WIP variance, reconciled in minutes

Most firms find a project in trouble at month end, after the hours are spent. Reusable Copilot prompts in Excel compare two months of project data, check that totals tie and flag the changes worth a conversation. Set up in one coaching session, it gives the controller and PMs the same early view of where fees are leaking.

A build we’ve deliveredMonth-end variance analysis in minutesMatching rows by hand, to a reconciled table in minutes · Set up in one coaching session

See the build →

Technical Director or QA lead

Office standards and QA procedures answered with the section cited

Junior staff ask the technical director the same questions every week: which detail, which master spec section, which QA checklist. An assistant limited to your approved standards answers with each part cited to its section, and refuses anything outside that set instead of guessing. Senior people get time back and everyone works from the current version.

A build we’ve deliveredStep-by-step procedures, cited to the sectionProcedure lookups without stopping work · Weeks, on the existing foundation

See the build →

Contract administration lead

Shop drawings and submittals checked against your drawings and specs

In contract administration, reviewers compare submittals with drawings and specifications sheet by sheet. An agent loaded with your review guide lists the mismatches for an architect to confirm. Built with an engineering team, it caught a real equipment discrepancy in its first session. The reviewer still owns the response. They just stop hunting.

A build we’ve deliveredDrawings checked against vendor submittalsSheet-by-sheet comparison, to a discrepancy list · Built in a coaching session, shared with the team

See the build →

Project architects and project managers

What did we agree with the client in March?

Scope decisions get made out loud in client and consultant meetings, then disputed months later when the fee is gone. Copilot answers from your recorded Teams meetings, with the context, instead of someone rereading minutes. That's how an additional services request rests on the record rather than on recollection.

A build we’ve deliveredAnswers buried in past meetings, found in secondsHours of rereading, to an answer on request · Done in a coaching session

See the build →

Browse all the builds in the Use Case Explorer →

The people side

Why AI adoption stalls in architecture firms

The licences are rarely the problem. These four patterns usually are.

1Designers try it on images and stop there

Your keenest users are often younger designers making concept images. Proposal writers, PMs and CA staff never see what AI does for their work, so adoption looks high and changes nothing. AI-First Moments workshops, where people bring a live RFP or submittal log, fix this faster than another lunch-and-learn. The AI Habit Builder then makes it stick over six weeks.

2Every senior hour is billable, so nobody owns AI

In a practice where partners and associates bill, AI becomes a side project for whoever is curious. Without a named sponsor and protected time, a pilot never reaches the second studio. Leadership sponsorship means one partner who sets two or three outcomes, reviews them monthly and clears the blockers.

3Nobody wrote down how the work is done

QA reviews, fee build-ups and RFP assembly live in people's heads and habits. AI can't help with a process no one can describe. Recording a walkthrough of the process once, and turning it into a written procedure, is often the fastest first step. Autodesk found 48% of design and make professionals say their organizations lack the resources to design internal training, which is why this rarely happens on its own.

4Usage gets counted, outcomes don't

Logins tell you who opened Copilot. They don't tell you whether proposals go out faster or submittals come back sooner. Measure cycle time and capacity created. An AI Adoption Diagnostic scores four scales (people capability, work design, AI's role and enterprise adoption) and names the one constraint holding AI back, so you fix that instead of buying more tools.

How our AI Transformation & Adoption practice works →

Your first year

A 12-month plan for the executive leading AI

Sequenced so each phase earns the next. Throughout, measure work won, hours returned and turnaround times, not logins. Our Three-Layer AI Playbook explains the thinking behind it.

First 30 days

Get the facts and set the guardrails

Run an AI Adoption Diagnostic across studios, marketing, finance and contract administration to find the constraint that matters most.
Publish a one-page AI policy: approved tools, what client data can go in, and the rule that a licensed professional verifies any output behind a design, code or contract decision.
Review Copilot permissions on project folders, fee proposals and HR files before you widen access.
Pick three outcome measures, for example RFP turnaround, hours per fee proposal and days to return a submittal.

Days 30–90

Prove value on one pursuit workflow and one delivery workflow

Hold an Immersion Day so partners and studio leads agree priorities from real examples.
Run AI-First Moments workshops where marketing, PMs and CA staff bring a live RFP, fee build-up or submittal log.
Build a ranked use-case backlog and send the top idea for an Engineer review.
Start the AI Habit Builder with one studio so daily use lasts past the workshop.

