Most AI advice for coaches is thin. Use ChatGPT for captions. Use Otter for notes. Use Midjourney for your carousel hero. Fine, but none of it moves the dial on the thing you actually sell. It just makes your admin slightly less annoying.
If you want AI to grow the business, not decorate it, you have to think in three layers: tools, agents, and workflows. Most coaches skip the second two and wonder why their “AI stack” feels like a ChatGPT subscription with extra steps.
This is the practical version. What tools to use, what agents to build, how to chain them into workflows that actually work, and a 30-day plan to ship something real.
The three layers, defined
These terms get used interchangeably and they shouldn’t. Being precise here saves you money and confusion.
Tools are off-the-shelf products with AI features built in. You pay a subscription, you get a useful thing. Otter transcribes your calls. Fathom summarises meetings. Claude and ChatGPT draft copy. You don’t build these. You buy them.
Agents are custom AI systems that do a specific job in your business. An intake agent qualifies new leads. A check-in agent runs weekly accountability touchpoints. A content agent surfaces the right resource at the right moment. Agents combine a model (Claude or GPT under the hood) with your instructions, your knowledge, and the tools to take action. You either build these yourself or commission them.
Workflows are the glue. They chain tools and agents together so a single event (a new booking, a completed module, a quiet week) triggers a sequence of AI-powered steps without you in the middle. Workflows are where 90% of the time saving actually shows up. They’re also where most coaches’ AI plans fall apart, because they never build any.
If you’re using tools only, you’ve saved five minutes here and there. If you’re using agents, you’ve automated specific tasks. Only when you have workflows does AI start to feel like it’s running the business alongside you. That’s the shift our complete guide to AI for coaching businesses covers in full.
Layer one: the tools worth paying for in 2026
You don’t need all of these. You need the ones that match your specific friction. Here’s how to think about it by use case.
Meeting transcription and summaries. Fathom, Otter, Fireflies, Granola. Pick one, use it on every call. The transcript alone saves you 10 minutes of note-taking per session. The summary saves another 15 when you’re following up. This is the highest-ROI tool a coach can add in a day.
AI writing. Claude and ChatGPT are the workhorses. Claude Sonnet 4.6 is better for long-form nuance (emails to high-ticket clients, framework docs). GPT-5.x is better for quick iteration (social copy, briefs, drafts). You can run both. Paid plans are around £15 to £20 a month each.
Content repurposing. Descript for video (turn a 45-minute webinar into 10 clips). Gamma for decks (rough Loom outline to polished deck in minutes). Opus Clip for short-form repurposing. If you publish content, one of these will pay for itself in the first week.
Research and briefs. Perplexity for quick researched answers with sources. Claude for deeper analysis of a document you paste in. If you’re doing any competitive research, market sizing, or background on a prospect before a call, these are faster than Google.
Client communication. Your existing inbox, augmented. Superhuman’s AI drafts are genuinely fast. Fyxer for email auto-triage. If email is eating your week, one of these is worth the £25/month.
Voice and scheduling. Bland.ai and Vapi for AI phone agents that actually sound human. Reclaim or Motion for AI-assisted calendar management. These are newer and rougher around the edges but moving quickly.
The rule: if you’re evaluating an AI tool, ask “what’s the specific hour this saves me each week?” If you can’t answer in one sentence, skip it. The AI tool market is flooded with solutions looking for problems.
Layer two: the three agents every coaching business should build first
Tools are off-the-shelf. Agents are custom. This is where you stop being a consumer of AI and start having AI work specifically for your business.
You don’t need a dozen agents. You need three good ones. Build them in this order.
Agent one: intake and qualification
The problem: every inbound enquiry eats 30 to 60 minutes of back-and-forth before you know if the person is even worth a call. Most coaches do this manually. Some use a generic form. Neither works.
What an intake agent does: holds a conversational intake with a new lead, asks the follow-up questions your rubric would ask, scores them against your ideal-client profile, and either books them directly into a qualifying call or sends them to a self-serve nurture track if they’re not ready. By the time you see the lead, you’ve got a written brief of their situation and a recommendation.
