AI
Outcome = time x knowledge. AI is collapsing both. Here is where value goes in AEC.
Tom Leyden · 16 October 2026
The standard answer to "AI is killing the billable hour" is "move to outcome-based pricing". Sell the result, not the time. It is good advice for the next two years. It is not a strategy for the next ten.
Here is the problem. For a knowledge firm, an outcome is mostly time multiplied by knowledge. If AI collapses the cost of both, it collapses the cost of producing the outcome. Once competitors have the same tools, competition pushes the price of that outcome down towards its new, much lower cost. Outcome pricing captures value only while the client's alternative is a firm still working the old way. That is an information gap, and information gaps close.
So the more useful question is not "how do we price?" It is "what is left in the outcome that does not collapse?" For architecture, engineering and construction (AEC), the answer is unusually clear, and it changes where firms should be heading.
Knowledge and time really are getting cheap
The evidence on the first two terms is now solid.
- In a preregistered experiment with 453 college-educated professionals, ChatGPT cut the time taken on writing tasks by 40 percent and raised output quality by 18 percent (Noy and Zhang, 2023).
- Across 5,172 customer support agents, AI assistance increased issues resolved per hour by 15 percent on average, with the largest gains for less experienced workers (Brynjolfsson, Li and Raymond, 2025). Expertise that used to take years to build is being handed to juniors on day one.
- In a field experiment with 758 consultants, those using AI on tasks inside the model's capability completed 12.2 percent more tasks, 25.1 percent faster, at more than 40 percent higher quality. On a task outside that frontier they were 19 percentage points less likely to get it right (Dell'Acqua et al., 2023). The gains are real and uneven, which is why judgement still matters.
- Around 80 percent of the US workforce could have at least 10 percent of their tasks affected by large language models, and around 19 percent at least half, with exposure higher in higher-wage, knowledge-heavy jobs (Eloundou et al., 2024).
None of this says knowledge stops being useful. It says the cost of producing knowledge work is falling towards the cost of compute. Susskind and Susskind (2015) predicted exactly this: professional work broken into tasks, most of which can be routinised, eroding the professions' bargain of exclusive expertise.
The full equation
An outcome in AEC is not just time and knowledge. It is closer to:
Outcome = Knowledge x Time x Physical execution x Risk bearing x Permission x Human coordination
| Term | Collapses with AI? | Why |
|---|---|---|
| Knowledge | Yes | Heading towards the cost of compute |
| Time (on knowledge tasks) | Yes | Agents work in parallel, around the clock |
| Physical execution | No | Materials, labour, site and energy are atoms, not tokens |
| Risk bearing | No | Someone has to carry it when things go wrong |
| Permission | No, and tightening | Registration, certification and approvals carry personal accountability |
| Human coordination | Mostly no | Agreement between people stays scarce |
The economics literature predicts what happens next. When a technology makes one input cheap, the value of its complements rises. Agrawal, Gans and Goldfarb (2018, 2019) frame AI as cheap prediction, which raises the value of judgement and action. Autor (2015) shows automation both substitutes for labour and raises the value of the tasks only people can do. Acemoglu and Restrepo (2019) show displacement is offset only when new tasks are created for people to do.
And Baumol's cost disease (Baumol and Bowen, 1966) tells us that sectors whose core work cannot easily become more productive get relatively more expensive over time. Construction is the textbook case. Global construction labour productivity grew about 1 percent a year over two decades, against 2.8 percent for the world economy and 3.6 percent for manufacturing (McKinsey Global Institute, 2017). In the United States, construction productivity has fallen for decades, and measurement error is probably not the only explanation (Goolsbee and Syverson, 2023).
That is the key to AEC. Knowledge is getting cheap. The physical parts of a project are not.
Why this matters more in AEC than almost anywhere
Consultant fees are a small share of what a project costs. There is no good peer-reviewed benchmark, but industry guides typically put them in single-digit percentages of total development cost. Even if AI drove every consultant fee to zero, the building would only get marginally cheaper.
The real money is in the physical and financial economics:
- Rework. Across 161 Australian projects, rework contributed 52 percent of project cost growth (Love, 2002). Studies commonly estimate direct rework at around 5 percent of construction cost (Hwang et al., 2009).
- Time. Every month on a project costs finance and holding costs, and delivers income later.
- Overruns. Large projects follow what Flyvbjerg (2014) calls the iron law: "over budget, over time, over and over again".
