Formation Without Payroll

The entrepreneurial base is widening faster than its employer layer

September brought the owner-only economy into sharper view. The United States had 31,985,874 nonemployer establishments in 2024, according to tax-record data released by the Census Bureau on September 29. That was 1.56 million more than in 2023. Their nominal receipts reached $1.861 trillion.

This is not a side market. It is the numerical center of American business formation.

But count is not capacity. Three-quarters of those establishments reported less than $50,000 in annual receipts, and 86.7% reported less than $100,000. Average nominal receipts increased only about 1% from 2023 to 2024 even as the number of establishments rose 5.1%.

The current formation data extends the pattern. Through August 2026, seasonally adjusted business applications were 16.7% higher than in the same eight months of 2025. Applications indicating planned wages moved the other way, falling 15.9%. Their share of all applications declined from 9.6% to 6.9%.

The strongest interpretation is not that hiring is disappearing or that every applicant intends to remain solo. It is narrower: the entrepreneurial base is widening faster than its employer layer. More people are entering business formation through structures that do not begin with payroll.

For CKOS, that strengthens a principle running through the Observatory since Issue 03, The First Hire: systems designed around departments, administrators, and delegated roles enter the market too late. A solo-first operating system must create value while one person still holds sales, delivery, finance, coordination, and judgment at the same time.

This issue separates four layers throughout:

• Observed evidence is what a source directly reports or what CKOS calculates from a released public dataset.
• CKOS interpretation connects signals without claiming that correlation establishes cause.
• Opportunity hypotheses identify tests, not validated markets.
• Unknowns remain open because the available evidence does not answer them.

Evidence brief

The month produced six signals that matter for owner-only firms and businesses with fewer than ten workers.

1. The owner-only stock expanded. Census counted 31.99 million nonemployer establishments in 2024, up 5.1% from 2023. Nominal receipts rose 6.2% to $1.861 trillion, while average nominal receipts rose only 1.0% to about $58,194.
2. Most nonemployers remained economically small. 75.4% reported less than $50,000 in annual receipts; 86.7% reported less than $100,000. Sole proprietorships accounted for 27.64 million establishments, or 86.4% of the total.
3. Applications rose while payroll intent fell. From January through August, seasonally adjusted applications increased 16.7% year over year to 4.19 million. Applications with planned wages fell 15.9% to 289,542.
4. One-to-nine-worker firms weakened in August. QuickBooks estimated employment down 11,100 jobs and average real monthly revenue down 1.52% to $49,850.
5. Openings fell sharply even though hires did not. BLS estimated 1.05 million openings at establishments with one to nine workers, down 335,000 from July, while hires increased by 7,000.
6. Customers spent more per visit but visited less. Fiserv reported sales up 1.3% year over year, average tickets up 3.0%, and transactions down 1.8% for a tenth consecutive month of traffic decline.

Taken together, these observations strengthen August's Growth Without Slack finding and make it more specific. The expansion is not only happening without much operating room. Much of the new base is forming without an initial payroll layer.

Lead essay / The smallest firm is not a smaller company

Most business software and professional services still imagine the small firm as a miniature version of the large one. The interface may be simplified and the price reduced, but the underlying model remains departmental: sales enters the opportunity, operations fulfills it, finance records it, marketing sustains demand, and management reviews the result.

The owner-only firm has none of those handoffs. One person changes roles repeatedly across the same day. The relevant constraint is not merely limited headcount. It is that the same person must preserve context while moving among incompatible kinds of work.

That makes scale assumptions consequential. A workflow that saves twenty minutes but requires a new system of record, another weekly review, and an integration may create a net loss. A service that assumes the customer can prepare a brief, organize assets, approve drafts, and analyze results may simply relocate the work back to the owner. A financing product that looks at receipts without cash timing may mistake activity for resilience.

The September evidence shows why this is not a niche design concern. The nonemployer universe grew by 1.56 million establishments in one year. The majority were sole proprietorships, and three-quarters were below $50,000 in receipts. At that scale, fixed administrative effort is not diluted across a team or a large revenue base. It lands directly on the owner.

CKOS interpretation: the central unit of small-business design is not the company. It is the owner's constrained operating day. Products and services should be evaluated by whether they reduce reconstruction, coordination, waiting, and switching inside that day.

This strengthens Issue 08, The Capacity Tax. It also reframes Issue 05, Administrative Fixed Cost. The same compliance task, software setup, procurement process, or reporting requirement consumes a much larger share of available capacity at $25,000 of receipts than at $2.5 million.

The implication is not that every small firm needs automation. It is that every proposed intervention must show where the owner's work actually disappears. Faster task completion is insufficient if the owner still has to frame, verify, correct, transfer, and remember the work.

