Marisol runs talent enablement at a 600-person healthcare-technology company, and this quarter her job has a single deliverable: get an AI screening and outreach toolkit into the daily workflow of all 14 recruiters on her team. She has the budget, the vendor contracts, and an executive sponsor. What she does not have, she realized after a deflating first training session, is adoption. Twelve of her 14 recruiters sat through a 90-minute demo of the new tooling, nodded politely, and went back to doing exactly what they did before. Two people used it. The tool was not the problem. Marisol had treated adoption as a software install when it is actually a learning problem, and learning for working adults follows rules that a vendor demo ignores entirely. This lesson is about those rules and how she used them to turn 2 adopters into 12.

Adoption Is a Learning Problem, Not a Login Problem

The mistake Marisol made in her first rollout is the most common one in talent enablement: she assumed that access plus a demo equals adoption. It does not. Her 14 recruiters are not blank slates waiting to be programmed. They are experienced professionals, several with a decade or more of sourcing and closing under their belts, each carrying a working theory of how recruiting gets done. A new AI tool asks them to suspend that theory and trust a system they did not build. That is a learning act, and adults do not learn the way children do. They will not absorb material just because it is on the agenda. They learn when the material solves a problem they already have, on terms they help set.

Technology adoption is a human change process, not just a technical implementation, and that is why understanding how adults learn is a leadership skill rather than a training-department skill. Marisol's reset started with a single reframe: stop asking "how do I train people on this tool?" and start asking "what problem does each recruiter have that this tool can visibly remove?" That question, repeated for every recruiter, became the backbone of her second rollout. It also pointed her toward a set of frameworks that gave structure to the instinct: Malcolm Knowles's principles of adult learning, the 70-20-10 model of how skills actually develop, the diffusion of innovations curve for sequencing who learns when, and Kirkpatrick's evaluation levels for proving any of it worked.

Andragogy: The Five Principles Marisol Built Around

Malcolm Knowles popularized andragogy, the theory of how adults learn, as a deliberate contrast to pedagogy, the teaching of children. It rests on five core assumptions: adults have an internal drive to learn because the material is relevant to their work; they bring substantial life and work experience; they are problem-focused rather than subject-focused; they are self-directed and want control over their learning path; and they respond better to intrinsic motivation than to external mandates. Traditional top-down training, the "here is the tool, use it" model, fails because it ignores every one of these. Effective training respects experience, connects to problems recruiters actually face, gives autonomy, and builds competence visibly. Marisol mapped each principle to a concrete change in her rollout.

Self-direction. Adults want control over how and when they learn. Marisol's first rollout was a fixed 90-minute lecture; her second offered a short core session plus a menu of optional deep-dives recruiters could choose from based on their own gaps. People who pick their own path show up engaged instead of resigned.

Drawing on experience. A seasoned recruiter's instinct for a strong candidate is an asset, not an obstacle. Instead of telling her team to trust the AI's screening, Marisol framed the tool as a way to handle the mechanical first pass so their judgment could focus on the close calls. She asked senior recruiters to share cases where they caught something a keyword filter would have missed, then showed how to encode that judgment into a better prompt. Their experience became the curriculum.

Readiness to learn. Adults learn when they are ready, which usually means when a real need is in front of them. Marisol stopped scheduling training for a generic future and started attaching it to live requisitions. A recruiter drowning in 200 applicants for a nursing-informatics role is ready to learn AI-assisted screening in a way no calendar invite can manufacture.

Problem-centered orientation. Adults want to apply learning to an immediate problem, not bank it for later. Every exercise in Marisol's second rollout used a real open requisition, real resumes from her ATS, and a real outreach sequence the recruiter would actually send on LinkedIn that week. No contrived sandbox data.

Internal motivation. External mandates produce compliance; internal motivation produces adoption. Marisol learned that different recruiters were driven by different things, and one message could not reach all of them.

Motivating 14 People Who Are Not Motivated by the Same Thing

The question underneath every rollout is rarely asked out loud: why should this recruiter care about learning this tool? To save time on boring work, to improve hiring quality, or to build a skill? Different team members are motivated by different outcomes, and when Marisol asked her recruiters what would make a new tool worth their effort, she got three distinct answers and stopped pretending one pitch would land for everyone.

