The headline number in AIA's June Architecture Billings Index was 47.3. That was an improvement from May, but it was still below 50, meaning more firms reported declining billings than increasing billings. AIA said the profession had now gone 41 months without a majority of firms reporting billings growth.
The number I would circle was several paragraphs lower: 3.1 months.
That was the average project backlog AIA reported for firms with less than $250,000 in annual billings in the second quarter of 2026, down from 4.9 months in the first quarter. Across all responding firms, backlog moved only from 6.6 months to 6.3. Larger-firm backlogs generally held steady, according to AIA. The sharpest size-segment compression was in the lowest annual-billings segment AIA identified.
That is a serious signal. It is not yet a diagnosis.
The two numbers are also different types of measurement. The ABI is a diffusion index. AIA asks firms whether billings increased, decreased, or stayed the same, then builds the index from the share in each response. A 47.3 is not a 2.7 percent decline, a 47.3 percent decline, or a dollar-volume estimate. The 3.1-month figure is an average reported backlog level for one billing segment. Putting them in the same sentence can be useful, but only if the reader does not treat them as interchangeable measures of magnitude.
Four different things get called pipeline
Architecture firms use the word pipeline so loosely that it can describe almost any work that might produce a fee. I think that makes a soft market harder to read. At minimum, I would separate four stages.
An inquiry would include a conversation, request for qualifications, feasibility question, or invitation to propose. I would treat it as evidence of interest, not commercial commitment.
A signed agreement would record the fee under contract. I would keep any phase that still requires a client's later authorization out of the active-work total.
Authorized backlog would include work the firm is permitted to perform but has not yet billed. That is the closest internal analogue to a workload cushion, although firms do not all calculate it the same way.
A collectible receivable would include work already billed that the firm reasonably expects to collect. I would keep it out of backlog entirely and track its aging separately.
Those stages can move in different directions at the same time. A studio can have a full proposal calendar, a healthy count of signed agreements, and a thin authorized backlog if clients are holding phases. It can also have months of authorized work and weak cash collections. Calling all four conditions a strong pipeline hides the point where the work is actually stopping.
The June release shows the gap, but not for the same firms
AIA's national June indicators illustrate why the distinction matters. The project inquiries index rose to 56.1. The design contracts index was 49.8, essentially flat. The billings index was 47.3. Those are three different stages moving three different ways.
But they cannot explain the 3.1-month backlog figure.
The public release gives the inquiry, contract, and billing indexes for the overall survey. It does not give those readings for the group below $250,000 in annual billings. It also does not publish the respondent count for that segment, the quarter-to-quarter composition of that segment, or a size-segment backlog series long enough to tell whether a 1.8-month move is unusual volatility or the beginning of a sustained break.
That means I cannot say the firms in this segment received more inquiries but failed to turn them into contracts. I cannot say their signed work failed to receive authorization. I cannot say their receivables weakened. The national readings and the segment backlog are not the same denominator.
AIA itself uses careful language, saying firms in this group may be working through existing backlog while struggling to find new projects. That is a plausible interpretation. It is not a finding that the published segment data can independently demonstrate.
Do not substitute headcount for billings
There is another easy denominator mistake. In its 2024 Firm Survey summary, AIA reported that about 75 percent of U.S. architecture firms had fewer than 10 employees. It described 28 percent as sole practitioners, 32 percent as firms with two to four employees, and 15 percent as firms with five to nine.
Those figures show that the profession is dominated numerically by firms with small teams. They do not tell us how many firms fall below $250,000 in annual billings, and they do not let us apply the 3.1-month average to every sole practitioner, boutique, or local firm. Headcount and annual billings are different units. A ten-person firm is not interchangeable with AIA's lowest billing segment, and neither is every one-person practice.
This matters because the phrase "small firm" invites readers to picture their own office and assume the number belongs to them. The only defensible description here is the one AIA actually measured: firms reporting less than $250,000 in annual billings.
Why the number still deserves attention
Refusing to overread a statistic is not the same as dismissing it.
AIA reported 6.3 months of average backlog across all firms at the end of 2025. In March 2026, the average rose to 6.6 months, the highest since December 2023. By the second quarter, it was back to 6.3. At first glance, the aggregate looked almost unchanged over six months.
The lowest annual-billings segment did not. Its reported cushion compressed from 4.9 months to 3.1 in one quarter while larger-firm backlogs generally held steady. Even without a causal explanation, that split tells principals not to use the profession-wide average as a proxy for their part of the market.
The longer backdrop makes the split more important. AIA's December 2025 survey found that 90 percent of responding firm leaders had experienced significantly delayed projects during the prior six months, 84 percent had experienced projects placed on hold or stalled indefinitely, and 71 percent had experienced cancellations or abandonment. Those figures count whether a firm experienced any such project. They do not mean most of all project dollars were troubled. In the same release, firms said an average of 70 percent of project dollars were proceeding normally, while 16 percent were significantly delayed, 10 percent were on hold, and 5 percent were canceled.
That is exactly why the stages matter. A firm can be busy and exposed at the same time. Current work can keep people occupied while future authorizations thin. A project can remain in a signed contract total while its next phase sits on hold.
Construction activity can look healthier because it is later in the cycle
The June 2026 Census construction-spending release estimated total construction at a seasonally adjusted annual rate of $2.1665 trillion. That was 3.2 percent below June 2025, with stated uncertainty of plus or minus 1.5 percentage points. The measure still described construction being put in place, not the design decisions that preceded it.
That does not contradict weak architecture billings. AIA describes the ABI as a leading indicator that tends to lead nonresidential construction activity by roughly 9 to 12 months. Census measures the value of construction put in place, which includes work already under construction. An architecture firm can therefore see active job sites, RFIs, and construction administration from earlier awards while its front-end authorized backlog is getting thinner.
The visual evidence of a busy construction market arrives later than the commercial decision that feeds a design studio. That timing difference is another reason not to let current activity stand in for future authorized work.
The internal report I would want
The public ABI can tell a principal whether the broader market is expanding or contracting. It cannot replace a firm's own stage accounting. I would want one project register with separate values and dates for inquiry, proposal, signed agreement, authorization, unbilled authorized fee, invoice, and collection.
I would also want stalled value shown separately from active backlog. A signed project waiting on financing should not carry the same operational weight as a phase with a notice to proceed and staff assigned. Nor should a past-due invoice be treated as evidence of future workload just because it appears in a financial dashboard.
The point is not to create a complicated forecast. It is to prevent one optimistic stage from covering a weak one. If inquiries are strong and authorizations are weak, that is one problem. If authorizations are healthy and collections are slow, that is another. They need different responses.
What I cannot conclude from 3.1 months
I could not find a public AIA size-segment backlog history that would establish whether the move from 4.9 to 3.1 months falls outside normal quarterly volatility. The June public page and infographic do not disclose the respondent count for the under-$250,000 segment, whether exactly the same firms supplied both quarter averages, or a margin of error for the backlog estimate.
The figure also does not measure cash, profitability, receivable quality, staff utilization, or survival risk. It does not describe every small architecture practice. And national inquiry and contract indexes cannot be assigned to this billing segment without segment-level results.
So I would not build a forecast from one quarter. I would watch the next two releases, ask AIA for the segment sample and history, and compare the published signal against the firm's own authorized-work register.
Still, I would not let the overall 6.3-month average erase the segment split. The June data says that one specifically defined group reported a much thinner cushion than it had one quarter earlier. That is enough to ask a better question. It is not enough to pretend we already know the answer.