Marcus Webb runs talent acquisition enablement at Northbridge Mutual, a 600-person fintech with a 9-person recruiting team. Three months ago he stood in front of that team and rolled out an AI screening and outreach assistant with a polished half-day workshop. Everyone nodded. Everyone logged in. Then he watched the usage dashboard flatten. By week four, four of his nine recruiters had quietly stopped opening the tool, and one of his most experienced sourcers told him over coffee, "I tried it twice, it gave me something weird, and I just went back to doing it the old way." Marcus had not built a bad tool or a bad training. He had assumed that a training day was the finish line, when it was actually the starting gun. What sustained adoption was never the workshop. It was the slow, human work of coaching, support, and confidence that comes after the slides are closed.

Why Training Builds Knowledge but Coaching Builds Confidence

A training class transfers information. People learn what a tool does, where the buttons are, and what the workflow looks like in a controlled demonstration. But knowledge is not the same as confidence, and confidence is what determines whether someone reaches for a tool on a busy Tuesday when they have six screens to run and a hiring manager breathing down their neck. The gap between "I understand how this works" and "I trust myself to use this under pressure" is exactly where most adoption dies.

Marcus learned this the hard way. His nine recruiters had all sat through the same workshop, yet a month later they had split into three groups: three confident daily users, two occasional users who reached for the tool only when work was slow, and four who had drifted away entirely. The training had been identical. The outcomes were not, because the thing that separated the confident from the lapsed was not knowledge. It was whether anyone had been there in the messy middle, when the tool produced a strange result and the recruiter had to decide whether to push through or quit. Coaching and support are what carry people through that middle.

Peer Coaching: Champions Who Carry the Adoption

The single highest-leverage move Marcus made was to stop being the only source of help. He identified two peer champions, recruiters who had adopted early, were respected by their colleagues, and had the patience to teach without making anyone feel slow. Peer coaching is more credible than manager-directed training for a simple reason: when a fellow recruiter says "I had the same problem and here is what worked," it lands differently than the same advice from the person who mandated the rollout. Champions answer questions in real time, in the flow of real work, and they model what good use actually looks like.

Not every early adopter makes a good champion. Marcus looked for five traits: early and genuine adoption of the tool, respect from peers, patience and a teaching instinct, willingness to help without resentment, and a real understanding of both the technology and the craft of recruiting. His fastest power user was actually a poor fit, because she lost patience explaining things twice. The two he chose were merely strong users, but they could explain a concept three different ways until it clicked and leave the other person feeling supported rather than judged. What you want is not the person who learned fastest; it is the person who can make someone else's learning feel safe.

The mistake most organizations make is treating champion work as a favor squeezed in around a full workload. Marcus refused to do that. He developed the coaching role formally, with four components: a clear charter naming what the role actually is, training on how to coach, meaning how to teach, give feedback, and handle resistance; time allocated by reducing other work rather than assuming coaching fits in the gaps; and recognition that makes the contribution visible. Concretely, he allocated 20 percent of each champion's week to coaching during the first quarter of the rollout, which on a 40-hour week is one full day, then planned to taper it once the team stabilized. Twenty to thirty percent is the working range depending on organization size. With nine recruiters and two champions, that gave him a ratio of about one champion for every four to five recruiters, close enough for each coach to know their people personally without drowning in requests.

One-on-One Support for the People Who Will Not Ask

Group training and office hours work for the people who are already comfortable. The recruiters Marcus was most worried about were the ones who would never raise their hand. His experienced sourcer who "went back to the old way" was not lazy or resistant. He was anxious, slightly embarrassed, and convinced everyone else had figured it out. People like this rarely ask for help. They struggle quietly and then disappear from the dashboard, and some of them learn best individually anyway, which a group workshop can never accommodate.

So Marcus made support proactive instead of reactive. He used the usage data not as a scoreboard but as an early-warning system. Any recruiter whose tool usage dropped below a handful of sessions a week in the first month got a low-key personal outreach, not a reprimand but an offer: "Want to grab 30 minutes and walk through a couple of your real reqs together?" He paired each at-risk recruiter with a champion for individual sessions built around four things: basic tool functionality, how the tool connected to their actual recruiting process, working through realistic scenarios using their own open roles rather than a generic demo, and addressing whatever specific concern that person was carrying.

