AI in practice

Automate Your Coaching Business: 9 Practical Uses for AI

By Artur Neumann. Updated October 21, 2026.

I run my business in Germany. Monetary examples use euros unless another currency is stated.

This week, I had a coaching session with a client. He’s a coach himself, and his business is thriving. He has a well-coordinated team, a clear sales process, and his closers have fully booked calendars.

Despite this success, he was missing something: time. For each new client, he recorded a personalized welcome video in the evenings. These videos were 16 to 18 minutes long, sometimes several per day. In addition, he sent voice messages to potential partners, created content for Instagram, and spent a lot of time managing data in his CRM system.

He showed me his workday on the screen. I told him what I would do differently. Here are the key takeaways. I’m omitting his name, industry, and specific numbers. The advice applies to almost any coaching or consulting business. If you’re interested in learning how to acquire clients as a coach, you can find more information at Acquiring Clients as a Coach.

The Best Automation Isn’t Always About Automation

Before you automate anything, ask yourself: Does this even need to be done by anyone?

This might sound trivial, but asking this question saves the most time in practice. If you can achieve a goal without doing anything, that’s better than any automation. Building, testing, and maintaining an automation takes time and effort. Eliminating a task completely removes its cost.

For example, my client manually entered cancellations into several different systems. However, many payment providers can handle cancellations themselves. In that case, it’s sufficient to simply register the cancellation within their system. The provider handles the rest.

Take a look at your week and ask yourself this question for each task. You’ll almost always find two or three tasks that no one would miss if they were eliminated.

Separate the Personal from the Repetitive

The most important idea from our conversation: Every task is made up of smaller sub-tasks. A small portion of those tasks is truly personal. The rest is repetitive.

Let’s take the welcome video as an example. My client didn’t want to give up on the video, and I think he’s right. His clients are primarily supported via chat, and they rarely see their coach in person. Personal interaction is very important to him. The video gives his clients a sense of security. It’s one of the reasons his coaching is so successful.

So, we broke down the video:

1. The first one to two minutes are personalized. He greets the new client by name, referencing notes from the sales team. He reassures them that they are in good hands.
2. The remaining 15 minutes are almost always the same. He explains how the support process works, showing the client’s profile within his coaching software.

This second part doesn’t need to be recreated every time. He records the explanation once, in a high-quality recording. For each new client, the AI pulls up their profile, records the screen, and clicks on the appropriate elements at the right moments. Then, it combines the greeting and explanation into a video. The link is then placed in the client’s profile, where he quickly reviews it and approves it for sending.

This reduces the time per client from 18 minutes to approximately two. The part that matters, the personal touch, remains authentic. We simply eliminate the parts he was repeating anyway.

Interestingly, his coaching software doesn’t have an API. That doesn’t stop us. Claude can operate a computer like a human: opening browsers, clicking, recording the screen. Almost everything you do on a computer can be automated today.

Ask yourself this question for every recurring task: What part of this is truly personal? Often, it’s less than you think.

Voice Messages: Just the Name is New

The same principle applies to voice messages. My client reaches out to potential partners on Instagram. Those who respond receive a three-minute voice message from him, which is almost identical each time. Only at the beginning does he mention their first name.

Here’s how I would build it:

1. He provides Claude with a few old voice messages. Claude analyzes them and suggests which parts remain consistent.
2. He records the consistent part once, in a clean, high-quality recording.
3. Using a voice cloning tool like ElevenLabs, we create a clone of his voice. The clone only speaks the first sentence: “Hey Julia, great to hear from you.”
4. The AI combines these elements and presents the message for approval.

The AI pulls the first name from the chat, not from their username. Otherwise, it might greet someone with “Hey FighterHeart87.” These small details are crucial for making a message feel personal.

Initially, a human reviews each message before it’s sent. They only need to check the first few seconds, the rest is the authentic recording. If the name sounds wrong, they click “regenerate.” If everything sounds right, they click “send.”

