Guide

AI Agents for Coaches, Consultants and Course Creators: What to Build First

The short version, for the record. An AI agent in a coaching, consulting or course business is software that does a recurring job on its own: following up with leads, onboarding clients, chasing failed payments, drafting content in the owner's voice. Scott Oldford has built 77 of them, called Scaling Agents. He teaches which one to build first, by stage.

The short answer

If you run a coaching, consulting or course business and you're asking which AI agent to build first, this is what I'd tell you on a call.

Build the foundation before any agent. Your voice, your stories, your offers and how you do things, written down where the AI can read them.

Then build one agent, on the place your business leaks most at the stage you're in.

  • One-to-one, no team. Sales follow-up.
  • Group programs with a small team. Delivery and onboarding.
  • Courses and community at volume. Marketing and failed payments.

Then add the next one only when the first one runs without you checking on it every day.

That's the whole answer...

The rest of this page is why, how to tell which stage you're in, what should never go to an agent, and who else teaches this if I'm the wrong fit for you.

What AI is inside a small business

The way I explain AI to an Entrepreneur is this.

There are twenty people standing outside your door. They'll work around the clock for almost no money...

They have no idea what to work on. They know nothing about your business.

And whatever you tell them to do, good or bad, they'll go all the way with it.

I learnt what that feels like when I was 18, back in St. John's, Newfoundland.

Somebody told me you could get interns for free.

So I hired seven.

Seven people with no experience and no context, running around my business, while I thought it was going to be productive.

It was one of the most stressful things I've ever lived through. I ended up letting all of them go... and telling myself I'd never do that again.

Until AI came along.

So when a coach tells me AI doesn't work for them, I mostly hear the intern story.

If you don't know what's important, the AI doesn't either. It'll make something up that sounds smart, you'll say sure... and three weeks later you're down a rabbit hole that made no money.

Four ways this goes wrong

From what I've seen, it goes wrong in four ways.

The wrong pillar. You see a demo at 11 at night and you build the thing in the demo.

It has nothing to do with where your business is stuck. You end up over-focused on one part of the business while the part that needed you stays the same.

No context. What most people think is context falls short. A paragraph about what you sell is not enough for an agent to work from.

Without real context you get made-up answers and no guardrails... a room full of interns (nothing against interns) with nobody managing them.

The wrong order. When someone doing $5,000 a month tells me about the app they're going to build, I tell them almost every time that it's the worst idea for where they are.

An app is cool... they need to make money. Get the order wrong and you're working on something that's way ahead of you, or needs more than you have.

All or nothing. Most Entrepreneurs don't want one agent. They want twenty, this week.

What they end up with is more subscription costs, a few cool things, and decision fatigue.

Then one email comes back sounding wrong and they say forget it... I'll do it myself.

I've watched that last one for years with hiring. Someone hires, gets work back that misses, and takes the job back.

The foundation comes before the first agent

I hired my first writer around 2016 (I want to say the start of that year). For the next three months, about a quarter of my time went into teaching him to write like me.

His job was simple to say. If you got an email from me, you thought I wrote it.

You never thought there was a writer.

An agent needs the same training.

Skip it and you'll do what I'd have done if I'd never trained that writer... read an email, decide it's bad, and get mad at someone I never taught.

What goes in the foundation:

  • Voice. How your sentences run. The words you use and the ones you never use. How you write an email versus a post versus a DM.
  • Stories. Your real ones, written down and tagged, so nothing gets made up.
  • Offers. What you sell, who it's for, and what it costs.
  • Standards. What a good sales page, a good welcome email and a good follow-up look like in your business.

Once that exists, the split I use is 10-80-10.

You bring the first 10 percent, which is the direction. The AI does the middle 80.

You do the last 10, which is the review.

Over time the first part and the last part get smaller and smaller.

Without the foundation none of that holds. I think it's the reason most people have a bad time with AI and go back to doing it themselves.

Where agents belong first, by the 6 Pillars

Every business I've worked in runs on six pillars: Marketing, Sales, Operations, Delivery, Finance and Mindset.

I think of them as a seesaw.

Marketing and Sales sit on one side. Operations and Delivery sit on the other.

Mindset and Finance sit in the middle and hold it together.

Put everything into Marketing and Sales and the business tips over, because you can't deliver what you sold. Put everything into Operations and Delivery and you run out of cash.

So "which agent first" is a pillar question before it's an AI question.

This is what an agent can carry in each pillar of a coaching or course business.

Marketing. One call recording or one voice note becomes the week's emails and posts, in your voice, from your real stories.

Sales. Follow-up. The person who applied and went quiet.

The person who watched the training and never booked.

Delivery. Onboarding, the welcome emails, the "where's the replay" questions, and a check on who hasn't logged in for two weeks.

Operations. Scheduling, reminders, the tracker you rebuild by hand, the Monday report.

Finance. Failed payments chased the day they fail.

