AI Automation Agency
Get paid to take the repetitive work out of other people's businesses, then keep a retainer.
Client automations
What ran, what broke, what recovered
- Quote follow-up
- Lead routing
- Quote follow-up
- Quote follow-up
- Lead routing
- CRM sync
- CRM sync
- Quote follow-up
- Lead routing
| Event | Workflow | Client | Result |
|---|---|---|---|
| Closed | Lead routing | Havenzicht | Complete |
| Assigned | Lead routing | Havenzicht | OK |
| Triggered | Lead routing | Havenzicht | OK |
| Closed | Quote follow-up | Bakker BV | Complete |
| Retried | CRM sync | Bakker BV | OK |
Example figures, shown to illustrate how the workflow behaves.
Open full screen · recording mode — sized for vertical video.
Follow these prompts to start building your income stream
7 steps. Paste each prompt into Claude, use what comes out, move to the next. The last one ends in something live — not in a plan to make something live.
- 01
Pick one industry, one process
Free sampleOne industry, five manual processes, and the one you lead with.
In this step you choose what you are the person for. Automation sells badly as a general service and well as 'I fix the thing where your orders get retyped'. You end with a shortlist of processes you can name in someone else's language, which is what gets you the first meeting.
1 — Find the manual processesI want to sell automation services to one industry: [INDUSTRY]. My own experience with this industry: [YOUR EXPERIENCE] Typical company size I would target: [COMPANY SIZE] List the ten processes in this industry that are most often still done by hand — copying between systems, retyping orders, chasing approvals, manual reporting. For each one: 1. The process in the industry's own words. 2. Which systems it moves data between. 3. Roughly how often it happens, and who does it. 4. Why it is still manual — the real reason, which is usually not "nobody thought of it". 5. What breaks when it goes wrong, and who notices. 6. How hard it is to automate with n8n or Make: easy, medium or hard, and why. Then pick the five that combine "obviously painful" with "actually buildable", and say what makes each a better opening than the others.
Claude. This is judgement, not retrieval — a strong general model is the right tool.
- [INDUSTRY]
- One industry whose language you already speak. Generic outreach about automation performs badly.
- [YOUR EXPERIENCE]
- Jobs, family business, anything that means you know how the work really runs.
- [COMPANY SIZE]
- Employee count or revenue band. A ten-person firm and a 200-person firm have different bottlenecks.
Then: Pick one process. That single process is what you talk about for the next month, in every conversation.
- 02
Put your offer online
In the packageA live page saying what you fix and what it costs.
In this step your service becomes something you can link to. It is one page, not a brand: the process you fix, what the client gets, what it costs, and how to start. Businesses buy automation from someone who sounds specific, and specificity is easier to show than to claim.
- 03
Book your first calls
In the packageFifteen named companies, the message that gets a reply, calls booked.
In this step you get in front of people. You are not selling an automation yet — you are asking to hear how the process runs today, which is a much easier thing for someone to say yes to. You end with conversations booked and a question list to run them with.
- 04
Map their process, send a quote
In the packageA process map they recognise, and a fixed price.
In this step the conversation becomes something you can sell. The map does two jobs: it proves you listened, and it makes the exceptions visible before you price the work. The quote then prices an outcome instead of your hours.
- 05
Build it so it fails loudly
In the packageA working automation that tells you when it breaks.
In this step you build the thing. The difference between an automation that survives and one that gets switched off is what happens on a bad day: the missing field, the API that times out, the duplicate order. You end with something running and a way to know when it is not.
- 06
Hand over and invoice
In the packageDocumentation they can use, the invoice sent, the monthly fee running.
In this step you get paid and the relationship turns into income that repeats. Documentation is the part most builders skip, and it is exactly what makes a retainer defensible: you are not charging to keep a secret, you are charging to keep it running.
- 07
Turn it into five more leads
In the packageA case study, and five companies with the same problem.
In this step the first client becomes a sales asset. Every industry has the same process running in dozens of companies, and you now have the only thing that opens those doors: a specific story about fixing it. This is where the second client costs a tenth of the first.
6 of the 7 steps are in AI Money
The sample above is a real step, not a teaser — it is the same depth as the rest. The package adds the other 6 for this strategy, and the same treatment for the other eleven.
Your first 7 days
- 1Pick one industry you already understand and list five processes you know are manual there.
- 2Build one of them end to end for yourself, with error handling and alerts.
- 3Write the case study for that build: before, after, hours saved, what could break.
- 4Approach five businesses in that industry about that one process — not about automation.
Your first 30 days
- 1Sell one paid discovery, even at a low price, and produce a written process map from it.
