Agency AI Niche & ICP Finder
A free course for SMMA, marketing and AI agency owners. Six prompts turn your client history into a chosen niche, ...
Show more
Instructor
David Tainsh
Course details
Lectures
:
4
Level
:
Beginner
- Description
- Curriculum
- FAQ
- Reviews
Free course · For SMMA, marketing and AI agency owners
By the end you have chosen a market — and checked it before telling anyone.
Six prompts that pull candidate niches out of your own history — jobs, freelance work and any clients you have had — score them against six tests, send you to verify the winner in a single evening, and turn it into an ideal client and one sentence you can commit to.
- A decision, not a driftOne niche, chosen from evidence you already have and argued for rather than compared.
- Checked before committedSix free public checks in an evening, run before the website or the headline changes.
- A list you can build todayAn ideal client specific enough to produce two hundred named businesses this afternoon.
Who this is for
New agency owners choosing their first niche.
You do not need any clients yet. The course works from what you have already done — previous jobs, freelance work, favours and the industries you know from outside work — so you choose your market at the start instead of drifting into one.
- 01Starting out with no clients yet
Deciding at the start rather than drifting into it — which is cheaper, and the evidence is in your job history rather than a client list.
- 02Leaving a job to go out on your own
Years of experience in an industry is often the strongest niche candidate you have. This turns it into a market you can name and check.
- 03Anyone who tried niching and stopped
Usually because the niche was chosen on aspiration. This one is chosen on evidence, and checked before it is announced.
- 04Early agencies with a few mixed clients
A homepage that says “we work with ambitious local businesses”, because nothing more specific would be true yet.
You do not need market research, a data subscription, an agency coach or a sector you already love. You need an honest list of what you have already done.
What you walk away with
Eight things, from your own history.
All of it built from evidence you already have rather than from a sector somebody recommended. The measure of this course is whether you could change your homepage next week and still believe it in a year.
- Five candidate niches, each traceable to a line in your own history
- The unfair advantage you already hold in each one, or an honest “none”
- The candidate you were about to pick for the wrong reasons, named early
- Six scores per candidate, with any fatal score named rather than averaged away
- One niche chosen and argued for — not a balanced comparison
- A six-check verification brief you can run free, in a single evening
- An ideal client in four parts, with disqualifiers you can spot from outside
- A positioning line, and the specific work you are now turning down
The ideal client is also what the free Cold Outreach course assumes you already have. Take this first and you skip its opening prompt entirely.
Why it never gets chosen
Nothing is failing. Nothing is compounding either.
That is why an accumulated niche survives for years. It does not hurt enough to fix, and it quietly caps everything built on top of it. Four things keep agencies general.
- ×You accumulated instead of deciding
Three years of whatever came in, then you counted which kind of client you had most of. That is a pattern, and a pattern cannot be priced or prospected against.
- ×You chose on aspiration
A revenue figure somebody posted, in a sector nothing in your own history points at. You start from zero against specialists and conclude eighteen months later that niching does not work.
- ×You averaged the scores
Five candidates, totalled columns, highest wins. The winner failed one structural test — usually reachability or proof — and the total still looked respectable.
- ×You never wrote down the no
So when an enquiry arrived from outside the niche on a slow month, you took it. Eighteen months later you count your clients and find you have accumulated one again.
The drift is not a failure of discipline. It is what happens when the no was never written down while nothing was at stake.
The trap
A niche you cannot list is a preference.
“Ambitious founder-led businesses.” “Companies that value quality.” “Businesses ready to scale.” Every one of those describes a temperament, and you cannot build outreach, pricing or proof on a temperament.
A preference
- Filters you cannot actually apply from outside
- No register, directory or trade body behind it
- Impossible to build a list of two hundred
- Every message has to work for everybody
- Proof from one client transfers to nobody
A market
- Something observable decides who is in it
- A public register, directory or map search exists
- Two hundred names in an hour, without buying data
- You know their words and what makes them look
- One result makes the next four sales easier
The through-line
AI scores the candidates. The history has to be yours.
The prompts can inventory, score, pressure-test and narrow. What they cannot do is know what you have actually done — and every way this goes wrong goes wrong on the wrong side of that line.
- 1You write the honest history
Every client, including the ones that ended badly and the favours. The freelance work before the agency. And the industries you know from outside work altogether.
- 2The prompts find and test
Five candidates, each traceable to a line you wrote. Six scores each. And a refusal to average a fatal score into a respectable total.
- 3Nothing fashionable gets included
A candidate has to come out of your evidence. If your history genuinely points nowhere, the prompt is told to say so rather than inventing five plausible industries.