The cadence mattered as much as the content. After the initial workshop, Marcus did not vanish. He built a deliberate follow-up rhythm: a check-in at the end of week one, again at week two, and again at week four, each one a short and genuine "How are you doing with the tool, and what questions do you have?" That ongoing contact improves adoption and confidence far more than the training itself does. One-on-one support is expensive in time, and Marcus could not give every recruiter this level of attention, which is exactly why he reserved it for the three or four people most likely to slip away. That targeted investment recovered two of his four lapsed users within six weeks, including the sourcer, who eventually became one of the team's steadiest users.

Feedback Loops That People Trust Enough to Use

Coaching only works if Marcus could hear what was actually going wrong, and people only share what is going wrong when they believe it will lead somewhere. He collected feedback systematically through four channels, because different people surface concerns in different ways: regular office hours where a champion was reliably present, a short feedback survey after the first month of use, a dedicated Slack channel for quick questions, and one-on-one interviews with selected users, especially the recruiters least likely to post in public. Some people will type a question into a channel at midnight. Others will only say it out loud to one trusted person. He wanted both, and the questions he asked were deliberately plain: what is confusing about the tool, how could we make it easier to use, what are you struggling with?

The part that made the difference was closing the loop. Feedback has to be actionable, and there are three distinct responses depending on what the feedback reveals. If the tool itself is the problem, fix or reconfigure the tool. If the issue is understanding, clarify the training. If the workflow around the tool is wrong, change the process. When three recruiters separately reported that the outreach assistant kept producing messages that sounded too generic, Marcus did not file the comment away. He worked with a champion to adjust the team's shared prompt template, rolled out the fix, and then went back to those three people: "You flagged that the outreach felt generic. We reworked the template, here is the new version, thank you for catching it." That small act of reporting back is what keeps feedback alive. Once people see that their input quietly disappears, they stop offering it, and the coach loses the ability to see problems before they spread.

Psychological Safety: Making It Normal to Not Know Yet

None of this works in a culture where asking a question feels like admitting failure. Psychological safety, the shared belief that you will not be embarrassed or punished for speaking up, is the soil that coaching grows in. It is also fragile. A single comment like "you should know this by now" or "everyone else got it" teaches the whole room a lasting lesson: keep your struggles hidden. Marcus had to actively protect against that, because his team had a few high performers whose offhand impatience could shut down a less confident colleague in one sentence. Nobody who is slower to adopt should be shamed for it.

He worked the small moments. When someone asked a basic question in the Slack channel, the champions were coached to respond with "great question, I wondered the same thing" rather than a terse answer that implied the asker was slow. He named the team norm out loud: in the first ninety days, nobody is expected to be fluent, and questions are a sign the system is working, not failing. Most importantly, he modeled fallibility himself. In a team meeting he openly described how he had fumbled a feature for two weeks until a champion showed him a shortcut. When the person who ran the rollout admits he struggled, it gives everyone else permission to struggle too. Leaders set the tone whether they intend to or not. If they project effortless mastery, people hide their confusion. If they show the work, people show theirs.

Recognizing Progress to Sustain the Effort

Confidence compounds when progress is seen. Marcus made recognition specific and public rather than vague and private. "Good job everyone" motivates no one. "Dana ran her first 50 screens with the assistant this month and has been helping Priya get set up, thank you Dana" does two things at once: it rewards Dana, and it shows the rest of the team what success concretely looks like. Public recognition motivates the person named and models positive adoption for everyone watching.

He celebrated at three levels so that everyone could see themselves somewhere on the ladder. At the individual level, he marked personal milestones, of the form "you completed your first 50 screenings with the tool." At the team level, he shared real movement, of the form "our team has improved screening time by 20 percent in the last month." At the organizational level, once the function crossed a steady majority of active users, he named that milestone the same way, "we have reached 85 percent adoption across recruiting, well done." The point was never empty cheerleading. Recognition tied to a concrete contribution tells people their effort produced something real, and that is what convinces them to keep going through the awkward early weeks.

