Real Cost of a Solo Business on AI: Software Is 20%

🆕 Rewritten August 2026 — including the parts of this page that were made up. The original version of this article claimed we had spent a year tracking costs across a panel of solo businesses at three revenue stages. We hadn't. No such study existed, and several of the numbers built on top of it — a 2018-versus-2026 savings table, a "visible costs are 30-40% of the total" ratio, a set of setup-hour estimates multiplied by opportunity-cost rates — were invented to fill it out. Those sections are gone rather than quietly reworded.

What replaces them is arithmetic you can redo. Every price below was opened on the vendor's own pricing page in August 2026, every total is the sum of the table above it, and the three costs that turn out to be larger than the software all come from the IRS, the Census Bureau or KFF.

Search for what it costs to run a one-person business with AI and you will find a lot of people arguing about tools. Twenty dollars or nine. Six tools or twelve. Claude or ChatGPT or both.

Here is the thing that argument misses. On a one-person business of typical size, the entire software stack — even a bloated one, even billed the expensive way — comes to about a fifth of what the business actually costs to run. The other four fifths are three numbers that almost never appear in these breakdowns, and one of them changed by 114% this year.

This page is the arithmetic. It reconciles with our complete 2026 solopreneur AI stack guide, which re-derived every price in the same week, and it corrects that guide in one place where it was wrong.

How big is a one-person business, actually?

Before pricing a stack it is worth knowing who the stack is for, because most content on this topic quietly assumes a business far larger than the typical one.

The US Census Bureau counts businesses with no paid employees separately, as nonemployer businesses. In its most recent data — 2022, released May 2025 — there were 29.8 million of them, with $1.7 trillion in receipts between them. That is about 6.8% of the US economy, run by people with no payroll.

Divide it out and the average one-person business turns over roughly $57,000 a year. Census doesn't publish that division itself — it publishes eleven receipt-size bands starting under $5,000 — and a long tail pulls the mean up. But it anchors the scale. When a breakdown pitches a "$295/month full stack" at businesses doing $150-400K, it is pricing for the far right edge of that distribution and describing it as normal.

So the numbers below use $57,000 as the reference business. Scale them if yours is different; the shape of the answer doesn't change.

One aside on failure statistics, because they get quoted constantly in this niche. The "half of businesses fail within five years" figure comes from the Bureau of Labor Statistics' Business Employment Dynamics series, which tracks establishments — defined by the employment they report. Nonemployer businesses aren't in it. BLS also counts an establishment as not surviving when it stops reporting employment, which includes the owner retiring, selling or merging. Whatever it measures, it isn't the odds of a solopreneur still trading in five years. We've quoted it in the past. We won't again.

Stack one: the free floor — $1/month

This is the actual bottom, and it is lower than most articles admit.

That total is honest, and it is also the number most likely to mislead you, because two of those free tiers have limits that arrive fast and one has a limit that isn't a limit at all:

• Make Free allows two active scenarios. Not two runs — two live automations. The third one you build is the one you pay for — $9 a month billed annually, $10.59 month to month. We have a full breakdown of where that ceiling bites in Zapier free vs Make free.
• Zoho Mail's free plan has no IMAP or POP access. It's web-only. Your custom-domain email will not open in Apple Mail, Outlook or the Gmail app. That's the real constraint, and it's structural rather than a usage cap you grow out of.
• Tidio's free plan gives 50 Lyro AI conversations as a one-off allowance, in Tidio's own wording — not 50 a month. It's a trial wearing a tier's clothes, which is why it isn't in the table above.

The full map of what genuinely stays free is in the $0 solopreneur stack and free AI tools for freelancers.

Stack two: the working stack — $128.58 or $95.60

This is our recommended nine-category stack, taken directly from the pillar guide so the two pages agree by construction rather than by coincidence.

One correction to the pillar, found while writing this. That guide carries the Wave receipts add-on at $8 in both columns. On Wave's pricing page, $8/month is what a Wave Pro subscriber pays; a Starter user — which is what this stack has — pays $11/month, or $96/year. So $8 was right for the annual column and $3 low for the monthly one. The pillar's annual total of $95.60 stands; its monthly total was $126.99, not $123.99 — and is now $128.58, after the second correction below.

That figure had spread to six of our articles. All six are now corrected. It is a small error and a useful one, because it is the same mistake in miniature that this whole page is about: a number that is true on one billing basis quoted as though it were true on both.

A second correction, 10 September 2026 — the same mistake, on a different line. Both stacks on this page carried Make Core at $9 in the billed-monthly column. $9 is Core's price billed annually, and it is the figure Make's pricing page shows until you click its monthly toggle. Billed monthly, Core is $10.59. That moves this stack from $126.99 to $128.58 a month and the loaded stack below from $331.99 to $333.58, and every figure derived from them has been recalculated. The annual columns were right.

