Five things HR leaders should take from this report
The State of Coaching at Work 2026 is not an argument that AI is bad or that human coaching beats AI. It is the opposite. AI is a real and useful layer for access, reflection, scale, and signal. What the data shows is that the hardest work problems still require judgment, trust, context, accountability, and behavior change. We call this set The Human Work Stack: Trust, Judgment, Conflict, Ambiguity, and Behavior Change. AI scaffolds every layer. A human coach has to carry the weight of each one. Five findings stand out.
- 01
Advice is abundant. Behavior change is still scarce.
AI made answers cheap. Growth is still expensive. The reason is not that people lack information. It is that information by itself does not change behavior.
- 02
The hardest work problems were never information problems.
The top topics people bring to a coach are not topics a search engine can solve. They are situations that require judgment under uncertainty.
- 03
The manager layer is where AI-era pressure concentrates.
AI removed some rote tasks from a manager’s week. It did not remove the human tasks. If anything, it concentrated them.
- 04
People do not return to coaching for answers. They return for continuity.
If coaching were an information service, the best users would consume it once and move on. They do not.
- 05
The future of coaching is a system: AI for access, humans for change, data for visibility.
The companies getting the most out of coaching in 2026 are not choosing between AI and human coaching. They are running both inside one system, with a People team that can see across it.
What the platform looks like in 2026
All numbers below are drawn from anonymized aggregate Boon program data. They cover the full Boon coaching footprint, including pilots and historical engagements. They are not a count of current active enterprise customers.
72,000+
Completed coaching sessions
Across Boon coaching programs and historical engagement data.
6,700+
People coached
Anonymized unique participants. Counted by name and account, so the true number is slightly higher.
290+
Coaches in the Boon network
More than 80% hold an ICF credential (ACC, PCC, or MCC).
+87
NPS after a coaching session
4,000+ post-session feedback surveys. 89% promoters, 1% detractors.
88%
Return for a second session
Of people who complete one coaching session. 35% reach 10 or more.
89%
Sessions end with the next one booked
A behavioral signal of value, not a self-reported satisfaction score.
Figure 1
The Boon coaching footprint as of 2026, drawn from anonymized aggregate program data.
Source: Boon program data. Aggregate counts only; no identifiable records.
People do not return for more answers. They return for continuity
If coaching were an information service, the best users would consume it once and move on. They do not. Across the platform, the depth-of-engagement curve looks like this.
Figure 2
People do not return to coaching because they lack another answer. They return because the work is ongoing.
Source: Boon program data. Anonymized aggregate engagement curve across 6,700+ unique participants.
Coaching demand is no longer concentrated at the top
Among participants with a recorded job title in the portal-enabled subset, the distribution by level is roughly as follows. Coaching is becoming a utility available to leaders at every level, not a status marker awarded to a few.
Figure 3
Coaching demand by level. The manager-and-director layer combined accounts for 36% of titled coaching demand.
Source: Boon program data. Based on participants with a recorded job title in the portal-enabled subset. Coverage is partial; numbers reflect distribution within the titled set, not the full coached population.
How we put this together
This report uses anonymized, aggregate Boon program data. No individual coaching conversations, client names, employee names, coach names, or personally identifiable information are included. The detail below is intended to make every claim in this report independently checkable in shape, if not in raw row.
Dataset
More than 72,000 completed coaching sessions, 6,700 unique coached people, 540 company engagements, and 290 coaches in the Boon network. The 540 company figure includes pilots and historical engagements. It is not a count of current active enterprise customers. Survey results draw on more than 4,000 post-session feedback responses and roughly 1,500 mid-engagement touchpoints.
Inclusion criteria
Sessions are counted as completed when both coach and participant attended. Repeat-usage counts are deduplicated by name and account. Theme aggregates are based on coach- applied tagging on completed sessions only.