Putting It Together: Marcus's Quarter

By the end of the quarter, the picture had changed. Marcus had started with three confident users out of nine and a dashboard that was sliding the wrong direction. Two champions with protected coaching time, proactive one-on-one support for his four at-risk recruiters, feedback loops he visibly acted on, a deliberately safe culture for questions, and steady specific recognition turned that around. Seven of his nine recruiters were active weekly users, and the two who remained occasional users were at least still in the system rather than gone. The tool had not changed. The support around it had. Coaching did not make the technology better. It made the people confident enough to keep using it long enough for the technology to prove its worth.

Keeping Coaching Alive After the First Quarter

The threat to a coaching program in month six is not resistance. It is erosion. Coaching requests keep arriving after the protected time has been quietly reabsorbed into the champion's req load, the office hour survives on the calendar but nobody attends, and the recognition that made the role feel worthwhile stops being said out loud. Marcus planned the taper deliberately rather than letting it happen by drift: 20 percent of a champion's week during the first quarter, then a reduced allocation once the team stabilized, with the reduction announced as a decision rather than discovered as an absence. A commitment that quietly shrinks reads to a champion as a commitment that was never real.

The other half of sustainability is who gets coached next. Every new recruiter who joins after the rollout inherits the same conditions the original team faced without the same support, unless someone makes onboarding into the tool a standing part of the role rather than a launch-period exception. Marcus assigns each new hire a champion and the same four-part one-on-one path his at-risk recruiters received, which costs a few hours and prevents the slow re-accumulation of exactly the group he spent a quarter recovering. Coaching that only exists during a rollout produces a team that is confident about one tool at one moment in time.

Three Anti-Patterns That Quietly Kill Coaching

The first anti-pattern is coaching without structure. An organization assigns coaching informally, with no clear role, no training, and no time allocation, usually because naming a champion feels like the whole intervention. It fails because coaching then becomes an additional burden on already-busy people; champions burn out, coaching turns inconsistent or stops, and people fall back to struggling alone. The failure has a recognizable arc: you identify a peer champion, they help colleagues on the side of their regular work, it goes well at first, requests increase, the champion is stretched thin, they start avoiding questions, and coaching ends without anyone announcing it. Avoid it by formalizing the role. Define expectations, time allocation, and training explicitly. Champions should have 20 to 30 percent of their time for coaching depending on organization size. This is real work, not extra work. Invest in coaching training, recognize the contribution, and make the arrangement sustainable rather than heroic.

The second anti-pattern is one-on-one support only for the people who ask. Organizations make coaches and office hours available, then serve whoever shows up, which feels fair and is not. It fails because at-risk adopters usually do not ask; they struggle quietly, and without proactive support they eventually give up, and the organization loses capability it already paid to build. The pattern in practice: you have coaches and office hours, but the people who most need help are the least likely to come. They feel embarrassed, they assume they should be able to figure it out, they struggle alone, and they stop using the tool or stop trying to improve. Avoid it by being proactive. After training, identify who is at risk, whether from low usage, visible frustration, or a slower learning curve, and reach out first: "How are you doing? Can I help?" That removes the barrier of having to ask and signals that support is genuinely available.

The third anti-pattern is feedback ignored. Organizations collect feedback from coaching and support interactions and never act on it, often because collecting is easy and acting requires a decision and a budget. It fails because feedback that goes nowhere stops arriving; the coaching relationship deteriorates, and problems that were cheap to fix compound instead. The shape of it: coaches report that people are confused about a particular feature, nothing changes, and months later people are still confused about the same feature. The coaches notice that their reports vanish and stop filing them, so the problems that could have been fixed simply are not. Avoid it by closing the loop every time. Acknowledge the feedback, act on it where you can, and report back: "Three people reported confusion about this feature. We clarified the training. Thank you for the feedback." That single sentence is what tells the team the channel is real.

Practice

Each of these produces a document you could hand to a champion or a manager, which is the point. Coaching that is not designed tends not to happen.