Most AI research in construction so far has gone into computer vision, language processing, optimisation and process mining (Pan and Zhang, 2021): tools that sharpen how work is done rather than change who captures the value.
So the value of knowledge has come apart from the cost of producing it. Knowledge is now valuable for what it changes in the physical and financial outcome: a month off the programme, rework avoided, contingency released. And that value flows to whoever carries the risk or owns the asset, because they are the ones who keep the savings.
The AEC value chain, mapped against the equation
Ratings show how much each stage's value depends on each term (H high, M medium, L low). They are my judgement, not measurements, and will vary by project type.
| Stage | Knowledge | Time | Physical | Risk | Permission | Coordination | Exposure to the collapse |
|---|---|---|---|---|---|---|---|
| Land, capital, demand (developer, owner) | M | L | L | H | M | H | Low |
| Finance and insurance | M | M | L | H | H | M | Low |
| Feasibility and investment advice | H | H | L | L | L | M | Very high |
| Planning and approvals | H | H | L | L | H | H | Medium |
| Client-side project management | H | H | L | M | L | H | High |
| Architecture: concept and design | H | M | L | M | M | H | Medium |
| Documentation (drawings, specifications) | H | H | L | L | M | L | Very high |
| Engineering design | H | H | L | M | H | M | Medium to high |
| Specialist consultants (fire, acoustic, ESD) | H | H | L | M | H | L | Medium |
| Cost consulting and quantity surveying | H | H | L | M | L | M | High |
| Certification and building surveying | M | M | L | H | H | M | Low |
| Procurement and tendering | H | H | L | L | L | M | Very high |
| Main contractor | M | M | H | H | M | H | Low to medium |
| Contract administration and claims | H | H | L | M | M | M | High |
| Subcontractors and trades | L | M | H | M | M | M | Very low |
| Suppliers and manufacturing | L | L | H | M | L | L | Very low |
| Legal and disputes | H | H | L | M | M | H | Medium |
| Handover and defects | M | M | M | H | M | M | Medium |
| Operations and facilities | M | M | H | M | L | M | Low to medium |
| Software and data platforms | H | M | L | L | L | L | Medium |
What the map shows
1. The exposed pool is concentrated. Feasibility, client-side project management admin, documentation, cost estimating, tendering and contract administration are almost pure knowledge and time, with little risk, permission or physical work attached. Together they make up most of what consultants bill, and almost all of it is priced on hours AI removes. Documentation is the biggest single pool, and it is the work done by the junior base of the traditional pyramid.
2. The durable stages hold the money. Land and capital, finance, certification, the main contractor, trades and suppliers are about risk, permission or physical work. AI improves how they operate, but it barely changes their economics in three years.
3. Sign-off is the dividing line in the middle. Architecture, engineering, specialist consulting and legal work keep their value through accountable sign-off and trust. Regulation is moving the opposite way to the technology. The NSW Design and Building Practitioners Act 2020 requires registered practitioners to declare compliance and creates a statutory duty of care. AI can produce the design. A registered, insured person still has to stand behind it. The registered principal does well. The documentation pyramid beneath them does not.
4. The prize is knowledge applied to the durable stages. The largest value is not in doing project management, documentation or tendering more cheaply. It is in using knowledge to cut programme time, rework and contingency in the stages where the money actually is. Only the party carrying that risk captures it.
Why the squeeze in Australia is delayed, not avoided
Australia's AEC industry is capacity-constrained. The National Housing Accord targets 1.2 million new well-located homes over five years from mid-2024, alongside a large infrastructure pipeline and persistent shortages of engineers and project managers. In the short term AI will mostly absorb that shortage rather than remove hours, and Jevons (1865) reminds us that cheaper inputs often increase total demand.
That buys time. It does not change the direction. The cost per unit of knowledge work keeps falling, and once clients and competitors catch up, prices follow.
What firms should do with the window
Price against the physical economics, not your cost of production. Fees tied to measured programme days saved, rework avoided or contingency released hold their value. Fees tied to the effort of producing documents do not.
Own the sign-off and the assurance. In a world of unlimited cheap AI output, the scarce thing is someone accountable who checks it and stands behind it. Registration, certification and insured assurance are where professional margin survives.
Turn the archive into a data asset. A firm's decades of project history (costs, fees, programmes, contract terms, defects) become more valuable as everything else gets cheaper. Benchmarks and predictions built on proprietary data are hard to copy.
Redesign the pyramid before it redesigns you. Fewer graduates doing documentation, more experienced people reviewing and deciding. The hard problem is the missing rung: how juniors become seniors when junior work is automated. Firms that solve training will win the talent.