What the nonemployer data does not tell us

Nonemployer establishments are not identical to unique people. One owner may operate more than one establishment. Some partnerships and corporations are included, although sole proprietorships dominate. Receipts are not profit, owner income, or cash available to invest. Annual tax data also arrives with a long lag: the newly released file describes 2024, not September 2026.

Those limitations narrow the claim. They do not erase it. The data establishes the scale, receipt distribution, and legal form of the owner-only business base. It does not establish wellbeing, survival, or willingness to pay.

Feature / The formation boom is changing shape

Business applications are easy to celebrate because they are timely and large. They are also easy to misread.

The Census Business Formation Statistics distinguish all applications from high-propensity applications, applications with planned wages, and projected employer formations. Those measures describe different stages and intentions. An application is not an operating business. A planned-wage indication is not a completed hire. A projected formation is a model estimate of future payroll-tax liability, not a count of durable firms.

Through August 2026, all applications reached 4.19 million on a seasonally adjusted basis. The comparable 2025 total was 3.59 million. Yet applications with planned wages fell from 344,344 to 289,542. The share dropped from 9.6% to 6.9%.

The August cross-section is even clearer. Total applications were 531,728. Only 34,263 indicated planned wages, or 6.4%. In August 2024 the comparable share was 10.6%; in August 2025 it was 8.4%.

Projected employer formations did not collapse. Census projected 28,501 payroll startups from the August 2026 cohort, 9.8% more than from the August 2025 cohort, even though the measure fell 4.6% from July. That is an important counter-signal. The formation pipeline can produce more future employers in absolute terms while payroll intent becomes a smaller part of a much larger application pool.

CKOS interpretation: formation is bifurcating. The United States can have both abundant entrepreneurial entry and a thinner employer-intent layer. The relevant market is therefore not simply “startups.” It contains at least three operating states:

1. income-generating owner activity with no near-term employment intent;
2. durable owner-only firms that use contractors, platforms, and partners instead of payroll; and
3. pre-employer firms approaching the discontinuity of a first hire.

The same product proposition will not serve all three.

This strongly reframes Issue 03, The First Hire. The first hire is not the default next step for every new firm. It is one branch in a formation system increasingly populated by firms designed to remain owner-led or externally networked.

Opportunity hypothesis: the pre-payroll operating layer

A plausible territory exists between formation paperwork and full business software: a lightweight operating layer for commitments, customer work, cash routines, repeat decisions, and external collaborators before payroll exists.

The test is not whether owners say they need “one place for everything.” The test is whether a bounded system can reduce missed commitments, repeated setup, payment delay, or owner review without demanding a new administrative habit.

Status: watch and research. The evidence establishes population size and operating structure, not demand for a particular product.

Feature / Hiring did not send one signal

September's labor evidence appears contradictory.

QuickBooks estimated that firms with one to nine employees lost 11,100 jobs in August. BLS estimated that job openings at establishments of the same size fell by 335,000, with the openings rate dropping from 6.0% to 4.3%. NFIB reported that 35% of surveyed owners still had openings they could not fill and 82% of those hiring or trying to hire encountered few or no qualified applicants.

ADP, meanwhile, estimated that establishments with one to nineteen employees added 20,000 jobs in August. BLS also estimated that hires at one-to-nine-worker establishments increased by 7,000 even as openings fell.

These figures do not describe the same population or the same event.

• QuickBooks targets businesses with one to nine employees and models monthly employment using anonymized payroll records calibrated to official statistics.
• BLS JOLTS measures establishments, not firms, and separately estimates openings on the last business day and hires across the month.
• ADP uses payroll records and reports a wider one-to-nineteen-worker size class.
• NFIB surveys members and measures reported conditions and intentions rather than universe counts.

Observed evidence: openings at the smallest establishments fell sharply in August; hires did not fall in the same measure; one-to-nine employment declined in QuickBooks; one-to-nineteen employment increased in ADP.

CKOS interpretation: the safest conclusion is not “microbusiness hiring collapsed.” It is that hiring demand became less visible and more selective at the smallest establishments while completed payroll counts remained mixed. The discontinuity between wanting capacity, posting an opening, finding a qualified person, and adding payroll remains unresolved.

This operationalizes Issue 07, The Small-Business Data Blind Spot. Definition is not a footnote. It changes the direction of the headline.

No-action conclusion: do not build for a generic hiring shortage

The month does not justify a recruiting marketplace, labor-matching product, or first-hire platform. The signals conflict, the sector mix is unknown, and the pain may sit in qualification, wage economics, schedule fit, or management readiness rather than discovery.