Roughly half her team were efficiency-driven, energized by tools that eliminate tedious work. For them she led with time, quantifying the hours freed for higher-value work on their own requisitions: a resume-screening prompt that turned a three-hour shortlist into a 25-minute one was the entire pitch. About a third were quality-driven; they cared less about speed and more about not missing a strong candidate or letting a weak hire through. For them she emphasized consistency, showing how a structured prompt evaluates every applicant against the same rubric so fewer good people are lost in the pile, and how that reduces the variability that creeps in when a human reviews the two-hundredth resume of the day. The remaining recruiters were growth-driven, earlier in their careers and thinking about their resumes. For them, Marisol framed AI fluency as a durable, market-valued skill, and she made the tool champions visible so that becoming one was a recognized credential.

Same tool, three honest pitches, each tuned to a motivation that was already there. The discipline this requires is negative as much as positive: avoid one-size-fits-all messaging. A single all-hands announcement pitched at efficiency reaches whichever segment happens to match it and quietly loses the rest, and the people it loses will not tell you they were unpersuaded. They will simply not use the tool, and you will read that as resistance rather than as a messaging failure.

Assess Readiness Before You Train Anyone

Before Marisol scheduled a single session in her second rollout, she asked a question her first rollout had skipped entirely: is this team actually ready to learn this? Readiness is not one thing, and treating it as one thing is why so many rollouts stall. It has three dimensions, and a recruiter can be strong on one and weak on another. Skill readiness asks whether they have the prerequisite technical skills to operate the tool at all. Conceptual readiness asks whether they understand why the tool exists and what problem it solves. Confidence readiness asks whether they believe they can successfully use it. A recruiter with strong technical skills but low confidence will struggle, and no amount of feature training will fix that, because the obstacle is not knowledge. A recruiter with good conceptual understanding but weak technical skills needs remedial support before the tool session, not during it.

The practical consequence is that readiness assessment comes before formal training, and it produces a segmented plan rather than a single calendar invite. Which team members can jump straight into tool training? Which need foundational digital skills work first? Which need confidence-building before anything technical? Marisol segmented her 14 recruiters against those three questions and built a different entry point for each group. Skipping this step is what creates the bottleneck in most rollouts: you train people who are not ready, they experience the session as confusing or threatening, and the result is frustration and poor adoption that gets misattributed to the tool. If readiness is low, foundational work comes first. Do not skip it, and do not let a launch date decide it for you.

Experience Is an Asset, and Saying So Is the Whole Reframe

This is the point most often overlooked, and it is where experienced teams are won or lost. Experienced recruiters may resist AI because it feels like you are telling them their expertise is obsolete. That reading is not paranoia; it is a reasonable inference from how most rollouts are pitched. The opposite message, that AI amplifies their expertise, is both more honest and more motivating. Experienced recruiters are excellent at judgment: weighing soft skills, reading between the lines in a resume, assessing whether someone will thrive on a particular team. Those are exactly the places where AI adds no value. What AI does well is the mechanical, rule-based first pass that your most experienced recruiters are over-qualified for anyway.

So say it plainly. "This tool will handle the mechanical screening so you can focus on the judgment work you are good at" does real work in a room full of tenured recruiters. Marisol made the acknowledgment structural rather than rhetorical. She explicitly named the expertise in the room at the start of every session. She collected and shared examples where an experienced recruiter had caught something the tool missed, treating those as evidence of the design working rather than as embarrassments. And she framed the tool as a partner rather than a replacement in every artifact, including the ones her recruiters read without her present. That reframing, from threat to augmentation, removes a major barrier at almost no cost, and it is the barrier budget cannot buy through.

The 70-20-10 Model: Why the Demo Was Always Going to Fail

The 70-20-10 model describes how professional skills actually develop: roughly 70 percent from doing real work, 20 percent from learning through other people such as coaching and peers, and 10 percent from formal instruction like courses and demos. Marisol's first rollout was 100 percent of the 10 percent. She had poured all her effort into the smallest slice of how adults learn and wondered why nothing stuck. The model does not say formal instruction is worthless; it says formal instruction alone cannot produce a skill, and that a design which stops at the demo has funded the least productive third of the work and skipped the rest.