Months 3–6

Build the first custom tool

Build the highest-value tool in your own Microsoft environment. For most practices that's proposals or project references.
Connect it to the real archive: past submissions, project sheets, résumés and office masters.
Name AI Ambassadors in each studio to coach peers and report what isn't working.
Report to partners monthly on cycle time and capacity created.

Months 6–12

Extend into delivery and margin

Add a second build on the same foundation: fee build-ups, a standards assistant or submittal review.
Make fee burn and WIP variance a monthly AI-assisted review with PMs.
Update the policy as tools change and as the RAIC, the OAA and your insurer publish guidance.
Decide what to scale from measured adoption by tool and team, then set next year's backlog.

Guardrails

Risk, data and governance

Architects already manage liability, confidentiality and contracts with care. AI adds new ways for each to go wrong. Settle these four first.

The licensed member still owns the work

The OAA has been clear that the member supervising the practice remains fully accountable for AI-assisted work, and the RAIC's 2026 statement says AI doesn't replace the architect's duty of care. RIBA notes it is uncertain whether relying on AI output you know could be wrong would even count as covered negligence. In Ontario every Certificate of Practice participates in Pro-Demnity's mandatory liability insurance, so talk to your insurer before AI touches anything you seal. Build verification into the workflow and use tools that cite their sources.

Client confidentiality and your agreements

Plans for a hospital, courthouse or data centre, a developer's unannounced site, a family's home layout: none of it belongs in a public chatbot. RIBA warns that data entered into public AI tools can't be taken back and may breach contractual obligations to clients. Check your client agreements for confidentiality and security clauses, keep AI work inside your own Microsoft 365 or Azure tenant, and treat your design work as intellectual property worth protecting.

Copilot sees what your people can see

Microsoft states that Copilot grounds its answers in data users already have permission to access. If project folders, fee proposals or HR files are shared too widely, Copilot will surface them to anyone who asks. Fix oversharing and apply sensitivity labels before broad rollout, and keep walls between teams serving competing clients.

Privacy law and public-sector clients

Résumés, site photos and residential projects contain personal information, and Canada's privacy commissioners have said generative AI sits inside existing privacy law, not outside it. Ontario's Bill 194 (2024) lets the province set AI accountability requirements for public sector entities, including municipalities and school boards. Expect public clients to ask how you use AI on their projects, and have the answer in writing.

Client stories

Firms that bid and deliver like you

We don't have an architecture firm story to show you yet, and we won't invent one. These two clients win work through proposals and qualifications the way you do, and their builds are the ones described above. More in Client Stories.

“Our first build changed how we pursue work. Our proposal content, resumes, and project history now live in one place. Our team can research, draft and build submissions much faster than before, with better consistency. ForwardPath took the time to understand how our business actually bids, and we’re already looking at what’s next with them to find and qualify tenders earlier.”

Jerry RyndaVP, Business Development & Marketing, Buttcon

A general contractor whose proposal content, résumés and project history now live in one place: the same project-reference and résumé problem every architecture marketing team has.

“ForwardPath has delivered AI tools that our teams use every day. More departments keep adopting these tools, and ForwardPath has been a reliable, long-term partner every step of the way.”

Phil LangilleVP, Business Development, Rimkus

An engineering and consulting firm, a professional practice like yours, that automated proposal generation and tender discovery and keeps adding departments.

Read all our client stories →

Before you spend

Questions to ask before you spend another dollar on AI

Ask these of your own team, and of any partner who wants your budget.

Which three workflows cost us the most senior hours, and how do we know?

If the answer is a guess, start with a diagnostic, not a tool.

What did our last ten RFP responses cost, and how many did we pass on for lack of time?

Pursuit cost and capacity-driven no-bids are the clearest business case in most practices.

Where does client project data go when someone pastes it into an AI tool?

You need a one-line answer for every tool in use, including the free ones on people's phones.

Who verifies AI output before it touches a drawing, a spec or a code decision?

Name the role and the check. If everyone is responsible, no one is.

Is Copilot showing people files they shouldn't see?

Run an oversharing review before you widen licences, not after a fee proposal travels between studios.

Does our insurer know how we use AI?

Ask before AI contributes to sealed work. The answer shapes your verification rules.

Will we own the code and the data of anything built for us?

Custom tools should run in your own cloud, so they stay yours whatever happens to the vendor.

How will we know in six months whether this worked?

