AI generates a code-compliant building design in minutes. Clients are asking what they’re paying you for.
What’s happening
Tools like TestFit, Maket, and Archistar let developers generate site plans, unit mixes, and feasibility studies before calling an architect. TestFit optimizes site layouts by zoning, parking, and unit count in seconds. Maket generates residential floor plans from natural language, marketed as requiring “no architecture degree, no CAD software, no learning curve.” Archistar, used by Brookfield, Mirvac, JLL, and 25+ city councils, automates compliance checks for investors. By the time the client calls an architect, they’ve already made design decisions.
The impact is most visible in multifamily residential and commercial office, where design parameters are well-defined and AI optimization is straightforward. Autodesk Forma generates massing studies across thousands of options, with real-time environmental analysis for sunlight, wind, energy, and noise. These tools produce schematic-level output, not construction documents, but that’s enough for a developer to evaluate a site, choose a direction, and reduce the scope of what they need an architect for.
The question from clients is becoming more direct: if AI generated three design options that meet code, what’s the architect adding? The answer is real: design judgment, spatial quality, user experience, community context. But it’s never been separately priced. It was always embedded in the design fee, invisible to the client who saw only the drawings.
Why the obvious responses don’t work
“Emphasize design quality and creativity”
Design quality is subjective and difficult to price in a proposal. When a client has three AI-generated options that meet their program requirements, arguing that yours is “better designed” without a measurable outcome is a losing pitch.
“Adopt the AI tools yourself”
Using the same tools the client already used doesn’t create differentiation. It makes you faster at producing what the client can already get. Speed is valuable, but it isn’t a new offering. It’s a cost reduction.
“Focus on complex building types AI can’t handle”
AI design tools are expanding from simple to complex building types every year. Healthcare, education, and mixed-use, today’s complex types, are next quarter’s solved problems. Retreating to complexity is a shrinking island.
What’s working instead
Separate design judgment from design production. Offer pre-design strategy as standalone fixed-fee services before any design begins: programming, site selection advisory, community engagement, entitlement navigation. Design becomes one phase of a broader engagement, not the entire engagement. When the client arrives with AI-generated options, the response is not “let me redesign this” but “let me tell you which one to build and why.”
The pattern is the same across every firm that gets this right: they stop optimizing the old model and build new offerings around what AI cannot do. The Workshop is the facilitated day we do this work with you. You leave with 2–3 offering briefs, specified and priced. Your team tests them with named clients, then builds what earns it.
Other pressures on Architecture Firms
The same pressure in other industries
Related reading
McKinsey Put Numbers on Your Repricing
Your clients have decided AI should make your work cost less. McKinsey’s QuantumBlack report puts 32 pages of evidence behind that and prescribes a Fortune-500 cure you don’t need.
YC Published the Playbook for Replacing Your Firm
Y Combinator is teaching founders to build AI-native firms that sell outcomes in tax, audit, law, and insurance. Its four targeting criteria double as a diagnostic for which of your services are exposed.
$15,000
A full day with your senior team, then 2–3 offering briefs. Test what clients will pay for before you build.
Book a conversation30 minutes with Shawn Yeager. No pitch.