• Peer coaching program design. Design a comprehensive peer coaching program for your organization. Identify potential coaches: who would make good ones, what are their strengths, and how would you train them? Define the coaching role, including responsibilities, time allocated, and how coaches are recognized. Design the coaching interactions themselves, including how coaches prepare and how quality is ensured. Create a coaching manual covering principles, teaching techniques, common questions, and the tools to use. Then plan for sustainability: how will you avoid coach burnout, and how will you keep coaching vital and valued a year in?
• One-on-one support program. Design a proactive support program for at-risk adopters. How will you identify at-risk people, whether by low tool usage, expressed frustration, or slower learning? What support will you offer, from baseline retraining to personalized coaching to flexibility on deadlines? How will you reach out, through proactive contact from a coach or manager, invitations to office hours, or personal offers? What is the timeline, meaning when help is offered, for how long, and how many follow-ups? And how will you measure success, through adoption rates, confidence surveys, or performance metrics?
• Feedback system design. Create a system for collecting, acting on, and closing the loop on coaching feedback. Which channels will you use across surveys, interviews, forums, and office hours? How will feedback be captured and organized, whether in a centralized list, by category, or with tracking? How will you decide what justifies action, and who decides? What actions are available to you, from training clarification to tool modification to process change to a communication? And how will you close the loop and update people on what happened?
• Psychological safety audit. Assess your current state and design improvements. Does your team feel safe asking questions, and what is your evidence, whether interviews or surveys? Identify the barriers: what makes people afraid to ask for help, and what experiences shaped that? Design leadership actions such as modeling vulnerability, celebrating questions, and responding to mistakes with learning. Design team practices such as no-blame post-mortems, peer support, and shared learning. Then plan measurement: how will you know safety has actually improved?
• Recognition strategy. Design a strategy for recognizing progress and successful adoption. Which milestones are worth celebrating, from first tool use to proficiency to helping others to genuine innovation? Who should be recognized, across individual contributors, teams, managers, and coaches? How should recognition happen, whether through public celebration, rewards, advancement, or peer recognition? How will you make it meaningful, meaning specific rather than generic and connected to the value created? And how will you sustain it on a regular cadence without it becoming rote?

Reflection

These questions are worth answering before your next rollout, because the honest answers usually name the gap.

• Who on your team is most likely to struggle quietly with a new tool, and how would you currently find out?
• What is your biggest obstacle to making coaching real work rather than a favor, and who would have to agree to remove it?
• What is the last piece of feedback your team gave you about a tool, and what did you visibly do about it?
• When did you last admit publicly that you struggled with something, and what did that change in the room?
• How will you know your coaching program worked, in a measure that "everyone was trained" cannot satisfy?

Glossary

• Peer champion. A respected colleague who adopts early and coaches others. More credible than manager-directed training, and effective only when the role is formalized with a charter, training, time, and recognition.
• At-risk adopter. Someone likely to abandon a tool without proactive support, usually recognizable by falling usage or expressed frustration rather than by a request for help.
• Psychological safety. The shared belief that speaking up will not lead to embarrassment or punishment. Fragile, destroyed by a single dismissive comment, and the precondition for anyone asking a question.
• Closing the loop. Acknowledging feedback, acting on it, and reporting back what changed. The step that determines whether a feedback channel keeps producing anything.
• Organizational capacity. The resources, capability, and attention available to execute initiatives. High-capacity organizations can run several at once; low-capacity ones must sequence them, which is why coaching time has to be taken from somewhere explicit.
• Adaptive capacity. An organization's ability to learn, change, and improve in response to new information. High adaptive capacity is what turns coaching feedback into actual changes rather than filed comments.

Related Lessons

Coaching connects to several neighboring lessons that supply what it depends on.

• Change Leadership: Driving Adoption While Managing Resistance covers the resistance mapping and messaging that determine what your coaches will be asked about in the first place.
• Adult Learning Principles: How Recruiters Learn and Adopt New Tools explains why practice and support outperform lectures, which is the assumption underneath this entire lesson.
• Creating Communities of Practice: Learning Networks takes the peer coaching model beyond a pair of champions into something that sustains itself as the team grows.
• Measuring Adoption: Tracking Usage, Proficiency, and Impact gives you the metrics that turn a usage dashboard into an early-warning system rather than a scoreboard.
• Training Program Design: Curriculum, Delivery, and Evaluation handles the workshop half of the problem, which coaching is meant to extend rather than replace.

Closing

The through-line of Marcus's quarter is that nothing he did required a better tool or a bigger budget. He named two champions and gave them real time. He reached out to the people who would never reach out to him. He asked what was confusing, fixed what he could, and told people what he had fixed. He made it normal not to know yet, and he said out loud who was making progress. Each of those moves is small, and together they moved his team from three confident users to seven.