Agree cycle time and capacity measures now. Logins and prompt counts won't answer a partner's question about return.

FAQ

Frequently asked questions

How are architecture firms using AI today?

Mostly for concept images and writing. RIBA's 2025 survey found 70% of AI-using practices use it for early design visualisations and 89% for report writing, with 58% using it for bid creation. Far fewer use it for fee calculation or project cost management. The firms getting measurable value are moving AI into proposals, fee build-ups, office standards, QA and contract administration, where senior hours and margin are actually spent.

Can AI write architectural specifications?

It can draft them, but it doesn't make them more accurate on its own. Half of RIBA's AI-using respondents disagreed that AI had improved the accuracy of their specifications. The safer approach is an assistant that works only from your office masters and approved standards, cites the section behind every answer, and leaves the final edit to the specifier who is accountable for it.

Is it safe to put client drawings into ChatGPT or Copilot?

Not into public tools. Data entered into public AI services can't be pulled back and may breach confidentiality clauses in your client agreements. Use enterprise tools that run inside your own Microsoft 365 or Azure tenant, check that Copilot permissions don't overshare project folders, and write a short policy that says which data can go where. Sensitive projects, such as healthcare, justice or security work, deserve extra care.

Who is responsible if AI makes a mistake in an architect's work in Ontario?

The architect. The OAA has said the member supervising the practice remains fully accountable for AI-assisted work, as if they had done it personally, and the RAIC's 2026 statement says AI doesn't replace duty of care. Ontario practices carry mandatory professional liability insurance through Pro-Demnity, so confirm with them how AI use affects your coverage and build verification into every workflow.

What AI use cases pay back fastest for an architecture firm?

Usually the ones closest to winning work and protecting fees: RFP responses drafted from past submissions, matching past projects and teams to evaluation criteria, finding public opportunities earlier, fee build-ups from historical hours, and month-end fee burn reviews. They are document-heavy, repeated every week and easy to measure in hours and turnaround time, which makes the return visible to partners quickly.

How long does it take to see results from AI in an architecture practice?

Training and workflow coaching can change how a team works within weeks, especially with Copilot prompts for writing and reporting. A custom build such as a proposal or project-reference tool typically takes about three to four months to a first release. Agree the measures before you start, such as proposal turnaround or hours per fee proposal, so you can show results rather than describe them.

Start the conversation

Lead AI in your architecture practice with confidence.

Start with a conversation with our leadership team. We’ll look at where you are, where AI can move the needle first, and what your next 90 days could look like.

Book a strategy call Have our engineers review your idea →

Sources

RIBA AI Report 2025 (full report), Royal Institute of British Architects, 2025
Second RIBA AI Report shows surge in usage among UK architects, RIBA, 2025
RAIC Statement on Artificial Intelligence in Architecture, Royal Architectural Institute of Canada, 2026
RAIC issues statement on artificial intelligence in architecture, Canadian Architect, 2026
OAA Council Exploring Its Role in the AI Conversation, Ontario Association of Architects, 2025
Practice Advisory, September 19, 2025, Ontario Association of Architects, 2025
Professional Liability Insurance, Ontario Association of Architects, 2026
A Guide to Determining Appropriate Fees for the Services of an Architect, RAIC, 2019
Artificial Intelligence Adoption in Architecture Firms: Opportunities & Risks (The Architect's Journey to Specification 2024), American Institute of Architects with Deltek and ConstructConnect, 2025
New research explores perceptions and opportunities of artificial intelligence in architecture, American Institute of Architects, 2025
AIA AI Firm Toolkit, American Institute of Architects, 2026
Latest Deltek Clarity Industry Studies (47th A&E Study), Deltek, 2026
What the 46th Annual Deltek Clarity A&E Study Reveals, Deltek, 2025
Analysis on artificial intelligence use by businesses in Canada, second quarter of 2026, Statistics Canada, 2026
Autodesk 2025 State of Design & Make, Autodesk, 2025
Principles for responsible, trustworthy and privacy-protective generative AI technologies, Office of the Privacy Commissioner of Canada, 2023
Secure & Governed Data Foundation for Microsoft Copilot, Microsoft Learn, 2026
Bill 194, Strengthening Cyber Security and Building Trust in the Public Sector Act, 2024, Legislative Assembly of Ontario, 2024

Figures are quoted from the sources above as published. Use cases describe builds ForwardPath has delivered; client names appear only with permission.