Status: no action. Continue interviews only where a current CKOS venture encounters a concrete capacity bottleneck.

Feature / Fewer visits make each relationship carry more weight

Fiserv's transaction data extends a pattern first surfaced in Issue 13. Small-business sales rose 1.3% year over year in August, while average tickets rose 3.0% and transactions fell 1.8%. Traffic had declined for ten consecutive months.

The composition matters. Restaurant transactions fell 2.9%. Service transactions fell 2.7% while average service tickets rose 3.9%. Essentials outgrew discretionary spending, and discretionary traffic declined twice as fast.

The Small Business Majority operator survey points in the same direction from a different method. Among 222 owners in its network—primarily women- and minority-owned firms with fewer than ten employees—73% reported higher expenses, 75% reported revenue unchanged or lower, and 83% said reduced consumer spending was a challenge. Seventy-three percent reported difficulty making payroll and/or paying business expenses.

NFIB's August survey provides a broader employer-owner counterpoint. Its optimism index remained just above the long-run average, but a net negative 9% reported higher nominal sales over the prior three months, the weakest reading since November 2025.

Consumer context weakened at month end. The Conference Board's Consumer Confidence Index fell 6.7 points to 81.9, and its Expectations Index declined to 63.6. The Federal Reserve's September Beige Book described widespread price sensitivity and small-business customers under operating pressure.

CKOS interpretation: when transaction density falls, relationship quality becomes more economically important. A retained customer, referral, repeat order, or trusted local channel carries more of the firm's commercial result. This strengthens Issue 01, The Relationship Gap, and Issue 09, Average Is Not a Customer.

The opportunity is not generic lead generation. It is to help an owner understand which relationships create repeat, margin, and reliable cash—and to preserve enough context to act on that knowledge.

Opportunity hypothesis: relationship yield

A lightweight diagnostic could connect customer source, repeat behavior, revision burden, payment timing, and referral value. It should begin as a service method or structured review, not as software.

Status: operationalize as a research method. Test with existing founder and pilot conversations. Do not create a standalone venture.

Feature / Technology adoption is not operating slack

The Small Business Majority survey reported that only 20% of respondents were using no AI tools, down from 36% in its first-quarter poll. That is meaningful adoption within its network. It does not show that technology relieved the financial strain reported in the same survey.

This is the month's most useful AI finding precisely because it is not an AI-market forecast. Tool use rose while expenses, weak revenue, reduced spending, and payment difficulty remained widespread.

CKOS interpretation: adoption and relief must be measured separately. A firm can use AI frequently and still have no additional cash, customers, time, or decision confidence. Any CKOS proposition that uses AI should therefore be evaluated on an operating outcome—reduced owner review, faster cash conversion, fewer missing-context incidents, better customer retention—not on generation volume or usage frequency.

Opportunity hypothesis: none at the generic AI layer. AI remains an enabling method inside a validated workflow.

Status: closed for generic AI analysis. Reopen only when a specific owner workflow shows measurable relief after maintenance and review costs are counted.

Capital is uneven, not universally closed

The third-quarter CFO Survey from Duke and the Federal Reserve Banks of Richmond and Atlanta adds a useful but broader signal. About 20% of “small” respondents—defined as firms with fewer than 500 employees—reported financing constraints, roughly twice the rate among large firms. More than 60% of constrained firms said financing prevented them from pursuing new opportunities.

The counter-signal is equally important: 80% of small respondents did not cite financing as a constraint, a plurality reported margins near normal, and overall hiring and spending plans remained intact.

This source cannot be translated directly to owner-only firms. Its small category is far too broad. It is included to bound the narrative, not to establish microbusiness prevalence.

CKOS interpretation: financial constraint is concentrated rather than universal. The observable business problem is likely eligibility, timing, documentation, or volatility for a segment—not a single missing product for “small business.”

Status: closed for a new finance venture. Continue to track cash timing and access barriers inside specific workflows.

Opportunity map

Counter-signals

A disciplined synthesis must retain evidence that pushes against the lead thesis.

• Projected employer formations from the August 2026 application cohort were 9.8% higher than from August 2025. Employer formation is not disappearing.
• ADP estimated 20,000 jobs added at establishments with one to nineteen employees in August.
• BLS estimated 7,000 more hires at one-to-nine-worker establishments even as openings fell.
• NFIB optimism remained above its 52-year average, and 68% of respondents rated their own business health good or excellent.
• The Richmond Fed found that 80% of small respondents were not financially constrained and that overall spending and hiring plans remained intact.
• Fiserv measured positive year-over-year sales, and retail foot traffic was slightly positive year over year.
• Nonemployer receipts rose faster than establishment count in nominal terms from 2023 to 2024, although average growth was modest and not inflation-adjusted.