I want to emphasize two boundaries. A human still writes the initial message to a potential partner. We only trigger the automation when someone responds, and we use the official Instagram API. I would never automate mass outreach to strangers.

Second, be transparent about using AI-generated voices. The EU AI Act often requires that AI-generated voices be clearly identified as such.

Start with Low-Risk Opportunities

My client had a valid concern. He said, “I’m worried about losing focus.” You can automate so much. If you try to do everything at once, you’ll end up achieving nothing.

I would start with opportunities that combine two factors: the task takes up a significant amount of time, and it can be automated without jeopardizing any client relationships. For him, that’s the carousel posts on Instagram.

His previous process worked like this: Claude wrote image ideas for him each morning. He copied them into an image tool, saved the images, sent them to his phone and uploaded them. Then he set up the keyword in his chat tool, built a separate landing page and connected it to his CRM. Roughly one hour per post.

The new process:

1. Claude generates images directly through the interfaces of the image models, mimicking the style of his previous posts.
2. Claude writes text for the images and the post description, and chooses a keyword.
3. When someone comments on the keyword, that person automatically receives a direct message.
4. The landing page is created as a copy of a proven template. Each entry is added to the CRM with the appropriate tag, along with information about which carousel it came from.
5. My client reviews everything and approves it.

If anything isn’t perfect, he won’t lose any customers. He sees each image and caption before anything goes live. And because his carousels receive a lot of comments, he can post multiple times a day instead of just one.

A Guiding Principle: First Simplify, Then Optimize. Claude replicates the existing process, pulling data from his inbox and CRM, and recreating it exactly as it is. He uses the same messages, the same page, and the same tags. We’ll explore ways to improve the funnel later. Changing both aspects simultaneously can make it difficult to determine what impact each change had.

I personally take it a step further on Instagram. For all Reels, I use a single keyword. The AI recognises the topic of the post and sends the relevant material. You can read about how I acquire leads using AI in Winning Customers with AI.

Prepare, Approve, Release

Many entrepreneurs worry that AI might send something inappropriate to customers. This concern is valid. Therefore, I recommend automating tasks in three stages:

1. Technical Assistants: Forms, tags, a chatbot with pre-written responses. You or your team still handle the most important tasks. Most entrepreneurs are at this stage today.
2. AI Prepares, You Approve: The AI writes the message, edits the video, and builds the page. A human reviews everything before it’s released. Even this stage saves most of the time.
3. AI Does It Alone: Only after you’re no longer needing to make corrections after hundreds of iterations.

For Stage 2, Claude has a strict rule: Nothing goes to customers or the public without my approval. This rule is documented in the instructions. A single sentence in the chat isn’t enough.

I also use AI in my own work. An AI assistant reads my emails every hour and drafts responses. I personally send the messages. It learns from every change I make.

Your Business Brain Runs Even When Your Laptop is Closed

Many of these processes need to run around the clock. If someone comments on a keyword at 11 PM, your laptop is probably closed. Therefore, you need a system that’s always running: a small server or hosting with SSH access, providing a secure connection for Claude to operate. This is where your business brain resides, on a dedicated subdomain with a password. It connects to Instagram, your CRM, the image models, and your voice system.

This sounds more technical than it is. Claude can set this up step-by-step. You simply log in where necessary and copy access keys. The initial setup is the biggest hurdle. Once it’s done, you’ll suddenly realize what’s possible.

The second part of the brain is memory. My client had a routine that sent him a daily overview. It looked slightly different each day, which made him feel uncertain. He ended up double-checking everything himself.

This was due to two factors. The routine was running on a “medium” intelligence setting. For verification tasks, I would recommend a setting of at least “extra high.” Also, the AI didn’t have a written memory. Claude doesn’t forget anything that’s stored in a file. Create a folder with briefings and learnings as Markdown files (simple text files). Any feedback you provide goes in there. This way, the routine improves with each passing day. I show you how to set up Claude in my guide to Claude Code in German.

