Pre-context: I am a mentor, coach, and SaaS entrepreneur for 18 years. One of the goals is to create a new home for coaches & creators, and after a lot of conversations about what people need and see for the future, something interesting came up: AI is already being used by successful coaches.
The implementation lacks, because it is mostly based on custom GPTs, which isn’t a solution in this case. It is still a pioneering phase, but what we have learned in the software engineering space is transferable to coaching and teaching.
Today, I want you to understand what I learned, why I am keen to integrate AI into my coaching & mentoring business, and why it became part of the Q1 roadmap for pagebar.site.
“AI coaching is on the radar far more than I expected”
In the last few weeks, something has been repeated in conversations with coaches and mentors: AI coaching is on the radar far more than I expected. Not as a gimmick, not as “another tool”, but as a serious question of leverage, quality, pipeline, and trust. That matters because coaching is not a content format; it is a process of change, and change is never just intellectual. It is emotional, behavioral, social, and often painfully practical.
2025 was the dawn phase. Curiosity, doubt, experimentation, quick wins, quick disappointments – and a lot of doubt. This year feels different. AI is no longer “new”; it is simply present, like email, like Slack, like the smartphone. AI is quietly becoming infrastructure. The question is not whether AI will be used in coaching. The question is what kind of coaching it will produce.
Here is my critical thesis, with a clear pro-AI bias in coaching: AI will not replace good coaches; it will amplify the gap between coaches with substance and those with only output.
And the most important part is not the model. It is the design.
AI is already the number one coach in usage, not in quality
I’m careful with that sentence, because it can be misunderstood. I do not mean AI is the best coach. I mean, it is already the most used reflection partner. People are using it to interpret a letter from the tax office, to sanity-check a symptom before seeing a doctor, to work through a conflict before sending a message, to understand a concept they did not understand five minutes ago. The moment the question appears, the answer is accessible. That changes behavior.
Coaching, mentoring, and teaching, these are all forms of learning. Learning is not a weekly meeting; it is a daily loop. So if AI is already in everyone’s pocket, always available, always responding, it becomes part of the learning environment, whether we like it or not. The relevant question becomes: do we integrate it into the coaching process in a way that improves quality, or do we ignore it and let it happen in the shadows, without guardrails, alignment, or context?
Experience is the input, judgment is the difference
An AI coach is only as good as the coach behind it, minus the judgment call. That sentence kills two fantasies at once: the fantasy that AI can magically create depth, and the fantasy that a “coach brand” equals a coaching system.
If there is no track record, no years of sessions, no proven structure, then the AI has nothing to compress into a real point of view. It will sound like polished generic advice, which is exactly what a standard model already produces. You can add tone, you can add vocabulary, you can add a few principles, but you cannot fake lived patterns.
Where AI becomes interesting is when the coach has years of consistent reps, recurring cases, refined frameworks, and a language shaped by real humans in real moments. Then AI becomes a compression layer, not a creativity layer; it can return the coach’s patterns back to the client at the moment the client needs them, not one week later.
Judgment is the non-automatable core
Pattern recognition can be automated; judgment cannot. AI can help someone map options, surface trade-offs, reduce blind spots, and clarify the decision space. But there are moments where responsibility is the work, where you need a human who can challenge you, hold you accountable, and carry the moral weight with you. If you are making a high-stakes decision, for example, a leadership decision that impacts livelihoods, do not outsource the decision to a chatbot. Use AI to fill information gaps, then take the judgment call to a human you respect.
Coaching is daily, sessions are weekly, and the gap is where momentum dies
If you see coaching as a holistic process, you know the real work happens between sessions. The coaching session is a checkpoint; the week is the battlefield. Discipline, Resistance, avoidance, shame, fear, negotiation with your own standards, that is daily weather.
Here is the recurring situation:
“It’s Tuesday evening. I finally have time to reflect. I’m stuck. I could message my coach, but I don’t want to disturb them. I keep it short, I’m not even sure how to phrase it, then I wait. When the answer comes, it’s on-the-go, and we both know it’s not the best version of the coaching.”
That pattern is not a character flaw; it is a structural gap. AI can bridge it, not by pretending to be the coach, but by holding the client inside the coach’s system when the coach is not present. This is where I see AI coaching as a scaling tool, and depending on positioning, even as its own product layer that feeds into higher-touch work. Not because it replaces the relationship, but because it protects momentum.
Reflection speed becomes a competitive advantage
People who can reflect and adjust in real time progress faster than those who wait for scheduled feedback. Coaching often fails, not because the session was bad, but because feedback arrives too late to shape the next action. AI shortens the loop between thought, insight, action, and correction. This is the part coaches should take seriously: if your client wants to do the work in the moment they have energy, Tuesday evening matters.