Her second rollout rebalanced. The formal demo shrank to a 30-minute core session covering only what recruiters needed to get started. The bulk of the design went into the 70: structured on-the-job application, where each recruiter used the tool on a live requisition within 48 hours of the session, while the need was hot and the steps were fresh. The 20 came from a deliberate peer structure, which is the piece most rollouts leave to chance. Knowles's point that adults trust peers as credible sources turned out to be the lever that moved her numbers, which is why she invested in tool champions rather than in more instruction.

Learning Transfer: Getting the Session to Survive Contact With the Job

A dimension often missed entirely is transfer: will learning in a training session translate to changed behavior on the job? A session can be enjoyable, well-attended, and comprehensible, and still change nothing on Monday. Adults are more likely to transfer learning when three conditions hold. The learning is directly relevant to their work. They practice on realistic job scenarios rather than contrived ones. And they have support and reinforcement on the job after training, when the first real obstacle appears and the session is a week behind them.

Transfer is central to design rather than an afterthought, and it has to be built in from the start rather than bolted on when adoption numbers disappoint. Practically, that means using realistic recruiting scenarios instead of invented examples, having recruiters practise on actual job descriptions and candidate profiles from your own system, arranging peer support and coaching for the period after training rather than only during it, and scheduling follow-up sessions to address the real questions that emerge once people apply the tool to live work. Those questions are the most valuable material in the programme, because nobody could have anticipated them in the design phase, and a rollout with no forum for them loses that learning entirely.

Sequencing the Rollout With the Diffusion of Innovations Curve

Everett Rogers's diffusion of innovations model explains how a new practice spreads through a population in five groups: innovators who try anything new, early adopters who are respected and influential, the early majority who adopt once they see proof from people they trust, the late majority who adopt under social and practical pressure, and laggards who resist until there is no alternative. Marisol's failure was trying to convert all 14 recruiters at once, late majority and all, with a single demo aimed at nobody in particular.

Her second attempt followed the curve. The two recruiters who had adopted on their own were her innovators and early adopters; she made them tool champions, gave them recognition, and had them run short peer demos using their own wins. The early majority, about six recruiters, adopted once those champions, not Marisol and not the vendor, showed real results on real requisitions. The late majority came along as adoption became the visible norm and as the easy manual fallbacks were gently retired. The lone laggard she handled individually rather than dragging through group training, which spared both of them a pointless session. Sequencing the rollout to match the curve meant each group learned from the group just ahead of it, exactly as the 20 in 70-20-10 predicts.

Incentives, Champions, and Why Mandates Backfire

Motivation systems are the quiet half of an adoption plan. Is adoption tied to performance reviews? Is there recognition for early adopters? Is there support and training for people who struggle? Incentive design is subtler and more powerful than it looks. Consider the case where adoption is formally optional but performance metrics still include speed and quality. Early adopters see immediate performance gains while non-adopters gradually fall behind, and that creates natural adoption pressure without any explicit mandate. Nobody has been told to use the tool. The tool has simply become the easier path to the outcome everyone is already measured on.

Peer recognition compounds this. Designating tool champions, early adopters and power users who become peer teachers, and recognising them publicly leverages the adult learning principle that peers are credible sources of information. A recruiter learning from another recruiter trusts the information more than the same content from training materials or an external expert, because the peer has the same requisitions, the same hiring managers, and the same reasons to be skeptical. Marisol's two champions were worth more than any additional session she could have run.

What does not work is the purely punitive approach: use this tool or lose your job. Mandates of that shape create resistance and produce surface compliance rather than genuine adoption, a distinction that matters enormously in practice. Surface compliance looks like adoption in a usage dashboard and delivers none of the benefit, because people run the tool to generate a log entry and then do the work the way they always did. Design systems where adoption is the natural path to success, and let the metrics you already have do the persuading.

Measuring Whether It Actually Worked: Kirkpatrick's Four Levels

Marisol's first rollout had one metric: attendance. Fourteen people showed up, so on paper it was a success, which is precisely how she fooled herself. To avoid repeating that, she evaluated her second rollout against Kirkpatrick's four levels of training evaluation. Level 1, Reaction, asks whether recruiters found the session relevant and useful; a two-question pulse after each session, not just a smile sheet, told her whether the problem-centered framing was landing. Level 2, Learning, asks whether they could actually do it; she checked whether each recruiter could run a screening prompt and produce a usable shortlist without a walkthrough.