How to build it: start simple. A Claude or GPT conversation with a custom prompt that encodes your qualification questions and scoring rules. Host it on a page, link it from your site. For a basic version you can build this yourself in a weekend. For a production version that ties into your CRM and calendar, commission it.
What it saves you: typically 4 to 6 hours a week for a coach handling 10 to 20 enquiries. Plus: dramatically better call-show quality, because the unqualified leads have been filtered out before they reach your diary.
Agent two: weekly check-ins and drift detection
The problem: with 20+ active clients, you can’t maintain real accountability without a calendar blocked full of 15-minute check-ins. With 50+ it’s mathematically impossible, so you don’t, and completion rates slide.
What a check-in agent does: runs a weekly asynchronous check-in with every active client. Asks the specific questions your framework calls for at their stage. Spots when someone is drifting (missed homework twice, shorter answers, change in tone). Drafts a response to handle routine stuff. Flags the 3 to 5 clients each week who actually need a human conversation.
How to build it: this one needs retrieval (your framework docs + each client’s history) plus an agent that follows a weekly rhythm. More work than an intake agent. Expect 2 to 4 weeks of build time for a first version, longer if you want it polished.
What it saves you: 6 to 10 hours a week for a coach with 30+ active clients. More importantly, the reason average online course completion sits at 5% to 15% without support vs 70%+ with it is accountability. This agent is the accountability.
Agent three: content and resource serving
The problem: you’ve got a library. Frameworks, Looms, worksheets, case study stories. Clients rarely find the right thing at the right moment. You’re the search engine, which means they ping you.
What a content agent does: sits on top of your library, matches client context to the relevant resource, and surfaces it automatically. When a client mentions an objection in a check-in, the agent pulls the Loom that addresses it. When they finish a module, it queues up the next best thing for their specific situation.
How to build it: this is a retrieval pattern. Your content gets indexed (embeddings), the agent retrieves the most relevant resource for a given query, and the LLM writes a short personalised wrapper. Straightforward to build with the right setup, typically 2 to 3 weeks.
What it saves you: less raw hours than the first two, but it dramatically improves client experience. Clients feel like you’ve got a photographic memory of every resource you’ve ever made. Testimonials improve. Referrals improve.
Build intake first, check-ins second, content third. Not because they get harder (they do), but because that’s the order of ROI for a typical high-ticket coaching business.
Layer three: the workflows most coaches never build
This is the unsexy layer where the real compounding happens. An agent alone is useful. An agent chained into a workflow is a business.
Here’s what I mean by a workflow. Start with a trigger. End with an outcome. Everything in between is automated.
Workflow example: new client kickoff
- Trigger: client books kickoff call
- Step 1: agent sends pre-call intake (client fills it out)
- Step 2: transcription tool captures the kickoff call
- Step 3: summary agent produces the session brief
- Step 4: content agent queues up the first week’s materials
- Step 5: check-in agent schedules its first weekly touch 7 days later
- Outcome: client has everything they need, you’ve spent 60 minutes instead of 4 hours
You don’t do any of steps 1 to 5. The workflow does. You show up for the call, do the coaching, and the system handles the wrap.
Workflow example: content publication
- Trigger: you finish a Loom explaining a concept
- Step 1: transcription agent pulls the transcript
- Step 2: writing agent drafts a blog post in your voice
- Step 3: writing agent extracts 3 LinkedIn posts and 5 tweet hooks
- Step 4: content agent indexes the new resource into your client library
- Step 5: you approve or edit, everything goes live
- Outcome: one recording becomes 10 pieces of content and one new library resource
Same principle. Workflows are how AI stops being “a tool I use” and starts being “a team member who runs the pipeline”.
The tooling for workflows has improved sharply. You can stitch simple ones together in Zapier or Make with AI steps. More complex ones usually need custom code, which is where commissioned platforms come in. Most high-ticket coaches outgrow Zapier-style workflow tools somewhere around the 5-workflow mark, because the plumbing starts to break and the maintenance becomes its own job. This is the frankenstack problem we cover in custom coaching platform vs Kajabi. Run your stack through the frankenstack cost auditor to see what the duct-tape approach is already costing you.