Knowledge does not “sit” in a PDF; it has to live
Many AI coaching experiments fail for a simple reason: they treat knowledge like a file upload. Dumping text into a model and hoping for the best is not a strategy; it is a lottery ticket.
If this becomes infrastructure, then the “coaching brain” has to be maintained like any serious system. It needs condensation, structure, revision, and drift control. It needs to turn transcripts into principles, principles into decision rules, and decision rules into language that still sounds like the coach, but stays precise under pressure.
This is why I believe coaching AI will move from “prompting” to “frameworks”. Prompting is brittle and personal; frameworks are stable and transferable. The coach should not be forced to become a prompt engineer. The platform should provide a framework for condensation that keeps the knowledge base usable, grounded, and consistent.
Guardrails have to be strict; confidence matters more than output
This is the part that decides whether AI coaching becomes trusted or becomes a reputational risk. Hallucinations, confident nonsense, and invented certainty are unacceptable in coaching because coaching often happens at the edge of someone’s identity, decision-making, and relationships.
The goal is not to deliver a response. The goal is to deliver the right response with the right confidence, or to refuse.
That means strict guardrails, clear escalation rules, and explicit confidence signaling. “I’m not sure” is not a failure; it is a feature. The system should be able to say: I can answer this confidently; I can give options, but I’m uncertain; I cannot answer this responsibly, bring it to a human.
This also unlocks something powerful: the AI becomes a feedback instrument for the coach. Not feedback in the fluffy sense, but operational feedback: which questions keep appearing, where the knowledge base has gaps, where users ask for something the system cannot answer without risk. That creates a review loop where the coach improves the coaching brain week by week, instead of shipping a static GPT and hoping it behaves.
AI needs to be coached, too
In software, we learned that tools do not save weak processes. A team that already ships reliably becomes faster with AI; a team that lacks discipline becomes louder. The same applies to coaching. AI amplifies substance, not credibility. It accelerates what is already there, good or bad.
If you want a practical analogy: TDD is not a programming language feature; it is a feedback framework. AI agents became powerful in software when they were wrapped in frameworks and workflows, not when they were just chatbots. Coaching AI is the same. Without a framework, you get vibes. With a framework, you get deliberate practice.
Personal note: disciplined people experience the gap more sharply
I work on myself daily, without exception. Quiet moments create openings, a thought appears, an insight is near, and I want to continue right now, not tomorrow. That is where session-based coaching hits a hard limit: waiting kills the loop.
This is why AI is so tempting, and why it can be so effective when used correctly. Not as a replacement for human coaching, but as a bridge that keeps the loop alive between sessions.
Privacy and governance are not “extras”; they are the product
A client needs a space to think out loud without being judged. That includes messy questions, awkward framing, and half-formed thoughts. Reflection works because it is private; a mirror book works because the page does not interrupt you. AI can bring that privacy into an interactive format, but only if it is designed as privacy-first rather than surveillance.
The coach should not receive raw messages by default. The coach should receive a partially anonymized, human-readable summary only when the client consents to share. The client stays safe; the coach still gains context; the next session starts deeper because the client has already worked within the coach’s system before entering the call.
Governance matters here. Without governance, AI becomes a parallel universe, a solo adventure with generic advice. With governance, it becomes part of the coaching relationship; boundaries, standards, and intent remain owned by the coach and chosen by the client.
A simple integration pattern that is already better than “use a custom GPT.”
If you want a concrete way to think about this, keep it simple:
First, private reflection: the client uses AI as a mirror to work through the messy version.
Second, structured summary: the system turns that into a short, human-readable summary, optionally anonymized.
Third, session alignment: the coach receives only what is needed to coach, and the next session starts with the real topic.
Fourth, feedback loop: the AI reports uncertainty and missing coverage back to the coach, so the coaching brain evolves.
This respects privacy, increases momentum, and makes coaching sessions more valuable.
So, will AI replace or enhance coaching
I don’t think AI can replace good coaching, because coaching is not just about knowledge transfer. It is timing, accountability, judgment, and relationship. But AI will enhance coaching in a way that changes expectations: clients will want support the moment they are ready to do the work, not only when the calendar allows it.
I’m pushing to build this capability into ᴘᴀɢᴇʙᴀʀ.site for a selfish reason: I want it for my own coaching and mentoring. I want the bridge. I want the privacy. I want the strict guardrails. I want the feedback loop. I want the client to maintain momentum on a random Tuesday evening, then bring the real topic into the next session with clarity rather than chaos.
One question I keep coming back to, and I’d genuinely like your take: what is the one boundary you would insist on, before you would ever let AI touch your coaching process?
—Adrian
