Some time ago I earned my Practitioner Diploma in Executive Coaching, but my curiosity and deep technical background wouldn't let me rest. A question kept nagging at me: "Can AI handle quality coaching and make it accessible to everyone?"

So I conducted two experiments simultaneously:

  1. Create an AI voice assistant capable of delivering coaching comparable to human Executive Coaches
  2. Build a production-ready app with AI coding tools from scratch in just 1 working day (8 hours)

(Spoiler alert: The answer to both questions is YES)

The AI Coaching Opportunity

During my coaching studies, I discovered that coaching can truly be game-changing for personal development when done well. If meditation is the still lake that reflects your mind, then coaching is the compass that guides it - making your next moves crystal clear. Good coaching helps you navigate the maze of possibilities by asking the right questions rather than providing direct answers.

During my Executive Coaching studies, I realized something interesting – what makes coaching effective  aligns perfectly with AI's strengths:

  • Non-judgmental listening
  • Asking powerful questions without offering direct advice
  • Creating a safe space for self-exploration
  • Following established frameworks like GROW model rather than improvising

This realization led me to wonder: could AI provide this guiding framework while making it affordable for everyone? After all, the structure of effective coaching follows patterns that AI can do remarkably well!

Also, since privacy is essential, I decided to develope a client-side solution that keeps all personal information on your device only – only voice data passes through OpenAI's API for transcription and response generation.

The 8-Hour Development Challenge

Using ChatGPT for generating app design and UI mockups and Claude.ai and Claude Code for creating the actual code, I built a complete AI coaching application in about 8 hours (split between multiple short sessions, having duties in-between).

The result? An AI coaching experience that costs approximately 1 eur per 15-minute session – compared to 50-70 eur for the same time with a human coach.

AI Coach App Features

AI Coach App Main View

App is pretty minimalistic yet has everything essential for quality coaching experience:

  • End-to-end Voice sessions with AI coach (similar to Advanced Voice Mode in ChatGPT)
  • Each session is built so that it is a stepping stone toward the coachee's (clients) broader vision. But, at the same time during each session coachee can talk about anything he/she wants
  • After each session app summarizes session and extracts action points what coachee must complete by the next session
  • Session-by-session app envisions bigger picture and long term goals of the coachee
  • AI Coach uses my personal coaching model that I have developed during my studies
  • App is client-side, meaning it keeps all personal information (sessions summaries, action points, transcriptions, etc.) on your device only, so there is no any servers where your information is stored
  • No registration or login required, the only thing you need to set in App Settings is OpenAI API key for using it for voice sessions 

What AI Coaching Does (And Doesn't) Deliver

After some testing, I have found AI coaching delivers what I was hoping it will deliver (it actually does even better!), especially for those uncertain about investing in traditional coaching. However, there are some trade-offs:

  1. Self-discipline is essential – Without the accountability of meeting a real person, you might skip sessions. You need to understand that the only person you're harming is yourself.
  2. It lacks some human elements – The beautiful imperfection of human coaches, their occasional desire to share relevant experiences, or those seemingly illogical but ultimately brilliant creative insights that human coaches sometimes offer.

My Tips for AI-Powered Development

For my technically-minded connections, here are the key takeaways:

  1. Focus on single functionality first – If you have a large solution with multiple modules/functionalities, pick a single one and focus on it.
  2. Make your homework before asking AI to generate code – Think through what exactly you want to build, find latest documentation related to your problem (e.g. provide openapi docs if you are going to use some 3rd party service, etc.), or at least give hints about what approach, libraries or methods to use.
  3. Take time for high-quality initial prompts – Be detail-oriented as possible when describing your goal. In your initial prompt, describe your final goal (for the specific functionality), focusing more on the main functionality of the solution – use more of a waterfall approach over incremental.
  4. Fixing bugs costs more than building – Fixing some small bug could take more time than creating 95% of the entire solution, which is why the previous point is so important!
  5. Software development expertise remains essential – While AI can replace a small development team and compress weeks of work into hours, it still requires a skilled developer to guide it. AI sometimes takes wrong directions with technical decisions, and you must be able to identify it and provide instructions for course correction. The efficiency is transformative, but AI functions best as an acceleration tool for developers rather than a complete replacement - someone who understands code and architecture must evaluate and direct its output.

The Verdict

AI isn't about replacing human coaches entirely, but about making coaching accessible to everyone.

Similarly, AI development tools aren't about eliminating developers, but empowering them to create solutions that were previously impractical due to time or budget constraints.

In both cases, we need to be thoughtful about how we use these tools, following our inner compass to determine when AI is sufficient and when the human touch is worth the premium.

If you're interested in trying the AI coach yourself, here is the link to the app or if you want I can share source code in GitHub (just DM me). Since it's client-side (ensuring your coaching data stays in your hands only), you'll need to set up an OpenAI API key first (just google "How to get an OpenAI API key" or find video tutorials on YouTube).

Not technically inclined? Message me in LinkedIn, and I'll help you get started.

What do you think about AI in coaching? Would you try an AI coach, or do you see unique value in human coaching worth the premium?

Share this post