The concept phase of an architectural project is the most underpriced, most overworked, and most uncertain phase of the project lifecycle — and AI has made the economics of it structurally worse, not better, for most practices that haven’t adjusted. Compressing the production time of concept work should have freed practices. In many cases it hasn’t. This article lays out the numbers, the incentives, and what “broken” actually means in 2026.
What is the concept phase supposed to do?
The concept phase exists to answer a single question: is there a design direction worth developing?
Everything that follows — design development, construction documents, permits, bidding, administration — is expensive and irreversible. Starting construction documents on a concept the client will reject or the site will not support is the single most expensive mistake a practice can make.
So the concept phase is where judgment concentrates. Site reading, program analysis, initial form, palette, material direction. Enough to know: yes, this direction is worth investing in, or no, we need to go a different way.
Historically, this phase represented roughly 15-25% of total design fee on residential work, and a similar proportion on commercial. Not because the concept produced 20% of the output — it didn’t — but because it required senior judgment on site, program, client, and form. The value was in the judgment, not the hours.
How the Numbers Worked Pre-AI
Typical residential concept phase fee math, pre-AI (2020-2022, USD, typical mid-market practice):
| Item | Hours | Rate | Cost |
|---|---|---|---|
| Senior architect — site, brief, judgment | 20 | $200 | $4,000 |
| Designer/junior — sketching, variants | 60 | $80 | $4,800 |
| Drafter — plan variants | 20 | $60 | $1,200 |
| Outside renderer — 3 concept images | — | — | $3,000 |
| Printing, presentation prep | — | — | $500 |
| Total internal+external cost | 100 hours + vendors | — | $13,500 |
| Fee charged to client | — | — | $18,000-25,000 |
Margin on concept phase: roughly 25-35% on paper. In practice, often lower because revision rounds weren’t fully priced in.
Two things to notice. First, most of the time and cost was not the senior judgment — it was the junior staff grinding through variants and the outside renderer producing the images. Second, the client paid for a package that was mostly production.
This structure was stable for decades.
How AI Broke the Numbers
Starting around 2023 and accelerating through 2025-2026, AI tools compressed the production component of concept work dramatically.
Same residential concept phase, post-AI, at a practice that has integrated tools like Nuit, Midjourney, Maket, and Nano Banana:
| Item | Hours | Rate | Cost |
|---|---|---|---|
| Senior architect — site, brief, judgment | 20 | $200 | $4,000 |
| Designer — AI-driven concept production | 12 | $80 | $960 |
| Drafter — plan variants (AI-assisted) | 4 | $60 | $240 |
| AI tool subscriptions (allocated per project) | — | — | $30 |
| Outside renderer — zero for concept | — | — | $0 |
| Printing, presentation prep | — | — | $500 |
| Total internal+external cost | 36 hours + subscriptions | — | $5,730 |
The production cost of concept work dropped roughly 55-60%. The senior judgment hours are unchanged — those are where the actual value lives. The junior-hour and outside-renderer components collapsed.
This is the mechanical shift. Now the economic shift.
The Fee That Didn’t Fall
A practice that cut its concept production cost by 60% should, in a competitive market, see fees reprice downward. Some did. Most didn’t.
The practices that kept the old fee structure:
- Saw concept-phase margin expand from 25% to 60%+ on paper.
- Lost that margin quickly to underpriced technical phases, because they hadn’t rebalanced the fee structure.
- Found that informed clients — the ones who understand the tool landscape — started negotiating harder on concept fees.
- Started losing bids to practices that priced concept honestly and rolled the savings into either lower total fees or proportionally higher technical-phase fees.
The practices that cut concept fees aggressively:
- Looked cheaper on paper; won more work.
- Ended up with more projects in the concept pipeline than their technical team could handle.
- Hit capacity bottlenecks in design development and construction documents, where AI doesn’t help nearly as much.
- Some ended up in worse shape than practices that didn’t change anything.
Neither response is correct alone. The correct response is restructuring the fee breakdown — less for concept in percentage terms, more for technical phases where the time and risk actually live.
Few practices have done this well.
The Capacity Trap
Here’s the structural issue that makes the economics worse.