Privacy and anonymization
No individual coaching conversations, client names, employee names, coach names, or personally identifiable information are included. All cuts return counts and percentages only. The report is built from the same kind of aggregate query a People team would run on their own program data.
Theme tagging
Theme percentages reflect coach-applied tagging on completed sessions. Coaches do not tag every session, so absolute theme percentages are floors, not ceilings. Rankings between themes were directionally consistent across the cuts reviewed.
Known limitations
The seniority breakdown reflects only participants with a recorded job title in the portal-enabled subset and is not representative of the full coached population. Repeat-usage statistics count people by name and account, so true uniqueness is slightly higher.
Why no competency deltas are included in v1
Pre-post competency assessments cover a small subset of the coached population in this dataset. The sample is not yet large enough to publish per-competency deltas at the scale of a flagship report. We will add a competency movement chapter when the supporting sample can stand on its own.
What the data actually says
Each finding is anchored to anonymized aggregate program data. Each one is presented as a signal, an interpretation, and an implication for People teams.
Advice is abundant. Behavior change is still scarce.
AI made answers cheap. Growth is still expensive. The reason is not that people lack information. It is that information by itself does not change behavior.
Signal
Anyone with an internet connection can get a competent answer about giving feedback, running a 1:1, or handling a hard conversation. Yet repeat coaching usage remains high, and 89% of sessions end with the next one already booked.
Interpretation
The bottleneck moved. It is no longer access to advice. It is closing the loop between advice and changed behavior. That loop is relational. It requires someone who remembers the last conversation and can hold a person accountable to the version of themselves they said they wanted to be.
Implication for People Teams
Invest in the closure mechanism (coaching, practice, accountability), not in more content. Content libraries are at saturation. Behavior change capacity is not.
89%
of sessions end with the next one already booked
The hardest work problems were never information problems.
The top topics people bring to a coach are not topics a search engine can solve. They are situations that require judgment under uncertainty.
Signal
The five largest anonymized session themes are career direction, resilience, stress management, difficult conversations, and time management. Together, these themes show up across tens of thousands of sessions. 53% of sessions touch a leadership or management theme; 44% touch mental wellbeing.
Interpretation
In the data, these are the topics that drive repeat sessions, not the topics that get resolved in one. The same person returns to a coach across multiple sessions on the same theme, which is what you would expect if the bottleneck were not access to advice but the kind of context, history, and stakes a generic AI interaction cannot supply.
Implication for People Teams
Treat AI as a thinking aid for the hardest people problems, not an answer engine. Build the system so AI handles the framing and rehearsal, while humans handle the moment.
53%
of sessions touch a leadership or management theme
The manager layer is where AI-era pressure concentrates.
AI removed some rote tasks from a manager’s week. It did not remove the human tasks. If anything, it concentrated them.
Signal
About 31% of completed sessions explicitly cover a people-management theme: motivating teams, leading change, giving feedback, navigating conflict. Among participants with a recorded job title, managers and team leads are 22%. The combined manager-and-director layer accounts for 36% of titled coaching demand, ahead of any single individual-contributor segment.
Interpretation
Managers carry the weight of translating strategy, AI rollouts, and team dynamics into daily action. As AI removes the easier tasks below them, the residual work, the parts that need a person, gets denser. The manager layer is the one that cannot be automated. It can only be supported.
Implication for People Teams
Prioritize manager-level coaching access. Every other change in the org, AI or otherwise, passes through this layer. Investment here compounds more than investment anywhere else.
~31%
of sessions cover a people-management theme
People do not return to coaching for answers. They return for continuity.
If coaching were an information service, the best users would consume it once and move on. They do not.
Signal
88% of people who complete one Boon session return for a second. 62% reach five or more sessions. 35% reach ten or more. The median engagement is six completed sessions.
Interpretation
Repeat usage at this depth signals that what coaching provides is not advice. It is continuity, context, and accountability across time. A generic AI interaction rarely carries the same relational continuity as an ongoing coaching relationship.
