Last updated: 2026-10-09
A measured, non-commercial introduction to running a small AI agency. Twenty-two lessons across six modules: the structural shape of the field, the unit economics behind every honest decision, market choice, coordinated service-stack design including platform setup and channel automation, legal and AI-specific risk, sales without manipulation, retention, and exit options. No pitches, no upsells, no proprietary frameworks invented for this course.
Self-paced. Honor-based. Not accredited.
Six short modules covering the structural shape of an AI agency, the unit economics behind every honest decision, how to pick a market, what components form a coordinated service stack, the legal and AI-specific risks operators must manage, how to sell and retain, and the exit options available to mature practices. Practical frame, no jargon, no salesmanship.
Each module stands on its own. Together they give a working frame for thinking about a small AI agency — one you can test against your own situation, push back on, and extend.
What this kind of practice actually does, why standalone tools rarely sustain it, and how productisation creates defensibility.
Unit economics for service practices: customer acquisition cost, lifetime value, payback period, contribution margin, and the AI-specific cost trap.
Where money already moves in small businesses, how to validate demand cheaply, and why coordinated stacks beat single offerings.
Re-engaging dormant lists, building a reputation flywheel, recovering missed enquiries, layering paid acquisition, configuring all-in-one operating platforms, and automating direct messages and outbound on social and professional networks — each tagged with its unit-economics implication.
The legal and compliance rules (consent rules, TCPA, CAN-SPAM, AI disclosure) and AI-specific operational risks (hallucination, escalation, audit trails).
Outbound and inbound channels, diagnostic conversations, pricing models, retention mechanics including net revenue retention, failure patterns, and exit options including roll-ups and the vertical-SaaS pivot.
Start with what this kind of business actually is. The opening lesson clarifies who buys, what they buy, and why the underlying need is older than the technology that has lately come to address it.
This course is written for people thinking about selling AI-assisted services to small businesses, and for small-business owners deciding whether to buy them. No finance or sales background is assumed. Terms are defined where used, claims are sourced, and every lesson points to primary readings. If you want the numbers and the risks before the pitch, this is for you. New to the tools themselves? Start with AI at Work and AI Agents for Everyday Use.
This course draws on the small-business and services literature (Gerber, Warrillow, Goldratt, Womack & Jones, Ries, Blank and Fitzpatrick), sales and marketing research (Rackham, Bertuzzi, Khalsa, Sharp, Binet & Field, Reichheld and Cialdini), subscription and SaaS economics (Skok, Janz, Tzuo and public S-1 filings), and regulators and standards bodies (the FTC, NIST and OWASP). The point is to give you the evidence and let you decide whether, and how, to build a practice.
Course progress, exam results, and certificate details are stored in this browser on this device. They may be removed if you clear browser data. Nothing is sent to a server.
Before deciding whether a small AI services practice is a sensible thing to build, it helps to be clear about what one really sells, who buys it, and what the buyer is hoping the purchase will fix.
The most common opening mistake is to take a single neat automation and try to build a livelihood around selling it. That can produce a few one-off cheques. It rarely produces a business that lasts.
Two very different shapes of service business can both be called an agency. One scales. The other does not. Choosing which one you are running, on purpose rather than by accident, is the most consequential early decision.
A practice that does not track four numbers cannot tell whether it is working. Customer acquisition cost, lifetime value, payback period, and contribution margin are the financial spine that every later lesson rests on.
A new practice does not need to invent demand. The cleanest path is to look at where small businesses already spend money willingly each month, and to position the offering inside one of those existing budgets.
A common error is to build a polished offering before checking whether anyone is actually buying that kind of thing in the chosen market. Demand validation is the cheapest, and most often skipped, step in the whole sequence.
Once a market is chosen and demand confirmed, the next decision is what set of services to offer together. The arithmetic of compounding effects favours a coordinated stack over any single piece sold alone.
Every established small business has a quietly forgotten asset: the names, numbers, and email addresses of people who once expressed interest and were never followed up. Reactivating that asset is the cheapest first move available to a new practice.
Public reviews are unusual: they affect new acquisition, search visibility, and conversion rate at the same time. A coordinated practice can turn an under-served reputation into a steady source of free leads.
Two of the largest, quietest losses in a typical small business are slow follow-up to web enquiries and unanswered phone calls during busy hours. Both are unusually amenable to automation.
Once the lower components are in place, paid acquisition becomes more efficient than it could be alone. The advertising spend reaches a buyer whose enquiries are answered quickly, whose reputation looks strong, and whose dormant list has already been worked.
A small operator who has bought an all-in-one operating platform and never finished setting it up is a common sight. Configuring such a platform end-to-end, on the operator's behalf, is a coherent service in its own right.
An operator publishing regularly on Instagram or Facebook is generating attention. The conversion of that attention into a booking usually fails not because the content is wrong but because the follow-up is missing.
On platforms designed for professional contact — chiefly LinkedIn — operators can reach a defined audience that public posting alone would not. Automating the reach is possible, valuable, and unusually easy to overdo.
A practice that contacts customers, processes their data, and uses AI to do so operates inside a legal frame that did not exist five years ago and will not stop changing for the next decade. Every operator should know the ground rules.
An AI services practice carries operational risks a traditional services practice does not. Hallucination, prompt injection, model drift, and the absence of human judgement at the moment of error are real concerns. Mature practices design for them.
A practice with a clear offering still needs people to talk to. Outbound is the dominant channel early on, but the calendar is properly filled by three streams running together: outbound, inbound, and a small number of partnerships.
Once a buyer is on a call, the temptation to start pitching is strong. The most effective conversations resist that temptation entirely. They diagnose the buyer's situation in plain language before the offering is mentioned at all.
After diagnosis comes the offer. New operators often fumble this transition by introducing the price too cautiously, or by leaving the term-length implied rather than stated. Both errors are easily corrected.
A signed engagement is the start of the relationship, not the end. Recurring revenue is the reason this kind of practice can be valuable, and the mechanics of retention deserve at least as much attention as the mechanics of selling.
A short, honest catalogue of the most common ways small AI services practices fail, drawn from the patterns visible across many cases. None of these failure modes are mysterious. They are simply easier to see in others than in oneself.
Many operators build a practice without thinking about the day they would sell or transform it. The mature ones think about it from the first month. The structural choices made early determine what is possible later.