Concept-phase production time compressed by 60%+. Technical-phase production time compressed by maybe 10% (some AI help with drafting, documentation, but nothing like the concept-phase gains).
So a practice that used to carry six concept projects and three technical projects can now carry fifteen concept projects — but still only three technical projects.
What happens?
The pipeline backs up. Concept work is approved. Projects enter the technical phase. The technical team is overwhelmed. Delays cascade. Clients wait. Invoicing slows. Cash flow suffers.
This is the capacity trap. AI gave practices a bigger front door and didn’t change the size of the back door. Practices that leaned into the front-door leverage ended up with logjam.
The fix requires one of:
- Deliberately capping concept-phase throughput to match technical-phase capacity.
- Hiring more technical-phase staff (CAD/BIM specialists), which is actually harder in 2026 because the traditional junior pipeline that produced those staff has been disrupted.
- Outsourcing technical-phase work, which practices generally resist for quality-control reasons.
- Partnering with executive architects for construction administration, which some practices are doing.
None of these are trivial.
The Training Pipeline Collapse
This is the deeper and more long-term issue.
The traditional architectural career ladder was: junior drafter → designer → project architect → senior architect → principal. The middle rungs — drafter and designer — historically did the work that AI now absorbs. Rendering. Plan variants. Concept exploration.
If AI does that work, practices hire fewer juniors. Fewer juniors means a smaller training pipeline for the senior judgment roles that AI cannot do.
In 2026, this shows up as:
- Architecture schools are turning out graduates for roles that are shrinking.
- Mid-career architects who trained in the old ladder find themselves in practices that no longer need the work they did on the way up.
- Senior architects approaching retirement are starting to see a talent gap behind them.
- A few practices are actively redesigning their training pipelines — pushing juniors into site work, client work, and technical work earlier. This is hard and few have solved it.
The industry hasn’t absorbed this yet. The economic fallout is still in front of the profession.
Why don’t clients want the compression they think they want?
Clients often assume that if AI makes concept work faster, they should pay less and get their concepts faster.
The second part is true. The first part is complicated.
What clients pay for in concept work is:
- Senior judgment on site, program, and form.
- Time to do that judgment well.
- Enough visual exploration to make a confident direction choice.
- Risk absorption — the architect is on the hook for the direction being viable.
Of those four, only “visual exploration” compressed meaningfully. Judgment still takes time. Site still has to be read. Program still has to be thought through. The architect’s professional liability on the direction is unchanged.
So a client who expects “AI made it faster, so it should cost 60% less” is confusing production compression with total-work compression. A practice that obliges that expectation ends up underpricing senior judgment and overworking senior staff.
Good client education on this is rare. Practices that do it well tend to retain margin.
The New Bottleneck: Client Decision Speed
An unexpected consequence of AI compression.
Pre-AI, concept approval rounds took weeks because production took weeks. Clients had time between presentations to consult spouses, stakeholders, investors — the pace of the project matched the pace of human decision-making.
Post-AI, concept production takes hours. Practices that try to match their pace to their new production capability find clients are the bottleneck. Clients can’t always turn around feedback in a day. Consulting stakeholders takes time. Deliberating on a direction takes time. The decision-speed of humans hasn’t changed.
This changes how concept phases run. The efficient practice doesn’t try to rush the client — it uses the production headroom to show more directions, iterate more completely, and absorb revision rounds without pain. The pace stays human; the variety expands.
Practices that try to accelerate client decisions to match their production pace tend to produce worse outcomes.
What a Healthy 2026 Concept Phase Looks Like
A residential concept phase, at a practice that has adjusted:
- Senior architect runs the brief, site, and judgment work. Unchanged.
- Concept production compressed from weeks to days.
- Multiple directions shown — three to five instead of one.
- Deeper iteration per direction — two or three revision rounds without pain.
- Concept package extends to exterior + plan + interior, all coherent. Pre-AI this would have been cost-prohibitive.
- Fee structure rebalanced: concept fee at 10-15% of total rather than 20-25%. Technical and administration fees adjusted upward to match their true proportion of effort.
- Client experience: faster, more visual, more options, clearer direction choice, better-priced.