If your five candidates are all sectors you have never touched, the history was too thin. Add the pre-agency work and the jobs before that — it takes ten minutes and it changes the answer.
The framework
Six tests a niche has to pass.
Scored one to five, never averaged. A single fatal score kills a candidate outright, because the structural failures cannot be worked around with effort.
- 1Reachable
Could you build a list of two hundred of them without buying data? Name where the list comes from. This is the test most stated positions fail.
- 2Already spending
Is there a budget line for something adjacent — an agency, a tool, a hire? Redirecting a budget takes two months. Creating one takes two years.
- 3Something is urgent
A visible event that makes some of them look this quarter. Without urgency you are selling improvement, and improvement is what everybody defers.
- 4Provable in ninety days
Could you produce one referenceable result inside a quarter? A niche where proof takes a year is a niche a small agency cannot afford to enter.
- 5You have a way in
Prior work, a relationship, a background, a language. Anything a stranger would not have. Starting cold in a crowded niche is the expensive way to begin.
- 6Repeatable
Does the second client buy roughly what the first one bought? If every engagement is bespoke, you have a job rather than a niche.
One and four are the genuinely fatal ones. Spend, urgency and your own enthusiasm can all be worked on; reachability and proof cannot.
What’s included
Every lesson comes with the prompts that build it.
The classes are short. The work happens in the prompt pack — a numbered SOP, the six prompts in order, and a note on what a good output actually looks like so you can tell when to run one again.
- SOPA step table in the pack
Seven numbered steps. Four are the decision, three are what you do with it. Work down it and you have a chosen market.
- AICopy-paste prompts, in plain English
No codes and no variables to install. Square brackets mark the few things only you can supply.
- QAWhat good looks like
A concrete standard for every prompt — including how to tell when the scoring has been generous to your favourite.
- DEMOA real agency at the moment of choosing
Apex AI’s full client history, untidy and complete, with the failures and the pre-agency work left in.
How the course runs
Three short lessons, then the whole chain run live.
Twenty minutes writing the history, about an hour on the prompts, and one evening checking the answer. Only the last step is public, and it comes last on purpose.
Lesson one
Why agencies never choose
Accumulated versus chosen, the six tests, and why one fatal score kills a candidate whatever the other five say.
Your evening
Write the honest history
Every client including the failures, the pre-agency work, and the industries you know from outside work altogether.
Lesson two
Find and score the candidates
Prompts 01 to 03. Five candidates, six scores each, and a committed choice rather than a comparison.
Lesson three
Verify it and say it
Prompts 04 to 06. Six checks, the ideal client inside the niche, and a positioning line with its no-list.
Live demo
All six prompts, end to end
The full chain on a real agency, including the candidate that scores well, feels right and dies on one test.
The live demo
Watch an appealing candidate lose to evidence.
Apex AI is eighteen months in with six clients across four unrelated sectors. Nothing is failing. The founder spent four years in hospitality before the agency, which feels like an unfair advantage right until the scoring starts.
The position
Six clients, four sectors
- No two builds quite the same
- Every proposal from scratch
- Homepage describes a temperament
The favourite
Hospitality
- Four years front-of-house behind him
- Reachable and urgent, both fine
- One restaurant client, already failed
The answer
Sitting in the client list
- Four clients, near-identical builds
- Proof already achieved in under 90 days
- A public register, listable in an hour
Is it for you
Agency owners prepared to lose their favourite.
A good fit
- You are choosing your first market, or are not sure the one you have was ever chosen
- You will write the history including the failures and the favours
- You want a decision you can argue for rather than a shortlist
- You will spend an evening checking before you announce anything
- You can accept the answer being a sector you were not excited about
Not a fit
- You want to be told which industries are currently lucrative
- You have already decided and want the exercise to agree
- You would keep a candidate that fails on reachability
- You want a market research report rather than a decision
- You will take work from outside the niche whenever it is quiet
What it costs
Nothing, and nothing attached to it.
Free
No card. Lifetime access, including anything added later.
4 lessons · 6 prompts · prompt pack · demo pack · live demo
No upsell inside the lessons. There is a paid course this leads to, and it is mentioned once, at the end, after you already have a chosen market.
Quick answers
The questions people ask before starting.
I haven’t got any clients yet. Can I still do this?
Yes. Your history includes previous jobs, freelance work and industries you know from outside work altogether, and several of the strongest candidates usually come from that last group rather than from a client list.
Which AI does this use?
Claude. The prompts are plain English with square brackets marking the few things only you can supply. Nothing to install, no account to connect, no research tool to subscribe to.
Does niching mean turning away work?
