This essay examines knowledge labour and hiring in architecture through a personal account of a layoff after five years in an office, difficulty finding real work since, and similar experiences among personal contacts. It asks whether readily available knowledge changes what a practice is willing to pay another employee to do, and considers project demand, production, judgment, and how future architects learn their work.
My viewpoint · Personal account, edited for clarity
Five years in an office, then no real work
I worked in an architecture office for five years before I was laid off. Since then, I haven’t been able to find real work. A lot of people I know are in the same position right now.
I see a trend. I believe part of it is that knowledge is now readily available, and that changes the value proposition of hiring someone.
Why pay someone $4,000 a month when you can get the same output from a $100-a-month subscription?
Codex · AI analysis
The knowledge inside the work
Architecture is knowledge labour made physical. Before a building exists, someone must interpret a brief, understand a site, compare possibilities, coordinate systems, and communicate decisions. Drawings, models, schedules, and specifications carry that knowledge into construction. Their value depends on what they allow other people to understand and do.
The comparison puts pressure on a familiar part of that value proposition. A practice has often paid for access to information and for the labour of applying it. When AI makes some of that information easier to retrieve, organize, and turn into a useful deliverable, those parts of a role become easier to obtain without hiring. The question shifts toward the contribution that remains scarce within that particular team.
Some of this knowledge is easy to record: a dimension, a product specification, an office standard. Other knowledge develops through experience: which assumption needs challenging, when a consultant’s answer is incomplete, or why a seemingly minor change could unsettle the whole project. A practice relies on both.
Hiring brings these forms of knowledge together. A junior employee may contribute production capacity while developing judgment. An experienced architect may make fewer visible outputs while preventing expensive mistakes and helping others work effectively. Counting drawings or billable hours captures only part of what either person contributes.
Codex · AI analysis
AI changes the cost of a first pass
Generative AI can assist with parts of the work between receiving information and making a decision: summarizing supplied documents, organizing a brief, drafting correspondence, or producing material for comparison. These tasks consume time even when the person doing them already knows what the project needs.
Consider a practice preparing for a client meeting. Someone gathers the brief, previous notes, and design options, then assembles a discussion document. An AI tool could help organize that material into a draft agenda, comparison, and list of unresolved questions. An architect still has to determine whether the comparison is fair, the questions are relevant, and the proposed direction makes sense. The potential saving lies in reaching that discussion sooner.
Research on knowledge work supports a task-specific view of these gains. In an experiment involving 758 management consultants, AI assistance improved speed and quality on tasks within the system’s capabilities, but made participants less likely to reach correct answers on a task outside them. These were consulting tasks; the study does not establish equivalent savings on architectural projects.
The essay itself provides a small example. A person supplied the subject, experience, and direction; an AI researched and wrote the article and adapted it for social media. The person commissioning it has been reviewing the result and changing the brief. The work includes human judgment about what the article should say, even though its drafts were produced by AI.
For a practice, the useful measure is the time required to reach a dependable result, including briefing, checking, and correction. A fast draft that takes hours to repair has limited value. A reliable first pass that gives an architect more time to examine the design can be valuable even if it changes no staffing decision.
Codex · AI analysis
The economics of another hire
The comparison identifies a real hiring incentive: where a low-cost tool can produce a comparable, usable result, a firm has less reason to buy additional labour for that task. The comparison depends on the full cost of reaching that result, including the existing team’s time spent supplying context, checking accuracy, and carrying the work through. The figures illustrate the commissioning brief’s comparison, rather than a quoted salary or product price.
When a tool reduces the labour required for a task, the employment effect depends on what the practice does with the capacity it gains. It might accept another project, investigate more options, shorten delivery, or improve its margin. It might also postpone a hire. The same productivity improvement can produce different outcomes in different offices.
Project demand remains central. The U.S. AIA/Deltek Architecture Billings Index for August 2026 was 47.2, below the threshold of 50 that indicates unchanged billings. Newly signed design contracts also weakened. Such conditions give firms a reason to be cautious about payroll, independently of AI.
The staffing evidence also resists a simple replacement story. In AIA’s May 2026 survey, 88% of responding firm leaders reported that AI had no effect on staffing over the preceding year. In the same survey, 23% described their firms as understaffed. These findings do not prove that AI will leave employment unchanged; they do show that widespread replacement is an inadequate explanation of the surveyed conditions.
The plausible longer-term change is in the amount and mix of labour a commission requires. If a team can complete more routine work with its existing staff, some projects may generate fewer additional positions. If faster production makes previously uneconomic work viable, opportunities may expand. Neither outcome follows automatically from installing a tool.
Codex · AI analysis
Attention becomes a constraint
A practice can produce more information than it has time to evaluate. More options, more images, and more drafts can increase the burden on the people responsible for deciding what belongs in the project. The constraint then moves from production to review.
This matters because an architect’s attention is already divided among design, coordination, administration, and supervision. Delegating a task requires explaining it, answering questions, inspecting the result, and resolving differences in interpretation. AI can shorten some of these exchanges by making a draft available quickly. It can also create new ones when its output conceals an assumption or misunderstands the brief.
Clear briefs, accessible project records, and explicit review criteria therefore become more valuable. The knowledge in an experienced architect’s head has to be expressed well enough for other people and tools to use it. Doing that can make the whole office more effective, while revealing how much essential context was previously carried through informal conversation.
The benefit is architectural when the recovered attention returns to the project: examining a section, questioning a circulation strategy, or understanding how a material will age. Producing more material is useful only to the extent that it helps the team make and carry through better decisions.
Codex · AI analysis
The apprenticeship problem
The pressure on entry-level work deserves particular attention. Much of the work that might be accelerated also provides a route into the profession. Preparing a schedule, updating a drawing, and recording a meeting teach a graduate how information connects across a project. Repetition creates opportunities to notice consequences, receive corrections, and develop confidence.
If firms reduce these tasks without redesigning that learning process, they risk weakening the route through which future expertise develops. Expecting graduates to arrive with judgment that normally takes years in practice creates a gap that software alone cannot close. Senior knowledge has to be renewed through people who are given the chance to acquire it.
There is also an opportunity. A graduate who spends less time assembling a first draft could spend more time discussing why it works, comparing alternatives, or attending a coordination meeting. That outcome requires deliberate supervision and access to meaningful work. It will not happen simply because production becomes faster.
For someone returning after a layoff, experience can have value that is difficult to show in a portfolio of finished images. Recognizing an incomplete instruction, anticipating a coordination problem, and knowing when to seek another opinion are contributions to a team’s capacity. Hiring practices need ways to recognize them alongside technical output.
Codex · AI analysis
What a practice needs from its next employee
As the cost of generating information falls, a practice has reason to look more closely at what another person enables it to do. That includes production, but also dependable decisions, sustained responsibility, effective collaboration, and the ability to develop others. The value of a hire lies in the work the team can carry together.
These qualities depend on production experience. Judgment develops through making things, checking them, and learning why they failed. A useful response to AI is to reorganize that work so people keep learning while tools handle suitable tasks. An office that optimizes only for immediate output may save hours today while losing the capacity it needs later.
The intersection of knowledge labour and hiring is therefore a question about how architecture practices build capability. AI may let some teams start or finish work without another employee. Other teams may use it to make new hires more effective. The difference will depend on demand, the work involved, and how each practice chooses to use the time it gains.
That choice deserves as much attention as the technology itself. Time saved can become a smaller payroll, a healthier workload, better architecture, or an investment in the next generation. The tools expand what is possible; the practice decides what the gain is for.