Direct answer: A $5,000/month AI agency is roughly five clients paying $1,000 a month in retainers, and the automations behind it can be built in no-code tools such as Make or n8n. The limiting factor is not coding but sales volume and retention: under the model below, landing one new client a month with 10% monthly churn reaches $5,000 in recurring revenue in about seven months, while landing one every two months never gets there.
Search this topic and every income figure you find traces back to someone selling something: a course, a coaching seat, a playbook, a book. One guide promises $5,000 to $30,000 a month within the first year. An audiobook promises a $10,000/month business by chapter twelve. A consultancy owner on LinkedIn says almost nobody makes more than $5K/month selling automations, then pitches a coaching program for the alternative. None of these publish verifiable income records for their buyers.
So this article does something different. It ranks what can actually be verified, runs the arithmetic in the open, and tells you which part of the plan is easy (the tooling) and which part decides the outcome (selling and keeping clients).
What counts as evidence here
| Tier | Source type | What it can support | What it cannot |
|---|---|---|---|
| 1 | FTC complaints, orders and rules | What regulators allege or have ordered; legal exposure for income claims | Anything about typical agency earnings |
| 2 | Vendor pricing pages and documentation | Tool costs and limits (first-party, checkable, and they change) | Whether a tool makes you money |
| 3 | Census Bureau survey data | How many US businesses report using AI | Whether they will pay an outside agency |
| 4 | Guides, courses, social posts | What the market is selling | Anything about what typical operators earn |
There is no independent dataset on how many people who start an AI agency reach $5,000 a month. I searched for one and found none. Everything about earnings below is therefore either arithmetic from stated assumptions or a claim from an interested party, and it is labeled that way.
The regulatory fact competitors skip
The strongest evidence in this niche is not about agencies. It is about the people selling the dream of one.
The FTC sued Air AI in August 2025, alleging deceptive claims about business growth, earnings potential, and refund guarantees. In March 2026, it announced a settlement that bans the company and its owners from marketing business opportunities. The FTC alleged that some buyers lost as much as $250,000. These are allegations resolved by settlement, not findings of fact, but the pattern is consistent with other FTC actions against AI-branded “passive income” storefront schemes.
The mechanism matters more than the company. The FTC’s Business Opportunity Rule requires sellers to be able to substantiate earnings claims with written documentation. In January 2025, the FTC proposed extending that rule to money-making opportunities such as business coaching. I did not verify whether that expansion has been finalized, so check its current status before relying on it.
What this means in practice:
- If you plan to teach people to build AI agencies and put income numbers in your marketing, you are in the zone these cases describe. Publish only earnings you can document, and state what is typical.
- If you run an agency for clients, the rule is not aimed at you, but the underlying principle is: don’t promise outcomes or refund terms you cannot back up.
- This is context, not legal advice. A lawyer who handles advertising claims should review any earnings-based marketing.
The math nobody shows
Define terms first. MRR is monthly recurring revenue: money that repeats without a new sale. A setup fee is not MRR. A month in which you collect $5,000 from two builds is not a $5,000/month business, because the next month starts at zero.
Retainer prices are the weakest input. Guide sources disagree by a wide margin. One puts maintenance retainers at $300 to $1,500 a month; another puts basic AI management at $2,000 to $5,000. The low and high ends differ by roughly 3x to 7x, and neither publishes client data. I use $1,000 below because it sits inside the first range and below the second. Treat it as an assumption you must test, not a market rate.
Ways to reach $5,000 MRR (arithmetic only):
| Active clients | Retainer each | MRR |
|---|---|---|
| 10 | $500 | $5,000 |
| 5 | $1,000 | $5,000 |
| 4 | $1,250 | $5,000 |
| 2 | $2,500 | $5,000 |
Tool cost is the strongest input and the least important. Make’s entry paid plan is listed at about $9 a month billed annually, or $10.59 billed monthly, for 10,000 credits, and sources report slightly different figures. Even a generous stack of $150 a month (an automation platform, an AI subscription, an email tool, a domain; this total is my assumption) leaves $4,850 of the $5,000 before taxes and your own time. Tools are not what stands between you and the number.
What decides it is a pipeline equation. If you land L new clients per month and lose a fraction c of your clients each month, your client count settles toward L ÷ c. That one ratio explains most outcomes. With L = 1 and c = 10%, you settle at 10 clients. With L = 0.5 and c = 10%, you settle at 5, which means you only ever approach $5,000 and never cross it.
