AI coaching tools are everywhere, promising scale, savings and 24/7 support. But should HR really be buying them, or sticking with human coaches? As organisations plan for 2026, the real question is how to blend both wisely. Jonathan Passmore explores what the evidence says, and what HR should do next
Coaching is now a mainstream development tool in many organisations. At the same time, AI coaching agents are moving from novelty experiments to products available off the shelf, like Microsoft 365. For HR, the commissioning question for 2026 is no longer: “should we use coaching?” It is: “what mix of human coaching and AI-enabled coaching support best fits our workforce, our risks and our budget?” This article summarises the latest research and offers practical guidance for HR leaders.
“AI coaching is no longer a gimmick, but nor is it yet a full substitute for high-quality human coaching”
What AI coaching is genuinely good (and where it falls short)
The past year has seen an explosion of experimentation with AI and, unsurprisingly, a multitude of evaluation studies, as organisations road-tested trials of their AI tools. What lessons can HR professionals draw from these trials and how should they be using AI and human coaching?
The short answer is that AI offers some benefits. The studies, mainly with small groups of users, suggest users are willing to experiment with it and found it valuable for some tasks. While on the surface AI tools can be designed to replicate human coach competencies, the feedback from users suggests they can experience these tools as shallow, lacking empathy and trustworthiness when compared to a human coach. What we describe as the three S’s of AI coaching: short, shallow and sycophantic.
Users in studies have also questioned their ability to undertake complex tasks, to be sufficiently culturally sensitive, or 100% accurate. One of the biggest challenges is predicting a tool’s ability on a new task, what’s become known as ‘the jagged edge’ of AI technology.
For HR, this points to a need for care, while recognising many on the board will be presenting AI rollout as a way to save money on learning and development, and with an eye on the medium-term savings that might be gained from wider efficiencies.
But AI coaching should be commissioned as a ‘pilot’, not as a ‘prestige’ intervention. It can provide consistent access to structured reflection between formal learning events, it can help managers who need a private thinking space at short notice, such as at 3am when their human coach is not available, but question marks remain.
The evidence so far suggests that AI coaching should not be procured as a generic replacement for all coaching. Instead, it’s better positioned as an additional channel, best suited to specific kinds of needs, with escalation routes to human support when the work becomes complex, relational or high stakes.
Further care is needed in deployment to train users on both its benefits and its risks, helping users to display discernment in applying the data generated. Secondly, supporting engagement to prevent a user-non-user divide opening, as take-up and sustained usage has proven patchy when the tool is not integrated into workflows or employee concerns are not explored and resolved.
A practical way to match needs to the right coaching channel
The table below integrates what these studies imply about when AI coaching is likely to add value, when human coaching is likely to be preferable and what HR needs to put in place to deploy AI coaching effectively.
When to use AI, when to use human coaching?
| Coaching need in the organisation | Where AI coaching is likely to add value | Where a human coach is likely to add value | What HR should build into deployment |
| Short, structured development tasks within the flow of work | Consistent structure, goal clarity, reflection, action planning, easy access and convenience | Added value if the task involves ambiguity, politicsand competing stakeholder pressures | Set expectations that the tool supports structured reflection and action, not expert advice |
| Emotionally rich,complex, orsensitive coaching topics | Limited fit because relational empathy and psychological safety limited – and can create pushback | Deeper alliance, nuanced empathy, lived experience, use of silence, metaphorand humour | Provide clear guidance on what topics are suitable and offer a straightforward route to human coaching |
| Scaling access across large employee populations | Low cost and high availability, potential to widen access beyond leader-manager tiers | High impact for populations with high stakes roles or complex change agendas | Commission pilots, then scale based on outcomes and adoption data rather than vendor claims |
How to deploy AI coaching so it is used, trusted and useful
The strongest ‘how to’ message across these studies is that AI coaching adoption is a system issue, not a software issue. Users raised concerns about data privacy, bias and transparency, and these concerns shape trust and sustained engagement. If employees are unsure who can see what they type, they are less likely to actively engage with the tool.
This means HR commissioning in 2026 should treat governance and communication as core requirements. Contracts need clarity on what data is collected, how it is stored, who can access it and how confidentiality is explained to users in plain language. You should also build in transparency about what the AI coach is and is not.
Quality assurance should also be upgraded. If a vendor cannot show you how their tool is tested for coaching behaviour, ethical practice and consistency, it is hard to treat the product as ‘coaching’ rather than a conversational support tool, which offers short reflective conversations followed by generic advice.
Finally, set realistic success criteria. HR should avoid treating AI coaching as a settled intervention with guaranteed outcomes. A sensible 2026 approach is to pilot the tool with a bounded population, setclear goals, then evaluate what matters in your context. If your goal is better day-to-day self-management and faster reflection cycles, adoption and user-reported usefulness may be a suitable first indicator. If your goal is deeper leadership shifts, sustained motivation and complex behavioural change, it’s likely that human coaching still holds a clear advantage on several core dimensions, particularly alliance, insight and commitment.
Human or AI coaching? The balanced answer
The overall message for 2026 is balanced. AI coaching is no longer a gimmick, but nor is it yet a full substitute for high-quality human coaching. Its organisational value comes from thoughtful task matching, credible quality assurance and careful deployment that earns trust.
Jonathan Passmore is Professor of Coaching at Henley Business School and a public speaker