Level 3, Behavior, asks whether they are using it on the job, weeks later, unprompted. This is the level her first rollout failed completely, and the only one that equals real adoption. Level 4, Results, asks whether it moved outcomes that matter, such as time-to-shortlist and recruiter capacity. Levels 1 and 2 are easy and seductive because they can be measured in the room and they almost always look good; Levels 3 and 4 require waiting and are where adoption is actually proven or disproven. A programme that reports only the first two is not reporting adoption, it is reporting that a session occurred.

A Worked Example: Marisol's Two Rollouts

Here is how Marisol's two attempts compared on her own illustrative figures. Both started from the same 14 recruiters and the same toolkit. The first rollout was a single 90-minute demo. Eight weeks later, 2 of 14 recruiters, about 14 percent, were using the tool on live requisitions, a Level 3 result of essentially zero, and the share of new requisitions touched by the new tooling sat at roughly 10 percent. Time-to-proficiency for the two self-starters was whatever they taught themselves; for everyone else it was never.

The second rollout inverted the design. Marisol spent about 30 minutes of formal instruction per recruiter, then roughly 3 hours of structured on-the-job practice on real requisitions over the following two weeks, supported by two peer champions running short demos. By week 8, 12 of 14 recruiters, about 86 percent, were using the tool on live requisitions, and around 70 percent of new requisitions were running through the AI screening step. Median time-to-proficiency, defined as producing a usable AI shortlist without help, landed near 9 working days. On Level 4, her efficiency-driven recruiters reported cutting first-pass screening on a high-volume requisition from about 3 hours to about 25 minutes.

The tool did not change between rollouts. The learning design did, and that is the entire point. Notice which lines are causes and which are effects: peer support and on-the-job practice are inputs Marisol controlled directly, and the adoption rows are outputs that followed. If you are looking for a place to intervene in a stalled rollout, it is almost never the tool and almost never more instruction. It is the 70 and the 20.

Anti-Patterns

• Treating adoption as a software install. Provisioning access, running a demo, and recording attendance produces a completed project plan and no behavior change. The gap between access and adoption is not effort, it is design.
• One-size-fits-all motivation messaging. A single pitch reaches whichever segment happens to match it and quietly loses the rest. Efficiency-driven, quality-driven, and growth-driven recruiters need different honest framings of the same tool, and the people your message missed will not tell you; they will simply not use it.
• Skipping readiness assessment. Training people who are not ready creates the bottleneck it was meant to remove. A recruiter short on confidence or on prerequisite skills experiences the session as confusing rather than useful, and the resulting frustration gets misread as resistance to the tool.
• Framing AI as a replacement for expertise. Telling experienced recruiters, implicitly or otherwise, that their judgment is obsolete guarantees resistance and is also untrue. The augmentation message is both more honest and more motivating, and refusing to say it out loud costs you your most capable people first.
• Relying on mandates and punishment. "Use this tool or lose your job" produces surface compliance, which registers as adoption in a usage dashboard and delivers none of the benefit, because people generate the log entry and then work the way they always did.
• Treating transfer as an afterthought. Contrived sandbox exercises, no post-training coaching, and no forum for the questions that only appear during real use. The first obstacle arrives a week later with nobody to ask, and the recruiter reverts to the method that never fails them.

Practice Prompts

• Assess readiness across your team on the three dimensions separately: skill, conceptual, and confidence. Who can go straight into tool training, who needs foundational digital skills work first, and who needs confidence-building before anything technical? Produce a segmented plan rather than a single invite.
• Survey a sample of your team on what would make a new tool worth their effort, and sort the answers into efficiency, quality, and professional growth. Then write the three pitches, one per segment, each concrete and each true.
• Audit your last rollout against 70-20-10. Estimate what share of your design and budget went to formal instruction versus on-the-job application versus peer learning. If the formal slice dominates, redistribute before you run anything again.
• Identify your innovators and early adopters by looking at who has already adopted something without being asked. Name two as tool champions, recognise them publicly, and have them run a short peer demo using their own results rather than your slides.
• Design the transfer, not just the session. For your next rollout, specify the live requisition each recruiter will apply the tool to, the deadline for that first application, who they can ask when they get stuck, and the date of the follow-up session where their real questions get answered.
• Write the Level 3 and Level 4 measures you will report before you run the training, and fix the week you will report them. Deciding this in advance is what stops attendance becoming the metric by default.