- Practice experience: more margin pressure on concept, more value creation downstream, more throughput, more discipline needed on capacity.
This is what the compression made possible. Very few practices operate this way end-to-end in 2026. Most are partially adjusted.
What should practices actually do?
Concrete suggestions, drawn from practices that have adjusted well.
Rebalance the fee structure honestly. If concept production compressed, the fee breakdown should reflect it. Don’t pretend the old allocation still holds.
Hold the line on senior judgment value. Concept fee compressed, but the judgment hours didn’t. Price senior time visibly — retainer, hourly, whatever — so the judgment value doesn’t disappear into a discounted concept fee.
Invest in technical-phase capacity. This is where the bottleneck lives now. Hire CAD/BIM specialists, outsource selectively, or cap pipeline throughput. Don’t let concept throughput overwhelm technical capacity.
Educate clients on what they’re paying for. The production-vs-judgment distinction is worth communicating. Clients who understand it are more comfortable with the fee structure.
Restructure junior training. If juniors aren’t doing rendering and variant grind, give them something else — site visits, client meetings, technical work, spec writing. Otherwise you’re hiring juniors who won’t develop into senior architects.
Use AI for variety, not speed. The best use of the compression isn’t to deliver concepts faster. It’s to deliver better concepts — more variety, more iteration, better direction choice. Clients value the quality lift more than the speed lift.
Be transparent about AI use. Clients in 2026 assume AI is involved. Pretending otherwise damages trust when discovered. Owning the workflow openly is the durable answer.
Related reading
- What Is a Design Concept Package? — A design concept package is the collection of visual and written materials that…
- How to Create a Complete Design Concept Package in One Day — A complete design concept package — exterior concepts, a floor plan, and interior…
- AI for Small Architecture Studios 2026 — A three-person architecture studio in 2026 can produce concept work at the volume a…
- AI Doesn’t Replace Architects — It Moves the Bottleneck — The question ‘will AI replace architects’ is asked often and answered badly.
- State of AI in Architecture: 2026 Annual Report — By mid-2026, AI tools are routine in architectural concept exploration, atmospheric…
Frequently Asked Questions
Is the concept phase really broken, or just changing?
Both. The economics changed structurally with AI, and most practices haven’t adjusted their fees, their pipeline capacity, or their training models. “Broken” means the structure is out of sync with the current reality. Practices that adjust fully return to a healthy state — just a different one than pre-AI.
Should concept-phase fees drop?
In percentage terms, yes. The production work inside concept compressed substantially. The proportion of total fee allocated to concept should reflect that. This doesn’t necessarily mean total fees drop — technical phase fees may need to rise proportionally to reflect where the real effort and risk live.
Why do some practices still charge pre-AI concept fees?
Because they haven’t adjusted — and because clients haven’t always pushed back. Informed clients in 2026 do push back. The old fee structure is unstable; practices that haven’t repriced will face pressure over the next few years.
What’s the biggest risk in the post-AI concept phase?
The capacity trap. Practices lean into faster concept production, take on more work, and hit a bottleneck in the technical phase. The result is delayed projects, unhappy clients, and cash flow problems. Technical-phase capacity needs to match concept-phase throughput.
Should small practices hire differently now?
Yes. The traditional junior-grinds-through-rendering training model is fading. Practices that still hire juniors should give them meaningful technical, site, and client work earlier. Practices that can’t do this should consider hiring experienced CAD/BIM specialists directly.
What happens to outside renderers?
Demand for concept-stage rendering has dropped sharply. Demand for high-end hero renderings for completed projects (marketing, awards, portfolios) still exists but is smaller. Rendering studios that survived specialized in the high end or pivoted to architectural visualization consulting.
How should clients think about concept fees in 2026?
Ask what percentage of the fee covers senior judgment (site, program, direction) vs. production (rendering, variants, drafting). The former is where the value is; the latter has compressed. A fee that reflects this honestly is more defensible than one that treats concept as a black box.
Try Nuit free — 10 generations, no card required. Run a full multi-direction concept package through Nuit and see what post-AI concept economics actually look like for your practice. Start your project →