A hiring-cost field guide for one-person businesses — US operators
Can AI agents replace employees? Not the way the headlines say. Agents replace tasks and workflows first — the defined, repeatable parts of a job — and a whole role only when it is almost entirely those tasks. Here is the hiring math, done honestly. This page treats the question as what it actually is: a hiring decision. Run it like one — what the role costs, what the replacement costs, which tasks go first, and where the honest answer is "not yet." Every number carries its source; the ones we couldn't verify, we left out. Built by Pulse, a working 14-agent company that sells the operating manual for self-serve AI courses for solo operators.
The answer
Four facts carry the honest answer, each attributed to its source. Everything else on this page is arithmetic and practice.
A job is a bundle of tasks, and McKinsey's analysis estimates current technologies could automate the work activities that absorb 60–70% of employees' time. The defined, repeatable parts — drafting, sorting, researching, scheduling, answering the known questions — are exactly what agents execute. That is where replacement actually happens: task by task, not headcount by headcount.
Some roles are little more than a task list: routine data handling, tier-one support, fixed ops and admin. Those are absorbable today. Roles with judgment baked in — pricing, hiring, deciding what to build — don't vanish when the task list shrinks; they get smaller and more human. Any page that promises whole-role replacement for every job is selling something.
The US Chamber of Commerce found generative AI use among US small businesses jumped from 23% in 2023 to 58% in 2025 — and among small businesses using AI, 82% increased their workforce over the past year. Adoption so far looks like leverage on people, not a swap for them. The replacement wave the headlines describe hasn't shown up in the small-business numbers.
The Census Bureau counted 29.8 million US businesses with zero employees in 2022 — the majority of all US business establishments — and the Nasdaq Economic Institute found new one-person business applications up more than 20% since early 2025, with AI the named driver. Agents are replacing employees most reliably at the moment a company is born: the hire that never happens because an agent took the task list instead.
Sources cited inline: McKinsey, US Chamber of Commerce, Census Bureau, Nasdaq Economic Institute. The surveys measure different populations on different definitions — we cite each on its own terms rather than splicing them into one trendline.
The human side
The honest version of this math starts by refusing the internet's favorite number.
A hire carries payroll taxes, benefits, equipment, software seats, recruiting time, onboarding hours, and a ramp of months before full output. You will see comparison pages quote one loaded-cost figure for "an employee" — a different figure on every page, depending on who's selling. Those numbers conflict because they are guesses. Anyone quoting a single universal loaded cost is guessing; the honest move is to refuse the number and keep the concept.
The scale of the gap, per Forbes reporting from August 2026: a comparable human team at salary scale runs $80,000–$120,000 per month, while a complete working AI stack for a solo-run company costs roughly $3,000–$12,000 a year. That is the ten-person company's arithmetic. Your decision is smaller and better: one role, one task list at a time.
You are not choosing between a team and a stack. You are deciding, one role at a time, whether its task list is defined and repeatable enough to hand to an agent you supervise — and what your own hours are worth back. The payroll-vs-payroll framing is a mid-size-company model. The solo version fits on one page: sections 03 and 06 are that page. And when you want both sides of the comparison built out line by line — loaded cost, stack cost, break-even — that's the full line-item cost worksheet.
The agent side
Three honest budgets. The third is where the reporting lands; none of them is a moat by itself.
The full stack lands inside the $3,000–$12,000 a year range Forbes reporting cites; the $300–$500 a month working-stack figure comes from independent reporting (aibusiness.vc's one-person-company field report). Two different sources, both attributed — and we won't annualize one into the other to make the gap look tidier. Note what no budget buys: the knowledge of which task to delegate first and how to wire it. That's the $30 part of this page.
The decision
Map the roles you'd hire for, then look at the task list inside each one. If the list is defined and repeatable, it's agent-shaped.
Posts, newsletters, and product copy, drafted against a voice document you maintain once. The agent drafts; you edit, approve, and publish. The role's entry-level work goes first — the editorial judgment stays yours.
Sort, label, and draft replies; keep the follow-up queue honest so no warm lead goes cold. Nothing leaves the building without your hand on it.
Build the list, enrich it, draft personalized first touches at volume no human maintains. What that earns once it's running is arithmetic — see the benchmarks in section 06.
Invoices, scheduling, bookkeeping prep, file hygiene, and a weekly numbers report you actually read. The unglamorous work that quietly eats a solo operator's evenings.
Answer what your own docs already cover, escalate the rest with a summary and a suggested reply. Customers get fast answers; you get pulled in only where judgment is needed.
What to build, what to charge, which market to enter, what "good" looks like. An agent proposes; a founder decides. The moment you outsource the deciding, you don't have a company — you have a feed.
Money conversations, partnerships, angry customers, referrals. "Your first ten hires are AI agents," the Founder Institute's guide puts it — the framing works because the hires that stay human are chosen, not defaulted.
The pattern: replace the task list, keep the judgment. It's why the 82%-grew-headcount finding (1.3) and the replacement headlines can both be true — different decisions, made at different scales.