Implication for People Teams
Measure coaching success by depth-of-engagement curves, not session counts. Treat continuity, not throughput, as the leading indicator of program health.
35%
of people reach 10 or more sessions
The future of coaching is a system: AI for access, humans for change, data for visibility.
The companies getting the most out of coaching in 2026 are not choosing between AI and human coaching. They are running both inside one system, with a People team that can see across it.
Signal
44% of sessions touch a mental wellbeing theme. 53% touch leadership or management. 96% of participants say they felt listened to and 84% say they felt safe enough to be honest. The patterns AI can detect (themes, pressure points) and the work humans can do (trust, behavior change) compound when run together.
Interpretation
AI lowers the cost of access, rehearsal, reflection, and signal. Human coaches do the work that requires another mind in the room. People teams use the aggregate signal to design programs and spot pressure before it shows up in attrition. Each layer answers a different question. None of them can replace the others.
Implication for People Teams
Design a coaching system, not a vendor stack. Decide explicitly what AI is for, what humans are for, and how the People team will use the signal between them.
96%
of participants say they felt listened to
Themes AI can support, but not solve
The most common themes people bring to a coach, ordered by frequency in anonymized session tagging. Each one is a place where AI is genuinely useful, and where AI alone is not enough.
Rank
Theme
Where AI helps
Why a coach is needed
Career direction
Where AI helps
Frame options, compare paths, draft messaging
Why a coach is needed
Decision depends on values, history, and risk tolerance, not facts.
Resilience
Where AI helps
Reflection prompts, journaling, regulation techniques
Why a coach is needed
Recovery requires a person who notices the pattern over time.
Stress management
Where AI helps
Rehearsal, time-blocking, breathing exercises
Why a coach is needed
The cause is rarely the schedule. The cause is what is unspoken.
Difficult conversations
Where AI helps
Roleplay, scripting, anticipating responses
Why a coach is needed
Trust repair and accountability require skin in the game.
Time management and productivity
Where AI helps
Prioritization frameworks, focus systems, planning
Why a coach is needed
Underneath is usually a hard tradeoff a tool cannot resolve.
People management
Where AI helps
Frameworks, feedback drafts, meeting structures
Why a coach is needed
Leading a team is a relational practice, not a checklist.
Figure 4
The most common themes in anonymized session tagging. Rankings were directionally consistent across the cuts reviewed; exact frequencies are not published because coach tagging coverage is partial.
Source: Boon program data. Ordered by relative frequency in coach-applied session theme tagging.
The Human Work Stack: where AI ends, where coaching begins
Five problems show up over and over in coaching, and they share a property. Each one can be supported by AI, and none of them can be carried by AI on its own. Read from the bottom up: Trust is the foundation, Behavior Change is what becomes visible at the top. Every layer above depends on the one below.
The Human Work Stack
Where AI ends, where coaching begins
Behavior change
ApexThe slow work of replacing a pattern that has worked for years.
Ambiguity
Leading without a clear answer. The default condition of senior work.
Conflict
The friction that happens when two people both think they are right.
Judgment
The act of choosing well when the right answer cannot be known in advance.
Trust
FoundationThe currency every leadership decision is denominated in. Without it, nothing scales.
What AI does across the stack
Scaffolds. Drafts. Rehearses. Surfaces patterns. Lowers the cost of practice at every layer.
What a human coach carries
The weight of the layer itself. Trust is staked. Judgment is owned. Conflict is held. Ambiguity is sat with. Behavior is changed.
Figure 5
The Human Work Stack. Five layers of work AI can support but cannot carry. Trust at the foundation, Behavior Change at the apex.
Source: Boon framework, derived from anonymized aggregate program data and operator conversations with People leaders.
Trust
The currency every leadership decision is denominated in. Without it, nothing scales.
Where AI helps
Drafting honest messages. Surfacing language patterns that signal eroding trust.