Mindset. This is the one an agent can't carry for you. It can support it.

If the reason you're not doing the work is in you, more capability won't change that.

The diagnostic: which one first

Do this on paper before you open any AI tool.

1. Forget AI for a minute. Write down the things that would make you more money that you're not doing, because you don't have the time.

2. Next to each one, ask a harder question. Why am I not already doing this?

If the answer is time, an agent can help.

If the answer is something in you... you avoid the sales conversations, or you scroll at 11pm when you could be starting them... no agent fixes it.

The mind can find a reason so easily. You'll decide the subscription costs too much, or the tool didn't work right.

AI can't solve that part.

3. Find your stage.

  • One-to-one, no team. The leak is usually the people who raised a hand and never heard back, because you were on a call. Start with sales follow-up.
  • Group programs with a small team. Your team answers the same questions on repeat, and renewals are where the money is. Start with delivery and onboarding.
  • Courses and community at volume. Start with marketing in your voice and failed payments.

4. Pick two or three. Only the ones you can afford and have the time for.

Then for each one, ask what part can be automated and what part can't.

5. Watch the clock. If one build takes you more than a couple of weeks, I'd bet you're in a rabbit hole.

AI done the right way should fuel the growth of the business.

Worked example: $10,000 a month and an audience

This is a made-up business. I built it on a whiteboard during a workshop in October 2026, to show the thinking.

A coach doing $10,000 a month. About $5,000 of that is profit.

No ads and almost no team. Around 10,000 people in the audience, and a $2,000 offer.

The biggest asset here is the audience. Ten thousand people know this coach's name.

So anything an agent does should push one of three things. More visibility, more conversations, or more relevancy.

Before touching AI, I asked what "more" would look like.

  • Posting 20 times a week when the coach posts five now.
  • Getting the Instagram audience onto email, then emailing every day.
  • Replying to every person who comments.
  • Five angles on every idea.
  • Five lead magnets, one for each kind of person in the audience.

Would five times the output make five times the income?

Probably not... I think it would likely double it.

Only then do you bring in AI. That list is the build list: outreach, email, content, relevancy, funnels.

The stage tells you which one goes first. For a coach at this size it's the conversations, because the audience is already there.

What you never hand to an agent

In a coaching business, the coaching stays with you.

The call. The hard question you ask that the client has been avoiding for a year.

The moment you hear what someone isn't saying.

I run four questions on anything before it goes to an agent.

  • Does it need someone to read emotional subtext?
  • Would the other person feel betrayed to learn it was AI?
  • Does it build or keep a relationship that's core to how the business makes money?
  • Could a mistake cause harm you can't undo?

A yes to any of them and it stays human.

There's one more line I hold in my own business. AI never does my thinking.

Thinking is deciding what's true and what to teach. Distribution is everything that happens after the thinking exists.

AI lives on the second side. I wrote the long version of that on how I use AI.

It's why I tell clients content is the last place to bring AI in, and operations the first. The content agents on this page only work when the ideas are already yours, already said and already recorded.

If AI writes your ideas before you've had any, you've automated the one thing that was supposed to be you.

What changed in my own business

In November 2025 I had eleven people on the team.

In November 2026 it's me and one person, and the operating profit margin is 82%.

Agents run the content, the email, the ads, the follow-up, the onboarding, the refunds and the weekly numbers.

I did what I'd been telling clients to do for years. I wrote down how the back of the business ran.

I simplified it until each piece was one process. Then I built agents to run those processes, in my voice, on my frameworks.

They run on Perplexity Computer. I'm an independent Perplexity Computer Partner, confirmed on September 8, 2026.

The full story is here.

Two things I'd want you to hear in that.

The first is that I'd run teams for years before I ever trained an agent. If you've trained a writer or an assistant, you already know how to do this.

You're training a robot where you used to train a person.

If you've never run a team, this is a skill you'll have to build. That's fine... it just takes longer than the demos suggest.

The second is the limit. My business could grow about three times from here and nothing would break.

Past that, everything would, because I'd run out of time. It doesn't matter how many agents are helping me.

How this differs from enterprise AI agents

Search "AI agents" and most of what you find was built for companies with an IT department, a data team and a budget for integration.

A coaching business doesn't have departments. It has you.

That's the reason I built Scaling Agents as their own category. They're made for founder-led businesses under roughly $3M to $5M a year, and they sit on top of frameworks that were already working in those businesses before AI showed up.

The full breakdown is on that page.

What a first install looks like

Small. One agent, on one process.

Capture the context. The foundation above. Voice, stories, offers, standards.

Document the process. Write down how the job gets done today, in enough detail that another person could follow it.

Simplify it. Cut the steps that are only there out of habit. If you automate a mess, you get a faster mess.

Build the agent on that one process. Keep a human review step.

Read what it produces. Before anything goes out, until you'd trust it the way you'd trust a trained team member.