It is worth remembering what the work is ultimately for. Leading responsible AI adoption in recruiting shapes how people are evaluated for opportunity, which means the confidence you build in a recruiter is not just a productivity outcome. A recruiter who understands the tool well enough to question it is the safeguard the whole process depends on, and a recruiter who quietly gave up is a safeguard you no longer have. You have real power over how hiring decisions get made. Use it carefully, and invest in the people who exercise it.

Key Takeaways

• Training is the starting gun, not the finish line. A workshop transfers knowledge, but confidence under real pressure is what sustains adoption. The work that matters most happens in the weeks after the slides close, when someone hits a strange result and has to decide whether to push through or quit.
• Peer champions outperform manager-directed help. Select for five traits: early adoption, peer respect, patience and teaching ability, willingness to help, and understanding of both the technology and recruiting. The fastest power user is often the wrong choice.
• Formalize the coaching role or it collapses. Champions need a clear charter, training on how to coach, allocated time of roughly 20 to 30 percent depending on organization size, and visible recognition. Without those four, coaching becomes uncompensated burnout and quietly stops.
• Be proactive with the people who will not ask. At-risk adopters struggle quietly and then vanish from the dashboard. Use usage data as an early-warning signal, reach out with an offer rather than a reprimand, and keep a deliberate follow-up cadence at weeks one, two, and four.
• Feedback only survives if you close the loop. Collect it systematically through office hours, surveys, a chat channel, and one-on-one interviews, then act, distinguishing tool problems from training problems from process problems, and report back what changed.
• Psychological safety is the soil coaching grows in. It is destroyed by a single dismissive comment and built by welcoming questions, naming the norm that nobody is expected to be fluent yet, refusing to shame slower adopters, and leaders openly modeling their own struggles.
• Recognize progress specifically and publicly. Name the person, the milestone, and the contribution at individual, team, and organizational levels. Recognition tied to real value is what convinces people to keep going through the awkward early weeks.
• Capability building is core, not optional. Technology adoption depends on developing the team, so training, coaching, and communities of practice are the investment that determines whether the tool produces value at all.

Frequently Asked Questions

How much of a champion's time should coaching actually take? Plan for 20 to 30 percent depending on your organization's size, and take that time from their other work rather than assuming it fits in the gaps. Marcus allocated 20 percent, roughly one day a week, during the first quarter of the rollout and planned to taper it as the team stabilized. The number matters less than the principle behind it: coaching is real work. Champions absorbing it on top of a full desk will start avoiding questions within a few months, and the program ends without anyone deciding to end it.

What if my best user is impatient with people who are slower? Then they are not your champion, whatever their usage numbers say. Marcus's fastest power user lost patience explaining things twice, so he chose two merely strong users who could explain a concept three different ways and leave the other person feeling supported. Select for patience, teaching instinct, and peer respect over raw proficiency. A brilliant coach who makes people feel slow will suppress exactly the questions your feedback loop depends on.

How do I find the people who are struggling if they never say anything? Use your usage data as an early-warning system rather than a scoreboard. Marcus watched for anyone whose sessions dropped below a handful a week in the first month and reached out personally with an offer rather than a reprimand. Frustration expressed in passing and a visibly slower learning curve are the other two signals. The point is that the outreach comes from you, because the barrier for an embarrassed adult asking for help is far higher than most rollouts assume.

Is one-on-one support worth the time it costs? It is expensive and it is not for everyone, which is why it should be targeted. Marcus reserved it for the three or four people most likely to slip away and recovered two of his four lapsed users within six weeks, including the sourcer who became one of the team's steadiest users. Spread the same hours evenly across nine people and it would have helped nobody in particular. Concentrated on the at-risk group, it is the highest-return coaching you can buy.

Our team is too small to have dedicated champions. What then? The components still apply even when the roles collapse into fewer people. Name who holds the coaching responsibility, protect some portion of their time explicitly, agree on how questions get asked, and commit to closing the loop on feedback. What breaks small teams is not the absence of a formal program; it is coaching that everyone assumes is happening and nobody actually owns.