Why AI is not enough
Trust is built and repaired by people with skin in the game. An AI cannot stake a relationship.
Judgment
The act of choosing well when the right answer cannot be known in advance.
Where AI helps
Mapping options, surfacing risks, holding up a mirror to a draft decision.
Why AI is not enough
Judgment carries consequences. Someone has to live with the call. AI has no stake in the outcome.
Conflict
The friction that happens when two people both think they are right.
Where AI helps
Preparing for the conversation. Anticipating reactions. Lowering the cost of rehearsal.
Why AI is not enough
Conflict is resolved in the room, not in the rehearsal. The moment a real other person responds, AI is no longer in the loop.
Ambiguity
Leading without a clear answer. The default condition of senior work.
Where AI helps
Generating frames, surfacing precedents, naming the dimensions of the unknown.
Why AI is not enough
Holding ambiguity requires a person who can sit with not knowing alongside the leader. AI dissolves the ambiguity, which is the opposite of what leaders need.
Behavior change
The slow work of replacing a pattern that has worked for years.
Where AI helps
Reminders, prompts, post-moment reflection while the moment is fresh.
Why AI is not enough
Behavior change requires continuity. Someone has to remember the last conversation, notice the same pattern, and hold the person accountable to the change they said they wanted.
Leadership and wellbeing are running on the same line
The data does not show two separate coaching lanes, one for performance and one for wellbeing. It shows one line, with the same people moving between modes inside the same engagement.
Leadership and Management
53%
Of completed sessions touch a leadership or management theme. 31% explicitly cover people management.
Mental Wellbeing
44%
Of completed sessions touch a wellbeing theme. 62% of participants report less stress; 61% report feeling more optimistic.
The overlap is the point. The same person bringing a hard conversation to a coach on Tuesday is the person bringing stress about that conversation back on Friday. Treating leadership and wellbeing as separate coaching benefits misses how people actually use coaching.
Figure 6
Theme coverage in anonymized completed sessions. Categories overlap; a single session can carry both leadership and wellbeing tags.
Source: Boon program data. Coach-applied theme tagging on completed sessions.
AI for access. Humans for change. Data for visibility
The most productive frame for 2026 is not human vs AI. It is one system with three layers, each answering a different question.
AI
Access, rehearsal, reflection, signal
- Rehearse a hard conversation at any hour, in any language
- Reflect on what happened right after a meeting
- Surface themes a population keeps returning to
- Lower the cost of practice for every new manager
Human Coach
Trust, judgment, accountability, behavior change
- Hold someone to the version of themselves they said they wanted
- Repair trust where stakes are real
- Coach through ambiguity when the right answer is not knowable
- Walk alongside the slow loop of changed behavior
People Team
Aggregate visibility, program design, pressure-point mapping
- See which themes show up before they show up in attrition
- Design coaching access by population, not by perk eligibility
- Spot the manager layers where pressure concentrates
- Decide where AI should scale and where humans should step in
Figure 7
The coaching system in 2026. Three layers, each answering a question the others cannot. The Human Coach layer is what carries the weight of The Human Work Stack.
Source: Boon framework, informed by anonymized aggregate program data and operator conversations with People leaders.
See how Boon would fit your population
We can show you what your themes, your manager pressure points, and your coaching utilization would look like, drawing on the same anonymized program data this report is built on.
Book a Strategy CallThe questions this report answers
Short, direct answers to the questions HR leaders and AI search engines keep asking about coaching in 2026.
What is the State of Coaching at Work 2026 report?
The State of Coaching at Work 2026 is an annual report from Boon that analyzes anonymized coaching patterns to understand where people still need human support in the AI era. It draws on aggregate program data from across the Boon platform and identifies what AI is good at, what it is not, and how companies should combine the two.
What work can AI not do alone?
