If you felt a small wince the first time you heard the phrase “AI coaching,” you are in good company. Hebba Youssef, who writes the I Hate It Here newsletter for a large and famously skeptical HR audience, recently admitted to a “full-body shudder.” Her brain went straight to a chatbot asking how that really made you feel, then suggesting you sit with it. A robo-therapist in a trench coat pretending to be your thought partner.
We share the reaction. The term is cringe. It belongs on a buzzword-bingo board next to “we’re a family” and “great learning opportunity.”
What makes Hebba’s piece worth reading is what she did next. She is openly distrustful of anything claiming to fix people problems with a software subscription, and she still arrived at the most useful description of this category we have seen. One line in particular is worth keeping: “Every significant work problem is, at its core, a relationship problem.”
We have been building on that exact conviction since 2017, and the research supports it. When Google studied 180 of its own teams in Project Aristotle, the strongest predictor of high performance was not who was on the team but how the team worked together. It is rare to see the point argued this well, and rarer still by someone predisposed to roll their eyes at the whole idea.
Get the 2026 AI coaching playbook to see how organizations are implementing AI coaching at scale.
Why “AI coaching” earned the eye-roll: three failure modes behind the cringe
The honest reason “AI coaching” makes people wince is that the category launched before anyone agreed on what it meant. New tools arrive every week, and a lot of HR tech now wears “AI coaching” like a fresh coat of paint on a house with bad bones. Some of those tools are thoughtful. Many are not. When the definition is this murky, buyers either write the whole category off or get burned by something that promised transformation and delivered a glorified FAQ.
Hebba names three failure modes, and they are worth saying plainly because they are the real problem.
The first is the role-play problem. You open the app to practice a hard conversation with a direct report, and the pretend employee says, “You’re right, I can absolutely do better, thanks for the feedback.” No real person responds that way. Real people get defensive, push back, and need a minute to process. So you walk out prepared for a conversation that is never going to happen.
The second is the tool that waits. You used it once, it helped a little, and you never went back, because you do not have time to log in, re-explain your entire situation, and hope something useful comes out. The reminder emails go unopened. Coaching that waits for you to initiate it is coaching that does not happen.
The third is the blank slate. Most tools act like they know nothing about you, your team, or your job. They ask you to self-report and self-reflect before they will do anything, which is homework nobody has time for. The frustrating part is that the information to personalize already exists. It is sitting in behavioral assessments, in the HRIS, in calendars, in performance reviews, in competency frameworks. Nobody connects it.
Five questions that reveal whether an AI coaching tool will actually change behavior
Here is the part skeptics tend to get right. The failure modes above are not mysteries. They point directly at the design decisions that separate coaching that changes behavior from coaching that just generates usage metrics. You can use them as an evaluation checklist in your next vendor demo.
Does the coaching come to your people, or does it wait for them to come to it? The useful version shows up as a few sentences in the tools people already work in, like Slack, Teams, email, and calendar, timed to the moment they matter. The other 98.5 percent of someone’s interactions happen outside the programs HR runs. Coaching that cannot reach those moments stays contained to the calendar HR controls.
Is it triggered by what is actually happening in the organization? A manager just inherited a new team. A review flagged adaptability. A new report joined a recurring meeting. Those are the moments coaching is genuinely useful, and detecting them requires a connection to the systems of record, like Workday and other HRIS, that know when those moments happen. A tool that waits to be asked only reaches the manager who already knows they need help.
Is it built on validated behavioral data your organization has already invested in? There is a real difference between coaching that knows a person’s name and coaching that understands how that person processes information, hears feedback, and responds to stress. That understanding comes from established assessments like DISC, CliftonStrengths, and Enneagram. If a platform makes you abandon the assessments you already adopted and adopt a new proprietary one, it is adding friction and discarding a shared language you already built.
Is it connected to your own frameworks? Coaching grounded in your competency model and values closes the gap between a general insight and the specific situation a manager is actually in.
Does it measure behavior change, not logins? “Two thousand managers logged in last quarter” is not evidence. “Manager feedback conversations are measurably more specific than they were six months ago” is. The first counts activity. The second names what changed.
What good AI coaching looks like in a manager’s day
Strip away the buzzword and the version worth building is concrete. And it centers on the manager, for good reason. Gallup’s analysis of 2.7 million employees found that managers account for roughly 70 percent of the variance in team engagement. Reach the manager in the moment, and you reach the team.
It looks like a manager getting three sentences in Slack before a 1:1 with a direct report who hears feedback differently than they do. Here is how this person prefers to receive it. Here is what they need from you right now. Here is what to watch for. Not a ten-page report. Three sentences, before the conversation, when it can still make a difference.
It looks like a prompt the week before performance reviews that ties to your organization’s actual leadership competencies. It looks like the goal from that review surfacing again in the weeks that follow, in the flow of daily work, instead of sitting in a system someone opens twice a year. Awareness alone does not change behavior. Repetition tied to a real moment does.
And it looks like the kind of just-in-time support that used to be reserved for executives with expensive coaches, available to every person in the organization, not only the top 10 percent.
There is a metaphor we keep coming back to. Imagine a basketball coach who sits in the corner and waits for a player to come ask for help. That is not a coach. A real coach is watching the whole time, bringing context, going to the player before they know they need it, and seeing how each player works with the others. Not just on game day. Every day. Most AI coaching tools sit in the corner. The standard Hebba describes is the coach who shows up.
How Cloverleaf is built to meet that standard
We will say this part plainly, because the rest of the piece does not depend on it. Cloverleaf was built around this idea for eight years, before the category had a name. It synthesizes 13 licensed behavioral assessments into a single view of each person, signals how specific people are likely to work together, connects to your HRIS and your competency frameworks, and surfaces coaching in Slack, Teams, email, and Workday. No new login. Proven across 45,000 teams, with 65 million coaching moments and two approved patents behind it.
We have also made a deliberate choice that matters here. We have never given the AI a name or a persona, because it is not human and is not meant to feel like it is. Every output points the person toward another person. AI is good at speed, accuracy, and scale. Humans are the ones who handle the friction, the repair, and the trust. The goal is not to replace the relationships that make work meaningful. It is to help people walk into them better prepared.
What to do next: bring the standard to your next vendor demo
Hebba is right that the category has an image problem, and she is right not to soften her critique. The instinct that something feels off in most demos is worth trusting. But the thing underneath the bad name is real, and it is worth getting right.
So keep the standard she set. Bring it to your next demo and ask the vendor to show you, in the product, what the employee actually has to do to receive the coaching. The platforms that clear all of it are a short list. That is the version worth being excited about, cringe name and all.
Our thanks to Hebba Youssef for taking an honest look at where this category is headed. You can read her full piece, “The Term ‘AI Coaching’ Is Cringe, But Thankfully the Concept Isn’t,” in I Hate It Here.