- 2Deliver the first workflow with monitoring and a handover document.
- 3Convert that client to a monthly maintenance agreement before the project closes.
- 4Write down your own delivery checklist so the second build takes half the time.
What you need
core means you cannot skip it. The rest only if you need it.
- n8ncoreWorkflow platform, self-hostable for client data
- Makecorefree planFaster to build in, easier to hand over
- Zapierfree planWhere the client already lives in mainstream SaaS
- Claudecorefree planBuilding, debugging and documenting flows
- Airtablefree planThe database most clients can actually maintain
- Retoolfree planInternal interfaces on top of the automation
- Activepiecesfree planOpen-source option when licensing matters
Want the research?— what it is, what it costs, and how we scored it
What you can actually watch
Workflow runs firing, nodes turning green and red, a failed step retrying itself.
What it is
You get paid to remove repetitive work from other people's businesses — an order that is retyped into three systems, a quote that waits two days for someone to copy it — and then you keep a retainer for running what you built.
Why this opportunity
Demand scored 8 of 10, the highest of the twelve. Every business above a certain size has processes held together by copying and pasting, and the people doing it can name them in one sentence. The supply side is real too: n8n raised a $180M Series C in October 2025 at a $2.5bn valuation, led by Accel with NVentures and Deutsche Telekom participating (Tier 1/2, blog.n8n.io/series-c). That proves confidence in the platform — not in the services market around it, a distinction guru content always drops.
Who this is for
For someone who likes mapping how work actually flows and can sit with an operations manager without panicking about the mess. Not for someone who wants to avoid responsibility: when an automation fails at 3am, it is yours.
The business model
Project fee for implementation, then a monthly retainer for monitoring, fixes and changes. The margin lives in the retainer; the project pays for the sales cycle that won it.
How money is made
Implementation fees per workflow, monthly maintenance, and change requests. Circulating rate ranges for this market ($5,000–$50,000 per project, $2,000–$15,000 a month retainer) come from Tier 3/4 marketing blogs written by firms selling these services. Our research marked them UNVERIFIED and refused to publish them as market fact; there is no audited rate index for this market.
Who pays
SMEs and scale-ups with manual, repeatable processes. The payer is usually an operations director or the owner — rarely IT, which changes how you pitch it entirely.
Finding customers
Network and referral, narrow outbound, partner directories, and case content. Cold outreach about "automation" in general performs badly; outreach about one named process in one industry performs markedly better.
Delivery
Discovery, process mapping, build, handover, maintenance. Discovery is where the money is decided: what you find there determines whether the build is three days or three weeks.
Where AI actually helps
High while building — agent-assisted development makes integration work far faster — and low in discovery and accountability. Nobody buys an automation because a model wrote it; they buy because someone is responsible for it.
Skills you need
Medium to high: process thinking, APIs, error handling, and business analysis. The analysis is the differentiator. Plenty of people can wire two systems together; few can tell a client which process is worth wiring.
What it costs to start
Low. n8n Cloud Starter is €20 a month and Pro €50 a month (Tier 1, n8n.io/pricing, LAST VERIFIED: 2026-09-07), plus model usage. No inventory, no staff to start.
How to price it
Fixed price per workflow after a paid discovery, then a retainer priced on the number of workflows under management. Price the discovery — free discovery is where unprofitable projects come from.
Common mistakes
Selling automation instead of a named process. Unbounded scope with no change procedure. No monitoring, so the client discovers the failure before you do. And underpricing the maintenance, which is the only part that compounds.
Scaling
Scalability 6 of 10. You scale by repeating one workflow across one industry, not by taking every request. Templating your own delivery and monitoring is what moves margin; selling does not automate.
The version that holds up
Specialise until you can quote a price in the first call because you have built that exact flow eleven times. Then productise it: fixed scope, fixed price, fixed timeline, with your own monitoring dashboard as the reason the retainer exists.
Time to first revenue
4–16 weeks
4–16 weeks to the first paid engagement, assuming (a) you already have a network or a credible niche, (b) you sell one concrete process rather than "automation", and (c) you accept a small first job. Without a network and without a niche, six months or more is just as realistic. This is a range, not a promise.
How it scored
Ten axes, scored. Risk scored 5 out of 10, where higher is worse.
The state of this market
The most crowded market in the set: competition scored 9 out of 10. n8n raised $180M at a $2.5bn valuation in October 2025, which proves confidence in the platform rather than in the services market around it. Only defensible with a genuine niche and an explicit maintenance model.
Not sure this one fits you?
Answer a few questions about budget, time and skills and see which of the eighteen match your situation — and which to avoid.
Run the Opportunity Finder