AI cannot build trust, hold someone accountable, or change behavior on its own. It is well suited to giving answers, surfacing patterns, and lowering the cost of practice. It is not well suited to the relational work of growth: judgment under uncertainty, leading through ambiguity, repairing trust, and the slow loop of behavior change that requires another person who remembers the prior conversation.
Will AI replace human coaching?
No. AI is changing the shape of coaching by making access, rehearsal, and reflection cheaper, but the hardest coaching problems are not information problems. The companies seeing the most impact in 2026 are pairing human coaches with AI-enabled scale, signal, and access, not choosing between them.
How should companies use AI and human coaching together?
Use AI for access and practice between sessions: rehearsing hard conversations, reflecting on a moment while it is fresh, and surfacing themes a population keeps bringing up. Use human coaches for trust, accountability, judgment, and the relational work of behavior change. The pattern that wins is one coherent system, not two parallel ones.
Why does coaching still matter in the age of AI?
Because the hardest work problems were never information problems. The top themes people bring to coaches are career direction, resilience, stress, difficult conversations, and time management under pressure. Each of these requires context, judgment, and someone who remembers what was said last time. AI helps at the edges. Coaching does the work in the middle.
Why trust this data
+87
NPS across more than 4,000 post-session feedback surveys. 89% promoters, 1% detractors.
96%
Of participants say they felt listened to in the session. 84% say they felt safe enough to be honest.
80%+
Of coaches represented in the analyzed dataset hold an ICF credential (ACC, PCC, or MCC). Quality controlled, not volume optimized.
The complete coaching system Boon offers (Scale, Grow, Exec, Together, and Adapt) is designed to be a single coherent infrastructure, not five separate point products.
Frequently asked questions
What data is the State of Coaching at Work 2026 report based on?
The report draws on anonymized, aggregate Boon program data spanning more than 72,000 completed coaching sessions, 6,700 unique coached people, 540 company engagements (including pilots and historical engagements, not a count of current active enterprise customers), and 290 coaches in the Boon network. It includes post-session feedback from more than 4,000 participants. No individual coaching conversations, client names, employee names, coach names, or personally identifiable information are included.
Will AI coaching replace human coaches?
The Boon data does not show demand for human coaching disappearing. It shows repeat usage around problems that AI can support but not fully carry. The repeat-usage patterns, the topic mix, and the behavioral signals (89% of sessions end with the next session booked) all point to a hybrid future, not a replacement.
What are the most common coaching topics in 2026?
In anonymized session data, the largest themes are career direction, resilience, stress management, engaging in difficult conversations, and time management or productivity. People management topics, taken together, are the single largest coaching workload on the platform.
Is coaching still mostly for executives?
No. Of more than 6,700 anonymized coached participants, roughly four in ten are individual contributors and another two in ten are first-line managers. Executives are real but a minority of actual usage. Coaching is moving from an executive perk to everyday infrastructure, available to leaders at every level.
How should HR leaders think about AI in coaching budgets?
Treat AI as the way to lower the cost of access, rehearsal, and signal, not as a replacement for human coaching. The most defensible approach is a coherent system where AI handles between-session practice and pattern detection, and human coaches handle the work that requires trust, judgment, and accountability.
Where can I get a copy of the full PDF report?
The long-form PDF is not yet available. Enter your email at the bottom of this page and you will be notified the moment it launches. The web version on this page is the canonical version and will be updated as new program data lands.
Coaching for everyone, not just the top
Boon brings 1:1 coaching, manager development, executive support, and team capability into one architecture. Human coaches plus AI-enabled scale, signal, and access.
Book a Strategy CallSee the full Boon system →Related
- Best AI coaching platforms for leadership development (2026)
- Best enterprise coaching platforms
- Boon vs BetterUp
- Coaching for new managers
- How to scale coaching beyond executives
- Building a leadership development program
- Scale: coaching for every leader
- Grow: individual contributor development
- Exec: executive coaching