Recalculating turned up an error of method too. This page said the recommended stack was 6.4% of what the business costs. That divided the nine-tool stack's annual cost, $1,147.20, by the loaded stack's total. Against its own total — the same nine tools plus the same three non-software costs, $15,955.20 — it is 7.2%.

Stack three: the loaded stack — $333.58 or $270.90

Not a recommendation. This is what accumulates when you say yes to every reasonable individual upgrade over about eighteen months. Every line here is defensible on its own.

Sixteen tools. $333.58 a month, or $270.90 — the same sixteen tools, the same features, the same access. (One line in that table is a floor rather than a price: beehiiv Scale is $49 at 1,000 subscribers and climbs with your list — $69 at 2,500, $109 at 10,000. The pillar guide has the full ladder.) The gap is $62.68 a month, $752 a year, and it is bigger than removing any two tools from the list.

We wrote a whole article about that gap — the AI tools under $20 a month, where six tools cost $68 or $108 depending on nothing but the billing toggle. It is the single largest lever a solopreneur has on software spend, and it is almost never framed as a decision at all.

The catch, stated plainly: an annual plan locks the price, not the product. Every tool in that table can change what it gives you mid-term. You are prepaying for twelve months of a product that the vendor may redefine in month three — as Notion did when it announced Notion Mail's shutdown with a seventeen-month lifespan behind it.

The traps that make the sticker price wrong

Four patterns, all verified on the vendors' own pages this month. Each one produces a number that is correct in a spreadsheet and wrong on an invoice.

Seat minimums on products sold to one person. Dext is $31.50/month with a five-user minimum and no single-user tier — a solo business pays for five seats to use one. QuickBooks Online Plus, which older versions of this article listed at $42, is $115/month and is likewise a five-user plan. Fathom's Team plans require two or more users. Monday.com's cheapest paid plan carries a three-seat minimum. If a price looks good, check how many people it assumes you are.

Per-unit pricing that reads as flat. Buffer Essentials is "$5" — per channel. Three channels is $15. Calendly Standard is $10 per seat. Vercel Pro is $20 for the team and $20 for each additional developer seat, plus usage above 10M edge requests, 1TB of transfer or 1M function invocations.

Same price, different thing. Make Pro is $16 against Core's $9, both billed annually, and earlier versions of this page said it bought "higher credit volume". It doesn't — both plans include 10,000 credits. Pro buys priority execution. That is a real feature and it is not more capacity, which is what most people upgrading assume they are paying for. If you need more volume, the honest options are in Make alternatives for solopreneurs.

Tier names that imply a ladder they don't have. This page used to list "Tidio Plus, $59". Tidio Plus starts at $300/month plus usage. The $59 belongs to Growth, one tier below. Anyone budgeting from that line would have been out by more than five times.

The three numbers bigger than the software

Now the part that changes the answer.

Payment processing

Stripe's standard US rate is 2.9% + $0.30 per successful domestic card charge. On top of that: +1.5% for international cards, +1% when currency conversion is needed, $15 per dispute whether or not you win it, and 1.5% for instant payouts.

On $57,000 collected by card at an average of $500 an invoice — 114 transactions — that is about $1,687 a year. Roughly half what the loaded sixteen-tool stack costs, for a line item that never appears in a stack breakdown because it doesn't arrive as a subscription. What the processor's agreement lets it do with that money — holds, reserves, closing the account — is a separate question, covered in processor holds, reserves and closures.

Self-employment tax

The IRS charges self-employment tax at 15.3% — 12.4% for Social Security, 2.9% for Medicare — on 92.35% of net self-employment earnings. The Social Security portion stops at a wage cap that is re-indexed annually; the Medicare portion doesn't stop.

On $57,000 of receipts with $10,000 of costs, that is $47,000 net, and:

$47,000 × 92.35% × 15.3% ≈ $6,641 a year.

That is more than the loaded stack costs, plus the payment processing, combined.

Two honest qualifications. Half of it — the employer-equivalent portion — is deductible above the line against income tax. And it isn't a penalty for working alone: it is the same payroll tax an employer would otherwise have split with you. You are simply paying both halves and seeing the whole bill.

Health cover

This is the one that moved this year, and it moved a long way.

KFF's analysis of the ACA enhanced premium tax credits expiring at the end of 2025: the average Marketplace enrollee's annual out-of-pocket premium payment goes from $888 in 2025 to $1,904 in 2026 — a 114% increase, $1,016 more a year. Separately, the average Marketplace deductible rose from $2,759 to $3,786, up 37%, which KFF describes as the steepest increase on record.