I ran the model month by month with a flat $1,000 retainer, constant churn, clients landing at the start of each month, and setup fees excluded:
| New clients per month | Monthly churn | Clients at month 12 | MRR at month 12 | First month at $5,000+ MRR |
|---|---|---|---|---|
| 0.5 | 10% | 3.6 | $3,588 | Never (ceiling is 5 clients) |
| 0.5 | 5% | 4.6 | $4,596 | Month 14 |
| 1 | 15% | 5.7 | $5,718 | Month 9 |
| 1 | 10% | 7.2 | $7,176 | Month 7 |
| 1 | 5% | 9.2 | $9,193 | Month 6 |
| 2 | 10% | 14.4 | $14,351 | Month 3 |
Illustrative model, not observed data. The chart plots four rows of the table above.
Read the table asymmetrically. Halving your sales rate costs far more than doubling your churn, because sales rate sets the ceiling and churn only sets the speed. Going from one client a month to half a client a month at 10% churn turns a seven-month path into a permanent shortfall. Going from 5% churn to 15% churn at one client a month delays the target by three months but still gets there.
Setup fees change the picture less than they appear to. A $1,500 setup fee on one new client a month is $1,500 of cash that stops the moment you stop selling. It helps you survive the ramp. It does not compound.
Where “zero coding” holds, and where it doesn’t
The no-code claim is true for building. Visual platforms let you connect apps and chain steps without writing code. What the guides rarely cover is cost behavior and failure behavior.
Cost behavior. Make counts each module that acts in a run as a credit, so a four-step scenario run 300 times uses about 1,200 credits. A lead-enrichment workflow handling 5,000 leads a month at six steps each needs 30,000, which is three times the base allowance. One source also notes that failed operations still count toward the quota. A client with more volume than you scoped can push you onto a pricier plan while your retainer stays fixed. The practical rule: estimate monthly credits per client before quoting, and write a volume cap into the contract.
Failure behavior. This part is my analysis, not sourced data, but it follows from how these systems work. Automations break quietly. A client edits the form your workflow reads, a connected app changes its interface, a trigger stops firing, or a model returns output in a shape your next step cannot parse. Nothing alerts you unless you built the alert. The consequences land on the client’s customers: a missed lead, a duplicated invoice reminder, a wrong answer sent under their name.
This is why the retainer exists, and it is the only honest justification for one. You are not charging for the build; you are charging to watch it. If you cannot name what you monitor each month, you do not have a retainer product, and the churn column above will punish you.
What sells as a retainer and what doesn’t
A workflow that fixes a one-time task ends the relationship when it works. That is the structural reason build-only agencies stall, and it matches the claim from the LinkedIn consultancy owner mentioned earlier (an interested source, but the logic holds independent of who said it). A one-off build has 100% churn at delivery.
| Service shape | Recurs? | Why (editorial judgment) |
|---|---|---|
| Daily lead response or follow-up | Yes | Runs continuously, errors cost the client money, someone must review exceptions |
| Monthly reporting or invoice chasing | Yes | The output is consumed every month and the inputs change |
| Support triage with human escalation | Yes | Answer quality drifts, and rules need updating |
| Data cleanup or migration | No | Finite job; sell it as a fixed project |
| Custom internal tool, one team | Rarely | Value is front-loaded; maintenance is thin |
Sell the left column as retainers and the bottom rows as projects. Do not blur them: a project priced as a retainer fails on delivery, and a retainer with nothing to maintain fails at renewal.
Is there demand?
Yes, but not for what the courses say you are selling. The Census Bureau’s Business Trends and Outlook Survey tracks AI use across US businesses. A summary of its latest wave (August 24 to September 6, 2026) reports 23.8% of businesses using AI in the prior two weeks, up from 17.7% at the start of 2026. Verify the current figure at census.gov, because the survey updates frequently, and note that Census changed the question wording in November 2025 to cover use “in any business function.” Older, lower figures are not comparable.
The detail that matters for you: a Census analysis covering December 2025 to May 2026 found AI use rose among firms with 20 or more employees but did not change significantly among firms with fewer than 20. The small businesses most agencies target are the slowest-moving segment.
My reading, which is analysis rather than data: owners of small firms already use consumer AI tools themselves, so “we do AI” is not a product. A named process with a measurable result (“every missed call gets a text within a minute”) is. Your offer should never contain the word “AI” in the first sentence.