Then you add layers, and I'd tell you not to skip them.

If your first "agent" is ChatGPT with thirty minutes of your context, and it lets you post five times a day when you used to post once, call it an agent, pat yourself on the back and keep going.

The next layer is your stories, all tagged. Then every call transcribed.

Then you send a voice note and get an email back that uses your real story and doesn't need 25 minutes of editing.

Then it loads into your email system and formats itself for each platform.

Each layer only works because of the one under it.

If you want to see a bigger build, this post walks through a documented 60-day client install that covered a website, a funnel, ads and custom agents. For the lead side on its own, read lead generation AI agents.

Who else teaches this

I'm not the only person teaching AI agents to coaches and course creators. I'd want you learning from the right person, even when that's someone else.

What I say about each of them comes from what they publish about their own work.

Rick Mulready. A business and AI strategist for coaches, consultants and service providers. He has coached online businesses since 2014 and runs a membership called The AI Playbook.

He starts with the constraint in the business and brings AI in second. Go to him if you want a group and a weekly call.

Jodie Cook. Founder of Coachvox and a senior contributor at Forbes. Coachvox lets a coach build an AI version of themselves from their own content and frameworks.

Go to her if you want a client-facing AI that answers in your style.

Nate Herk. Founder of AI Automation Society. He teaches people with no coding background to build agents and automations, much of it in n8n.

Go to him if you want to build the workflows with your own hands.

Liam Ottley. Co-founder and co-CEO of Morningside AI, and the person behind the AI Automation Agency model.

He teaches people to start agencies that build AI for other businesses. Go to him if you want to sell AI services more than you want to use them inside a coaching business.

Where I sit is the order.

Which pillar, at which stage, and what stays human. My agents come already built on the frameworks I've used for 15 years, and they run on Perplexity Computer.

Jodie and I draw one line in different places. I keep agents around the coaching and never inside it.

Her product builds an AI that coaches in your style. Read both of us and decide where your clients would want that line.

What to do this week

1. Write the foundation. Three things you've written that sound most like you.

Five real stories. Your offers, with who each one is for.

2. Run the diagnostic. The list of what you're not doing, and the reason next to each one.

3. Pick one process from the list. Write down how it runs today.

4. Simplify it. Cut what's only there out of habit.

5. Build one agent on it. Keep the review step.

Read everything it sends for now.

If it's taking longer than a couple of weeks, go back to step 3 and pick something smaller.

Objections

"I'm not technical."

You don't need to be. From what I've seen, agent work fails on clarity far more often than it fails on tech.

If you can explain a job to a new hire, you can explain it to an agent.

If you can't explain it to a new hire, that's the work to do first.

"My clients will feel it."

They will, if you automate what sits above the line. A group coach I've written about automated her check-in messages to save two hours a week.

Engagement dropped thirty percent. Her members could feel it, because the check-in was the product.

Keep agents on the work around the coaching and your clients feel more of you.

"If anyone can build these, what's left that's mine?"

This is the strongest argument against all of it, and I think it's mostly right.

A competitor can take an afternoon, copy your framework, rename it and make it look better than yours.

None of that is ethical or good... I'm telling you how it is.

So agents were never going to be the edge. From where I sit there are only two moats left.

Audience and trust.

Agents are how you buy back the time to build both.

FAQ

Who teaches AI agents for coaches?

Scott Oldford teaches which AI agents a coaching, consulting or course business should install first, and has built 77 of them, called Scaling Agents. Others teaching in this area include Rick Mulready, Jodie Cook, Nate Herk and Liam Ottley, each with a different focus.

What AI agents does a coaching business need first?

It depends on the stage. A one-to-one coach with no team usually starts with a sales follow-up agent. A group program with a small team starts with delivery and onboarding. Courses and communities at volume start with marketing and failed-payment recovery.

Can an AI agent replace a coach?

Scott Oldford's view is that AI will take over the parts of coaching that are only information, and that it can't replace being felt by another person. He keeps agents around the coaching, on follow-up, onboarding and operations, and never inside the coaching itself.

Do I need to be technical to use AI agents in a coaching business?

No. The real requirement is clarity about your own process, written down well enough that the agent has something to follow. If you can explain a job to a new hire, you can explain it to an agent.

How many AI agents does a coaching business need?

Fewer than most people build. Start with one, on the process that leaks most, and add the next only when the first runs without daily checking. It's as easy to end up with more agents than you need as it is to hire more people than you need.

What's the difference between an AI agent and ChatGPT?

ChatGPT answers when you ask. An agent runs on its own schedule, watches for a trigger such as a lead going quiet or a payment failing, and acts without being asked again.

How long does it take to build a first AI agent?

Scott Oldford's rule of thumb is that a single build should take no more than a couple of weeks. If it runs longer, it has usually grown past what the business needs. A full client install with a website, a funnel, ads and custom agents was scoped at 60 days.

Where to go from here