The 30-day quick-win plan
If you’ve read this far and want to actually ship something by next month, this is the plan. No theory, no “audit your business”. Just do these.
Week 1: Add transcription to every call. Sign up for Fathom or Otter. Install it on your calendar. Every call, recorded and summarised automatically. Spend the 20 minutes it would have taken you to write notes on something else. You’ve bought back 2 to 3 hours in week one.
Week 2: Build a simple intake agent. Spend 4 hours on a Saturday. Write out your ideal-client qualification questions. Paste them into Claude or GPT with a prompt that says “have a conversation with the lead, ask these questions as naturally as you can, score them against this rubric, and produce a brief.” Share the chat link on your website. You now have an AI that qualifies while you sleep.
Week 3: Automate one follow-up workflow. Pick the most repetitive post-call action you do (send resources, book next session, update notes). Connect your transcription tool to Zapier or Make. Pipe the summary into a Claude step that drafts the follow-up email. Approve or edit, send. You’ve built your first workflow.
Week 4: Measure and pick the next one. Add up the hours you got back. Figure out which layer is pressing hardest (still drowning in check-ins? build the accountability agent next. Still drowning in ops? automate CRM updates). Pick one thing for month two. Repeat.
In 30 days you’ll have a transcription habit, one custom agent, and one workflow running in production. That’s more than 95% of coaches have. It’s also the foundation for a platform if you keep compounding.
When off-the-shelf tools stop working
Here’s the unspoken truth about AI tools for coaches: they get you to a ceiling, then they cap you the way Kajabi caps you. The ceiling is usually around 30 to 50 active clients and 5 or 6 tools duct-taped together.
The signs you’ve hit it:
- Your workflows break every time a tool updates
- You’re spending more time fixing automations than using them
- Your “AI stack” has 4 separate subscriptions and none of them talk to each other
- You can’t tell a client where they are in your framework without opening 3 apps
- You’re the only one who knows how any of it works
When you hit the ceiling, the move isn’t a 7th tool. It’s a custom platform that owns the whole delivery layer, built around your specific framework. This is the shift our complete AI for coaching businesses guide covers, and the productisation path our pillar guide on productising your framework walks through step by step.
If you’re not sure whether you’ve hit the ceiling yet, the platform readiness quiz takes about 3 minutes and gives you a scored read across the four dimensions that actually matter: framework proof, demand signal, operational readiness, and business model fit.
The bottom line
AI grows coaching businesses in three places: the tools you buy (small wins, fast), the agents you build (bigger wins, slower), and the workflows that chain them together (the compounding win).
Most coaches use the first layer and wonder why they’re still drowning. A few build a couple of agents and start seeing real time saving. The ones who build workflows, or commission a platform that has workflows baked in, stop trading hours for revenue.
You don’t have to do all of it. You do have to do more than “I use ChatGPT sometimes”. Pick the layer, pick the quick win, ship it in 30 days. Then do the next one.
Related reads
- AI for coaching businesses: the complete guide. The pillar. The five layers of AI in coaching, the four technology categories, how it all fits together.
- How to productise your coaching framework into a scalable SaaS. What it looks like when the whole framework runs in software, not just individual agents.
- How to turn a 1-on-1 coaching programme into a scalable digital product. The practical move from “I use some AI tools” to “my programme is a product”. Covers the 3 product shapes, the 5 steps to get there, and how to price the new offer.
- Custom coaching platform vs Kajabi: what growing coaches actually need. Where off-the-shelf AI tools stop working, and what a purpose-built platform replaces them with.
- Platform readiness quiz. Eleven-question diagnostic. Tells you whether you’re ready to move from AI tools to an AI platform.
- Frankenstack cost auditor. Add up what your current tool stack actually costs, including the hours you lose to admin.
Built a few workflows and ready for something that runs the whole delivery? Book a free architecture audit. We’ll map your current stack, show you what a custom AI platform would replace, and give you a realistic scope before you commit.