Honest limits
This page refuses the hype in both directions. Here is the failure case, from one of the biggest believers of all.
Per Ars Technica's August 2026 reporting on Reuters' investigation of Meta's internal Project OT: the plan explored cutting some teams by up to 60% and handing much of the daily work to agents overseen by small teams of people. The second layoff round was canceled, and the agents meant to replace workers made what the report called "large-scale, disruptive" actions before the plan was dropped. If whole-job replacement stalls at that scale, the five-task plan in section 04 is the honest unit of change.
Fortune's May 2026 reporting documents solo founders doing former-hire work with agents while flagging real limits on what going it alone can achieve. And Business Insider's February 2026 essay profiles one solo founder running his company with a "council" of 15 AI agents that saves him about 20 hours a week — a named example, not a statistic. Both sides of that sentence are the truthful one. And the limit with a clock on it — the hours agents cost after they're hired — is counted in the management-hours math. And when a deployment dies anyway, the postmortem lives in the failure-mode field guide.
Start with one task, not a department. Define "done" on paper, wire one workflow, supervise every output for a week, then chain a second agent to the first. The full five-step deployment sequence — with the weekly supervision ritual — is in our first field guide, AI agents for solo founders.
The payback
This is where most comparison pages quietly annualize a good month. Four steps, done honestly.
Two or more hours a week, every week, averaged — not the worst-case week you'd like to justify. If the task is smaller than that, the wiring isn't worth it yet.
Most first deployments run on the $0 floor; a focused stack is about $50 a month; the full working stack is the $300–$500 a month tier from section 03. Buy the tier the task needs, not the one the sales page features.
Hours times 52 is the honest annual number. A workflow that saves two hours a week pays back a $30 course — and the next workflow compounds it. That's the entire ROI claim on this page, and it is deliberately boring.
Run the agent supervised for one week and log every miss. Keep what survives the log; kill what doesn't. The math is only honest if the output is real.
Worked example — outreach: the average cold email earns a 3.43% reply rate and the top senders clear 10.7%+ (Instantly's 2026 benchmark across 700k+ businesses); Backlinko's 12-million-email study puts overall response at 8.5%. The gap is segmentation, emails under 80 words, and weekly testing — research and drafting automate well; the send button stays yours.
The catalog
Everything above is the replace-or-not framework our $30 course catalog installs on your own role list — one course per layer. Self-serve only: buy it, and the files land in your inbox within 24 hours of payment. Start tonight.
Decide Course 1/3
8 modules · Self-paced
The replacement decisions, installed: which roles are agent-shaped, which task lists go first, agent roles and guardrails, and the operating cadence — the literal manual of our 14-agent company, ready to paste into Claude Code.
Wire Course 2/3
6 modules · Self-paced
Steps 6.1–6.4 as a repeatable practice: map the task on paper, wire the n8n or Zapier workflow with an LLM step that doesn't hallucinate, and measure the hours it hands back.
Sell Course 3/3
6 modules · Self-paced
Role 4.3 at production quality: build a list you can reach, write outreach that clears the benchmarks instead of the average, and run the weekly numbers ritual that keeps pipeline honest.
Questions
They replace tasks and workflows first; a whole role only when it is almost entirely defined, repeatable tasks. And the data pushes back on the layoff story: among small businesses using AI, 82% increased their workforce over the past year (US Chamber). Agents are leverage on a person — the one-person company's missing staff, not its laid-off staff.
Three realistic budgets — $0, about $50 a month, and about $300–$500 a month for the full working stack per independent reporting — against a complete human team at salary scale at roughly $80,000–$120,000 a month, per Forbes reporting. No single loaded-cost number for an employee is honest — benefits, taxes, and overhead vary too much to be reduced to one figure.
The five roles in section 04, by task list: the junior writer's first drafts, the assistant's inbox triage, the SDR's lead research and outreach prep, the coordinator's ops and admin, and the helpdesk's tier-one support. Judgment and trust-bearing relationships stay human.
Count the task's real weekly hours, multiply by 52 for the honest annual number, price the tier it needs ($0, about $50, or about $300–$500 a month), and compare against the hours you get back. A workflow that saves two hours a week pays back a $30 course.
The average cold email earns about 3.43% replies; the top senders clear 10.7%+ (Instantly's 2026 benchmark across 700k+ businesses); Backlinko's 12-million-email study puts overall response at 8.5%. They win with segmentation, emails under 80 words, and weekly testing — not volume. AI does the research and drafting; you keep the send button.
The three courses in section 07, $30 each or $79 for the Operator Bundle. Buy with PayPal — the button opens a pre-filled order email and we reply with a PayPal payment request within one business day — and the files arrive by email within 24 hours of payment. 30-day money-back, no interrogation.
Start
One course per layer, or all three as the Operator Bundle — the delegation decisions, the workflows, and the pipeline as one coherent system for $79. No calls, no cohorts: buy it, and the files land in your inbox within 24 hours of payment.