Code is far better than any human for these types of checks. My client is a stickler for order. He double-checks run times and commissions. A human will eventually miss a date. Code won’t.

Protect What’s Yours

My client keeps his contacts, processes, and videos with a single provider. If that provider closes tomorrow, a significant portion of his business will be lost. Therefore, I recommend automatically backing up everything every night:

• All contacts from the CRM to his own server.
• All videos and images to an affordable storage solution, such as a storage box or Google Drive.
• Everything Claude builds, with version history, to GitHub.

This costs just a few euros per month. You won’t even notice it.

Also, I give every entrepreneur this rule: Don’t buy any tool without an API. An API (Application Programming Interface) allows other programs to work with the tool. Without it, it’s difficult to automate. Ironically, for my client, this applies to customer support, which is his biggest expense. I wouldn’t recommend a disruptive move like that, as the risk is too high. However, for any new tool, the API is the first consideration.

I’ll share a personal experience. For my previous email tool, I was paying around 300 euros per month, and it didn’t have a usable interface. I canceled it and built my own system. Today, everything is interconnected.

Re-engaging dormant contacts

A single action doesn’t just save time; it generates revenue. My client has many potential customers who filled out their questionnaire but never booked a consultation. Most aren’t lost. For many, something simply got in the way.

I would send them a very simple email every month. No logo, no image, just text, left-aligned, like a personal message:

Hi Sandra, we talked about your goals back in the spring. Is that topic still relevant for you? Best regards.

The AI uses names and the initial reason for contact, but you have full control. Anyone who responds is brought back into the regular sales process. I explain the concept in detail in The 7-Word Email. Only send these emails to people who have explicitly registered with you.

Tell the AI what you want, not how

Now, the most important point. Many of us have been trained to think within limitations. We first ask: How does this work? Who can do this? What will it cost?

With AI, “how” is no longer your problem. Imagine you’re talking to a very intelligent employee who knows your business. You don’t need to explain how something works. You tell them what you want, and they find a way. A helpful question is: What would you do if you had an unlimited number of employees and unlimited time? Because that’s essentially what you have now.

A practical tip from a recent conversation: We recorded it. In some instances, I spoke directly with Claude, not with my client. He now simply gives Claude the transcript and says: “Read what we discussed. Create a ‘vision’ file outlining the desired outcome. Then, start with the smallest step that will immediately save me time.” The conversation is the blueprint.

For in-person meetings, there are small AI transcription devices available. I’m sharing my experience with one of them in my Plaud-Note-Pro Test.

After the conversation, he asked: Am I doing all of this in a single conversation with Claude? My answer: No, one topic per conversation. One for the carousels, one for the welcome video, one for the voice messages. That way, Claude works on multiple topics simultaneously. Because all conversations are stored in the same project folder, they share the same vision, rules, and learnings.

What AI doesn’t replace

Pure chat support will become cheaper in the coming years. Anyone who types their question into ChatGPT often gets a usable answer today. For standard questions, you won’t need a coach anymore. At the same time, people are switching to my client from competitors whose chatbots provide poor support.

My conclusion: Automate everything around the relationship, not the relationship itself. The personal touch is the reason people buy from you. The time you save should be invested exactly there. In more genuine conversations, better results, and closer relationships with your customers. Those who do this well will win customers from those who automate everything, and from those who don’t automate anything at all.

Your first step

Choose a recurring task. Not five. Consider these five questions:

1. Can I completely eliminate this task?
2. Which part is truly personal, and which part is always the same?
3. Does it save a significant amount of time without jeopardizing the customer relationship?
4. Who reviews and approves before anything goes out?
5. Where is the data stored, and is there a backup?

Once you’ve answered all five questions, tell Claude what you want. You’ll figure out the rest together. I describe how small steps like these can build a self-reinforcing system in Flywheels in Business.