Human & AI
AI Can Coach. What Should Professional Coaches Change?
What should professional coaches stop, start and continue doing as AI coaching develops? A first-person reflection grounded in coaching practice, HR leadership and ethics.
Core question
If AI can coach, what should professional coaches stop, start and continue doing?
In brief
If AI can generate strong questions, a professional coach needs to offer more than questions. Stop premature interpretations, actively steward the agreed process, protect confidentiality and keep the client's decisions with the client.
01
The question I asked in the session
During Coaching.com's ‘AI Can Coach. Now What?’ session, I asked David Drake: if AI can perform some coaching tasks as well as, or even better than, humans, what should coaches stop doing, start doing and continue doing?
In his response, Drake emphasized understanding the client's social and organizational context, alongside presence and listening. Capabilities of tools such as Ovida were also discussed in the session. Those were the presenters' statements, not findings from my own evaluation.
For me, the question connects to something I have encountered both as a senior HR leader and while reviewing my own coaching: a coach's value cannot be measured only by the quality of the questions they ask.
02
What should we stop?
We should stop competing to produce ever more impressive questions. AI tools can suggest questions, exercises and conversation scenarios. If asking a good question is our only distinction, our professional value becomes difficult to explain.
We should also stop rushing to interpret the client. A manager who delays a difficult conversation does not necessarily lack confidence. Perhaps their authority is unclear, or disagreement carries a real cost in their workplace.
That attention to context comes from my HR experience: before treating a behavior as an individual problem, I look at the surrounding structures and relationships.
03
What should we start?
We need to become more explicit about responsibility for the process we agree with the client. Reviewing my own sessions showed me that client-centered coaching does not mean a passive coach.
The client owns the topic, meaning and decision. The coach helps make the session's purpose, a change of direction and the progress being made visible.
In organizational coaching, this becomes more complex. The sponsor may expect behavioral change while the session itself must remain confidential. Purpose, boundaries around shared information and progress reviews should therefore be agreed at the outset. A tool may help record those agreements, but it does not remove the ethical responsibility of the coach and the parties involved.
We should also design AI use deliberately: for preparation, rehearsal and organizing information we are permitted to use, not for outsourcing definitive interpretations of a client's behavior to a model.
My experience using AI has taught me that a fluent response without enough context can still mislead.
04
What should we continue?
We should continue listening, being present and remaining curious. But we should not assume these qualities belong exclusively to humans, or that they automatically produce better results.
What matters in my work is whether clients can challenge the coach's interpretation, redefine the question and retain ownership of their choices.
When coaching becomes advice disguised as questions, that ownership suffers. When a coach simply follows every turn without attending to the agreement and desired outcome, responsibility for the process weakens. Holding this balance is part of my professional practice.
AI is forcing us to define our value more precisely.
For me, the answer is not defending ‘humans against machines’. It is defending a conversation that takes the person's context seriously, makes responsibilities clear and leaves decisions with the person who must live with them.
Research sources
Questions to carry forward
- 01If strong questions are no longer distinctive, where does my professional value lie?
- 02Do I protect client ownership while still keeping the agreed process visible?
- 03Which uses of AI help without compromising confidentiality or delegating professional judgment?