Yes, and the course makes you write down exactly which work, before you announce anything. A positioning line you contradict within a month is worse than not having one — and that is how an accumulated niche quietly reassembles itself.
What if I disagree with the recommendation?
Argue with it — but by naming the specific score you think is wrong and the evidence it overlooked. If you can do that you have learned something. If you cannot, you have probably found the wrong-reasons candidate from prompt 01 arriving on time.
How long before I can commit?
About an hour on the prompts and one evening on the checks. The checks are the part people skip, and they are the reason the choice is still reversible when you find the problem.
How does this connect to the other free courses?
The ideal client this produces is what the free Cold Outreach Engine assumes you already have. Take this one first and you can skip its opening prompt and paste this in instead.
What exactly do I get?
Four lessons — three teaching lessons across 45 slides plus a live demo of the full chain — six plain-English prompts, the prompt pack and SOP in PDF and Word, a worked agency with a complete client history, and lifetime access including anything added later.
Stop noticing your niche. Decide it.
An honest history, six tests, one evening of checking — and a market you chose on purpose, with the work you are turning down written next to it.
A Digital Plan · adigitalplan.com
Agency AI Niche & ICP Finder
-
1
Lesson 01 — Why agencies never chooseYou took whatever came in for three years, then counted which kind of client you had most of. That is a pattern rather than a decision, and nothing about it is failing, which is exactly why it persists. But nothing compounds either: every proposal starts from scratch, no case study makes the next sale easier, and the website says something about ambitious local businesses because nothing more specific would be true.
This lesson makes the case for choosing on evidence rather than aspiration. You will learn the six tests, why a niche you cannot build a list of is a preference rather than a market, and why one fatal score kills a candidate whatever the other five say.
-
2
Lesson 02 — Find and score the candidates
The input is an honest history: every client including the ones that ended badly, the freelance work from before the agency existed, and the industries you know from outside work altogether. That last category produces the strongest candidate more often than the client list does, and it is the one everybody leaves off because it feels irrelevant.
You will run the first three prompts and meet the demo: Apex AI, eighteen months in, six clients across four unrelated sectors, and a founder with four years of hospitality behind him that feels like an unfair advantage right up until you look at the evidence.
-
3
Lesson 03 — Verify it and say it
The instinct on choosing is to rewrite the LinkedIn headline that afternoon. It feels like commitment, and it is what makes reversing the decision expensive. Verification costs one evening: six checks against free public sources plus two short conversations, and you are looking for the thing that would make you regret it while backing out is still cheap.
Then the narrowing. The niche gets you a list; the ideal client gets you a message, a price and a reason to say no. The lesson ends on the no-list, because a niche is defined by what you turn down rather than by what you add to your website.
-
4
Lesson 04 — Live demo — all six prompts, end to end
No new teaching. The whole chain run start to finish on screen using Apex AI, so you can watch what each prompt returns before you spend an evening on your own history.
The useful moment is the scoring. Hospitality looks strong on reachability and urgency, and the founder has four years of front-of-house behind him — and it still dies, because the one restaurant engagement failed, the margins were poor and the build does not repeat. Watching an appealing candidate lose to evidence is the part most people need to see before they will do it to their own favourite.
Is it really free?
Yes. No card is needed and you get lifetime access, including anything added later. A paid course is mentioned once, at the end, after you already have a chosen market.
I haven't got any clients yet. Is this for me?
Yes. The course builds your candidate niches from your own history, and that includes previous jobs, freelance work and industries you know from outside work, not just clients. If you are just going out on your own, choosing your niche at the start is cheaper than drifting into one later.
What counts as experience if I have no clients?
Every job you have held, any freelance work or favours you have done, and the industries you know from outside work altogether. Some of the strongest candidates usually come from that last group rather than from a client list.
What will I have at the end?
One niche chosen from five candidates and argued for, a six-check verification brief, an ideal client in four parts with disqualifiers, and a positioning line with the work you are now turning down.
What do I need before I start?
An honest written history of what you have already done, which takes about twenty minutes. You do not need market research, a data subscription, an agency coach or a sector you already love.
How long does it take?
About twenty minutes writing your history, about an hour on the prompts, and one evening running the checks. The checks come before you announce anything or build your website, so the choice is still easy to change.
Which AI does it use?
Claude. The prompts are plain English with square brackets for the few things only you can supply, with nothing to install and no research tool to subscribe to.
What should I do once I've chosen my niche?
The ideal client this course produces is exactly what the free Cold Outreach Engine assumes you already have, so that is the natural next step for finding your first clients. Take this one first and you can skip that course's opening prompt.
Please, login to leave a review