For a solopreneur earning too much for subsidies, the number that matters is the gross premium. Peterson-KFF puts individual-market premiums at an average of $540 per member per month in 2024 — about $6,480 a year — against $587 for fully-insured employer coverage.

Read that comparison carefully, because it is the opposite of the usual complaint: the individual market is not meaningfully more expensive per member. The difference is that no employer is paying a share of it.

Putting the four numbers side by side

Reference business: $57,000 in annual receipts, $10,000 in non-software costs, all revenue collected by card, unsubsidised individual health cover.

The software is 21.3% of that. Switch the same stack to annual billing and it drops to $3,250.80, or 18.0% of a $18,059 total.

So: about a fifth, and that is with a deliberately overloaded stack. Run the pillar's recommended nine categories at $95.60 annually and software falls to 7.2% of what the business costs, measured against that stack's own total of $15,955.20. (This said 6.4% until 10 September; the correction is under stack two.)

The old version of this article asserted that visible tool costs were "30-40% of the real total". That number had no derivation behind it and it was too high in a way that flattered the topic — it made the tool argument look more consequential than it is.

What this doesn't tell you

The costs above are the ones that can be sourced. These can't be, and pretending otherwise is how the last version of this page went wrong:

Setup and switching time. Real, significant, and not quantifiable without measuring it. The previous version multiplied invented hour counts by invented hourly rates to produce confident-looking dollar figures. We're not doing that again. What we can say without inventing anything: migrating a CRM or an accounting system is the expensive kind of work, and a better tool that requires a migration is often worse than the adequate tool already holding your data. That's the argument in HubSpot vs Notion as a CRM and it applies to every category here.

What a regulated profession adds. Every figure here assumes nobody licenses you. If they do, the stack is the cheap part: the obligations attach to each thing the software produces, and a tool that generates thirty posts has generated thirty advertisements. We worked that through for one licensed trade in AI tools for solo real estate agents, where three of our own standard recommendations turned out to collide with rules that bind the licensee rather than the tool.

How much you're actually spending. There is no study of one-person-business software spend that we could find. The nearest evidence is consumer, not business: C+R Research surveyed 1,000 US consumers between 22 April and 2 May 2022, asked them to estimate their monthly subscription spending — average answer $86 — then walked them through it category by category. The real figure averaged $219. A $133/month gap.

That's streaming services and gym memberships, not SaaS, and we're not going to pretend it transfers. What it establishes is that the estimating error in this domain is large and runs one direction. Your stack costs more than your guess. Go and add it up.

Sales tax. Some US states tax SaaS and some don't, and the definitions vary enough that the published counts disagree with each other. Depending on where you are, the sticker price isn't the price.

Vendor risk. Not a number at all, but the thing most likely to invalidate any budget you build. A bookkeeper at $300 a month does not restructure their pricing model. Software does, constantly — that's why every price on this page carries the date we checked it, and why we now re-derive totals instead of carrying them forward.

So what should you actually spend?

The reference business turning over $57,000 a year has roughly $18,000 of unavoidable structural cost before it buys a single subscription. Against that:

• The free floor genuinely works. $1/month is not a rhetorical device. The binding constraints are Make's two active scenarios and Zoho's missing IMAP, and both are specific enough to plan around.
• The first $9 is the best money in the stack. Make Core — $9 a month billed annually, $10.59 month to month — for unlimited active scenarios. That's the upgrade we'd make first, and there's a case for making it the only one — see automating client follow-ups for what one automation tool actually buys back.
• Check the billing basis before you check the price. $752 a year on the loaded stack, and it costs you nothing but a commitment you were probably making anyway.
• Then re-derive your own total, quarterly. Not from an article. From your own card statement, against the vendors' current pages. Ours was wrong by $3 on one line in August and by $1.59 on another in September, and we're the ones who check.

If you want the buying order rather than the accounting, the pillar guide has it, and the free CRM comparison covers the category where free holds up longest.

---

Every price, limit and billing basis in this article was opened on the vendor's own pricing page in August 2026. Make's were opened again on 10 September 2026 with its billing toggle in both positions, after we found Core's annual price in both stacks' monthly columns. Tax figures are from the IRS; business-population figures from the US Census Bureau's Nonemployer Statistics; health-cover figures from KFF and Peterson-KFF. Where an earlier version of this page published something invented — the tracked-business panel, the 2018 comparison, the 30-40% ratio, the setup-hour costings — it is named and removed above rather than quietly deleted. No link in this article pays us — how we make money.