These signals prevent “formation without payroll” from becoming a decline narrative. The evidence describes a changing structure, not a failed economy.

Month-over-month lineage

No prior core finding is overturned. One common assumption is weakened: that more business formation should naturally produce proportionate near-term payroll formation.

Research agenda

1. Distinguish durable owner-only businesses from transitional pre-employer firms in interviews and datasets.
2. Ask what event—not aspiration—would make a first hire economically and managerially possible.
3. Measure the owner's full coordination burden before proposing shared services or software.
4. Test a relationship-yield review using repeat behavior, referral, margin, revisions, and payment timing.
5. Track the planned-wage share of applications monthly and compare it with projected and realized employer formations.
6. Compare owner income, profit, and hours across receipt classes when credible microdata becomes available.
7. Treat AI use as an input and measure downstream relief after review and maintenance are counted.
8. Search for current evidence on contractor use, informal help, and shared capacity among nonemployers.

Unresolved questions

• How many nonemployer establishments represent primary livelihoods rather than supplemental income?
• What share of the planned-wage decline reflects durable solo design, contractor substitution, industry mix, application composition, or delayed hiring?
• Which owner-only firms want to hire but cannot, and which never intend to?
• Does a pre-payroll operating record reduce failure or simply add another task?
• Which functions can be pooled without creating more owner coordination than they remove?
• What proportion of AI use produces measurable time, cash, customer, or decision improvement?
• How do receipt-size classes differ by sector, owner demographics, and geography after cost of living is considered?

Public source ledger

1. U.S. Census Bureau, 2024 Nonemployer Statistics, released September 29, 2026. Annual tax-record data for establishments with no paid employees and qualifying receipts. CKOS calculations compare the 2023 and 2024 U.S. files and sum released receipt-size classes. Receipts are nominal and reported in thousands of dollars. Establishments are not necessarily unique owners.
2. U.S. Census Bureau, Business Formation Statistics: August 2026, released September 11, 2026. Administrative EIN application data and modeled employer-formation projections. CKOS calculated year-to-date totals and planned-wage shares from the released seasonally adjusted monthly series. Applications are not operating firms.
3. Intuit QuickBooks, Small Business Index: September 2026, released September 1, 2026. Modeled employment and real-revenue indicator targeting U.S. firms with one to nine employees, calibrated with official statistics and payroll/accounting data.
4. Fiserv, Small Business Index: August 2026, released September 3, 2026. Point-of-sale activity across approximately two million small businesses; not isolated to owner-only or one-to-nine-worker firms.
5. NFIB Research Center, Small Business Economic Trends: August 2026, released September 8, 2026. Monthly survey of randomly selected NFIB members. Measures conditions and intentions, not universe counts; firm sizes vary.
6. U.S. Bureau of Labor Statistics, JOLTS establishment-size table: August 2026, released September 29, 2026. Preliminary total-private estimates by establishment size; openings are measured on the final business day while hires and separations cover the month.
7. ADP Research and Stanford Digital Economy Lab, National Employment Report: August 2026, released September 2, 2026. Payroll-based private-employment estimate; the most relevant published size band is one to nineteen employees, not one to nine.
8. Small Business Majority, Voice of Main Street: Q3 2026, released September 16, 2026. Network poll of 222 owners, primarily women- and minority-owned firms with fewer than ten employees; 33% reported revenue below $100,000. Margin of error ±6 percentage points. The sample is not a probability sample of all U.S. microbusinesses.
9. Federal Reserve Banks of Richmond and Atlanta and Duke University, CFO Survey: Q3 2026, released September 23, 2026. Survey of 517 financial executives; its “small” category is below 500 employees and is used only as contextual counter-evidence.
10. The Conference Board, Consumer Confidence: September 2026, released September 29, 2026. Consumer attitudes and expectations; a demand-context indicator, not a direct business measure.
11. Federal Reserve Board, September 2026 Beige Book, released September 2, 2026. Qualitative district evidence collected on or before August 24; useful for mechanisms such as price sensitivity, not prevalence estimates.

Editorial standard

Observed evidence is restricted to what each source reports or what can be reproduced from a released dataset. CKOS interpretation connects evidence without claiming causal proof. Opportunity hypotheses are bounded tests. Unknowns are preserved. Broad sources are never silently translated into owner-only prevalence.

Consulting-firm and generic AI-market releases screened during September did not provide sufficient owner-only or sub-ten-worker resolution to support a claim in this issue. Exclusion is part of the method.

Evidence window closed September 30, 2026. No confidential client, employer, founder, or pilot data were used.