Reflection

• Think of a tool you adopted enthusiastically and one you were told to adopt and never really did. What was different about how each entered your working life?
• Which of the three readiness dimensions is weakest on your team right now, and what would you have to do differently if the answer is confidence rather than skill?
• When you last announced a new tool, whose motivation was your message actually written for? Who on your team would have needed a different pitch?
• Who on your team would other recruiters believe if they said the tool was worth using? What is stopping that person from being your champion?
• If you measured your last training programme at Level 3 today, weeks after the session, what would you find, and would you want to know?

Glossary

• Andragogy. The theory of how adults learn, popularized by Malcolm Knowles as a deliberate contrast to pedagogy. Its five assumptions: adults are internally driven by relevance to their work, bring substantial experience, are problem-focused rather than subject-focused, are self-directed, and respond better to intrinsic motivation than to mandates.
• Readiness. Whether a learner is positioned to learn, assessed on three separate dimensions: skill readiness (prerequisite technical ability), conceptual readiness (understanding what the tool is for), and confidence readiness (belief that they can succeed with it).
• Learning transfer. Whether learning in a training session translates into changed behavior on the job. It is more likely when the learning is directly relevant, practice uses realistic scenarios, and support and reinforcement exist after training.
• 70-20-10 model. A description of how professional skills develop: roughly 70 percent from doing real work, 20 percent from other people such as coaching and peers, and 10 percent from formal instruction.
• Diffusion of innovations. Everett Rogers's model of how a new practice spreads through a population in five groups: innovators, early adopters, the early majority, the late majority, and laggards.
• Tool champion. An early adopter or power user designated as a peer teacher and publicly recognised, used because peers are more credible sources of information to adults than trainers or external experts.
• Kirkpatrick's four levels. An evaluation model for training: Reaction (did they find it useful), Learning (can they do it), Behavior (are they doing it on the job weeks later), and Results (did outcomes move).

Related Lessons

• Training Program Design: Curriculum, Delivery, and Evaluation takes these principles into the full design of a curriculum, including the delivery formats and evaluation instruments this lesson only names.
• Coaching and Support: Helping Individuals Build Confidence develops the confidence dimension of readiness and supplies the post-training reinforcement that learning transfer depends on.
• Change Leadership: Driving Adoption While Managing Resistance handles the resistance this lesson traces to a replacement framing, at the scale of an organization rather than a team.
• Creating Communities of Practice: Learning Networks builds out the 20 in 70-20-10 into a durable peer structure rather than two designated champions.
• Measuring Adoption: Tracking Usage, Proficiency, and Impact goes deeper into the Level 3 and Level 4 instrumentation, including how to tell genuine adoption from surface compliance in a usage dashboard.

Closing

Marisol's second rollout worked because she stopped treating her recruiters as an audience and started treating them as adult learners with existing expertise, real problems, and their own reasons to care. Nothing she did was expensive. She shortened the demo, asked people what motivated them, assessed who was actually ready, said out loud that experience was an asset, put the practice on live requisitions, let two peers do the persuading, and measured what mattered instead of what was easy. The frameworks here are useful mainly because they make each of those moves obvious in advance rather than in hindsight. If a rollout on your team has stalled, the diagnosis is rarely the tool and almost never a need for more instruction. Ask which principle you skipped, and start there.