Move towards risk and ownership. The long-run escape from the collapse is not being better at knowledge work. It is moving to where knowledge is applied to risk and assets: guaranteed programmes, shared savings, joint ventures and equity in the projects and platforms where the surplus lands.
The uncomfortable conclusion
The billable hour is going. Outcome pricing is the right bridge. But the bridge leads somewhere specific: towards risk, permission, physical delivery and ownership. AEC firms that sell only knowledge and time are on the wrong side of the curve, however good they are at it. The firms that do well over the next decade will use cheap knowledge to change the economics of the physical project, and will be positioned to keep the savings.
Tom Leyden runs Red Yellow Blue, a strategic AI advisory practice working with mid-market Australian firms in construction, professional services and property. The value chain ratings in this piece are his assessment; the references below support the underlying evidence.
References
- Acemoglu, D., and Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives, 33(2), 3-30. https://doi.org/10.1257/jep.33.2.3
- Agrawal, A., Gans, J., and Goldfarb, A. (2018). Prediction Machines: The Simple Economics of Artificial Intelligence. Harvard Business Review Press.
- Agrawal, A., Gans, J. S., and Goldfarb, A. (2019). Artificial intelligence: The ambiguous labor market impact of automating prediction. Journal of Economic Perspectives, 33(2), 31-50. https://doi.org/10.1257/jep.33.2.31
- Autor, D. H. (2015). Why are there still so many jobs? The history and future of workplace automation. Journal of Economic Perspectives, 29(3), 3-30. https://doi.org/10.1257/jep.29.3.3
- Baumol, W. J., and Bowen, W. G. (1966). Performing Arts: The Economic Dilemma. Twentieth Century Fund.
- Brynjolfsson, E., Li, D., and Raymond, L. (2025). Generative AI at work. Quarterly Journal of Economics, 140. https://doi.org/10.1093/qje/qjae044
- Dell'Acqua, F., McFowland III, E., Mollick, E. R., Lifshitz-Assaf, H., Kellogg, K. C., Rajendran, S., Krayer, L., Candelon, F., and Lakhani, K. R. (2023). Navigating the jagged technological frontier: Field experimental evidence of the effects of AI on knowledge worker productivity and quality. Harvard Business School Working Paper 24-013.
- Eloundou, T., Manning, S., Mishkin, P., and Rock, D. (2024). GPTs are GPTs: Labor market impact potential of LLMs. Science, 384(6702), 1306-1308. https://doi.org/10.1126/science.adj0998
- Flyvbjerg, B. (2014). What you should know about megaprojects and why: An overview. Project Management Journal, 45(2), 6-19. https://doi.org/10.1002/pmj.21409
- Goolsbee, A., and Syverson, C. (2023). The strange and awful path of productivity in the U.S. construction sector. NBER Working Paper 30845. https://www.nber.org/papers/w30845
- Hwang, B.-G., Thomas, S. R., Haas, C. T., and Caldas, C. H. (2009). Measuring the impact of rework on construction cost performance. Journal of Construction Engineering and Management, 135(3), 187-198. https://doi.org/10.1061/(ASCE)0733-9364(2009)135:3(187)
- Jevons, W. S. (1865). The Coal Question. Macmillan.
- Love, P. E. D. (2002). Influence of project type and procurement method on rework costs in building construction projects. Journal of Construction Engineering and Management, 128(1), 18-29. https://doi.org/10.1061/(ASCE)0733-9364(2002)128:1(18)
- McKinsey Global Institute (2017). Reinventing Construction: A Route to Higher Productivity. McKinsey & Company.
- Noy, S., and Zhang, W. (2023). Experimental evidence on the productivity effects of generative artificial intelligence. Science, 381(6654), 187-192. https://doi.org/10.1126/science.adh2586
- Pan, Y., and Zhang, L. (2021). Roles of artificial intelligence in construction engineering and management: A critical review and future trends. Automation in Construction, 122, 103517. https://doi.org/10.1016/j.autcon.2020.103517
- Susskind, R., and Susskind, D. (2015). The Future of the Professions: How Technology Will Transform the Work of Human Experts. Oxford University Press.
- Australian Government Treasury. National Housing Accord. https://treasury.gov.au/policy-topics/housing/accord
- Design and Building Practitioners Act 2020 (NSW). https://legislation.nsw.gov.au/view/html/inforce/current/act-2020-007