Getting to five clients
I have no sourced conversion rates for agency outreach, and competitors’ numbers are unverified. So treat the figures below as inputs to test, not benchmarks.
If you close 1 in 10 conversations, one new client a month requires about ten conversations a month. At 1 in 20, it requires twenty. The model above needs that pace sustained for six to nine months, alongside delivery work. That is a sales job with a technical side.
A 30-day test that bites:
- Pick one process in one trade where a missed step visibly costs the owner money, such as quote follow-up or missed-call response. One niche, one offer.
- Hold 40 conversations before buying any course, template, or community. Write down every objection.
- Quote a fixed setup fee plus a retainer with a stated scope: what you monitor, volume cap, response time. Avoid refund guarantees and performance promises you cannot measure; refund-guarantee claims are part of what the FTC alleged against Air AI.
- Stop rule: if 40 conversations produce zero paid pilots, change the offer or the niche. Do not change the tools.
Who should not do this
- Anyone who needs $5,000 within 90 days. Even the fast path in the table takes six to nine months, and a half-client-a-month pace never arrives.
- Anyone unwilling to do outbound sales. The pipeline equation has no tooling variable in it.
- Anyone targeting regulated data such as health or finance records without budgeting for compliance work. The retainer ranges in public guides do not account for it, and I did not find sourced guidance that does.
If this fails, the cheapest pivot is to keep the process skill and sell it as a fixed-price project to the same niche, or as freelance work to existing agencies, rather than rebuilding the offer from zero.
What this article cannot answer
Two questions remain open, and I would rather say so than guess.
What share of people who try this reach $5,000 a month? No independent data exists in what I found. Payment-processor cohort data or a registered, representative survey of agency owners would resolve it.
What does real client churn look like? The table shows how sensitive outcomes are to it, but no source I found publishes churn for small AI automation agencies. Your own first ten clients are the only dataset you will have, so track cancellation reasons from day one.
How the numbers were produced
- Retainer, churn, sales rate, and the $150 tool stack are assumptions stated above. The month-by-month results are computed from them as: clients(t) = clients(t−1) × (1 − churn) + new clients, with MRR = clients × $1,000.
- Make pricing and credit rules come from third-party pricing guides checked in 2026 and differ slightly between them. Confirm on Make’s own pricing page before quoting clients.
- Census figures are relayed through secondary summaries of the Business Trends and Outlook Survey. Confirm at census.gov.
- No tests, interviews, or personal agency results are claimed anywhere in this article.
Recheck within six months: Make pricing and credit rules, the Census adoption figure, and the status of the FTC’s proposed Business Opportunity Rule expansion.
Sources
- FTC case summary, FTC v. Air AI Technologies: https://www.ftc.gov/node/297672
- PYMNTS, report on the March 2026 Air AI settlement: https://www.pymnts.com/legal/2026/ftc-seeks-to-ban-air-ai-from-marketing-business-opportunities/
- FTC proposed Business Opportunity Rule changes (January 2025): https://www.ftc.gov/node/87160
- Klein Moynihan, summary of Business Opportunity Rule substantiation duty: https://kleinmoynihan.com/ftc-settles-work-at-home-lawsuit/
- Mondaq, FTC sweep of AI-driven business opportunity schemes: https://webiis10.mondaq.com/unitedstates/advertising-marketing-branding/1660472/ftc-shuts-down-ai-driven-business-opportunity-scheme-continues-sweep-of-deceptive-income-claims
- Jet Admin, Make.com pricing (checked September 26, 2026): https://www.jetadmin.io/blog/make-pricing/
- Toolradar, Make pricing and operation counting: https://toolradar.com/tools/make/pricing
- Use Apify, Make credit examples: https://use-apify.com/blog/make-com-pricing-2026
- Start Business by State, Census BTOS summary (survey of Aug 24 to Sep 6, 2026): https://startbusinessbystate.com/small-business-ai-use-by-state/
- Morris County EDC, Census BTOS analysis by firm size: https://morriscountyedc.org/large-firms-with-at-least-20-employees-biggest-ai-users/
- Market claims referenced: https://dupple.com/learn/how-to-start-an-ai-agency, https://www.marcillyaiplaybook.it.com/blog/how-to-start-ai-automation-agency-2026.html, https://www.audible.com.br/pd/B0GTFMMRTZ