Key Takeaways

• Adoption is a learning problem, not an install problem. Access plus a demo does not equal adoption. Working adults adopt a tool when it visibly removes a problem they already have, on terms they help set, not because it appeared on a training agenda.
• Design around Knowles's five andragogy principles. Self-direction, drawing on experience, readiness to learn, problem-centered orientation, and internal motivation each translate into a concrete change: choice in the path, respect for existing expertise, training tied to live needs, real requisitions instead of sandbox data, and pitches tuned to what actually motivates each person.
• Assess readiness on three dimensions before you train. Skill, conceptual understanding, and confidence are separate, and a recruiter can be strong on one and weak on another. Segment the plan accordingly, because training people who are not ready produces frustration that gets misread as resistance.
• Segment your motivation message. Efficiency-driven, quality-driven, and growth-driven recruiters need different honest pitches for the same tool. One-size-fits-all messaging reaches whichever segment happens to match it and loses the rest.
• Frame AI as augmenting expertise, and say it explicitly. Experienced recruiters resist when a tool implies their judgment is obsolete. The honest message is that AI takes the mechanical first pass they are over-qualified for, freeing the judgment work they are good at.
• Respect the 70-20-10 model and build transfer in. Roughly 70 percent of skill comes from doing real work, 20 percent from peers and coaching, 10 percent from formal instruction. Transfer needs relevance, realistic practice, and post-training support, designed from the start rather than added when adoption disappoints.
• Sequence the rollout with the diffusion of innovations curve. Do not convert all 14 people at once. Start with innovators and early adopters as tool champions, let the early majority learn from their proof, and handle laggards individually rather than dragging them through group training.
• Make adoption the natural path, not a mandate. Incentives that let early adopters visibly outperform create adoption pressure without coercion, while "use it or else" produces surface compliance that shows in the dashboard and delivers nothing.
• Measure adoption with Kirkpatrick's four levels. Attendance and good reactions are Levels 1 and 2, seductive because they are easy. Real adoption lives at Level 3, sustained on-the-job behavior, and Level 4, results that move outcomes like time-to-shortlist. The learning design, not the tool, drives those numbers.

Frequently Asked Questions

My team is fully remote and I cannot run a room. Does any of this still apply? All of it, though the delivery changes. The principles are about how adults learn rather than about physical rooms: relevance to real work, control over the path, respect for existing expertise, practice on live requisitions, and credible peers. Remote helps some of these, because a menu of short optional deep-dives is easier to offer asynchronously than to timetable in person, and on-the-job application is unaffected by location. The part needing deliberate design is the 20, peer learning, because informal peer teaching happens by accident in an office and by appointment when remote. Name your champions, give them a scheduled slot, and make their wins visible somewhere the whole team reads.

What if my most experienced recruiters are the most resistant? That is the expected pattern, not a sign something has gone wrong, and it usually means the augmentation reframe has not landed. Their resistance is a rational response to an implied claim that their expertise is obsolete. Start by naming their expertise out loud and specifically, then ask them for the cases where their judgment caught something a mechanical filter would have missed, and use those cases as teaching material. That does two things at once: it makes the tool's limits explicit, which experienced people find far more credible than a claim of accuracy, and it puts their expertise at the centre of the curriculum rather than in opposition to it. Recruiters who help build the training rarely refuse to use it.

How long should I wait before deciding a rollout has failed? Long enough for Level 3 to be meaningful, which means weeks of unprompted use rather than the days it takes to collect a reaction score. Marisol read her results at week 8 in both rollouts, which let her compare like with like. The temptation is to check early, see enthusiastic reaction scores, and conclude it worked, or to check at week 2, see low usage, and conclude it failed before transfer support has even happened. Set the measurement week in advance, keep the coaching and follow-up running until then, and resist reading the dashboard as a verdict before the design has finished executing.

Do I need tool champions if I already have a competent trainer? Yes, and they do different jobs. A trainer delivers the 10, the formal instruction, and does that better than a peer. A champion supplies the 20, credibility, because adults trust peers as sources of information in a way they do not trust trainers or external experts. The champion has the same requisitions, the same hiring managers, and the same reasons to be skeptical, which is what makes their endorsement worth something. Marisol's two champions were not better instructors than she was. They were more believable, and belief was the constraint.

Our usage dashboard says adoption is high but nothing has changed. What am I looking at? Most likely surface compliance: people running the tool often enough to register in the log, then doing the work the way they always did. It is the characteristic output of adoption pressure that is punitive rather than natural. Diagnose it by moving up a level. At Level 3, ask whether the tool's output is actually being used, meaning whether shortlists and drafts reach a hiring manager, rather than whether the tool was opened. At Level 4, check whether any outcome moved. If usage is high and outcomes are flat, the problem is not adoption depth but the incentive structure that produced it, and more mandate will make it worse.