Adult Learning Methods: 10 Ways to Teach Adults at Work

Train teams faster and more effectively with AI

Start Free

Imagine a new starter who spends a morning watching slides about a procedure, passes the ten-question quiz and then gets the procedure wrong on shift. A content quiz may check useful knowledge without showing whether someone can carry out the task. The training needs an appropriate attempt as well as an explanation.

I'm a cofounder of Coursebox, where we make an AI tool for employee training. It builds the online parts of programmes like the one in this guide; it does not run the in-person parts, and I'll say where it fits. Before Coursebox I was lead learning designer on three edX courses at the University of Adelaide, and every course I design still comes down to three kinds of page: a lesson, a practice activity and an assessment. In this guide I cover ten adult learning methods, the evidence for each, and which kind of training goal each one fits. Then I walk through one workplace programme that uses all ten, three real cases with numbers, a product test and a planner you can download.

What are adult learning methods?

Adult learning methods are the activities a trainer uses to help adults at work learn a job, such as worked examples, retrieval practice, role play and on-the-job practice, each suited to a different kind of learning goal.

A method is what people study, attempt and receive feedback on. A theory explains why a design may help; our guide to adult learning theory covers Knowles and andragogy. Delivery format means online, in person or blended. A method can use several formats, but the practice can change: a video-call role play allows people to choose their words, while a choice-based scenario rehearses decisions from supplied options.

The method has to fit the goal, and workplace training goals come in four kinds, plus a fifth step that every one of them needs:

• Knowledge: facts and concepts people must recall, such as product names or the fields on a form.
• Procedure: a sequence of steps done the same way each time, such as taking a measurement or entering an order.
• Judgement: a decision where the answer depends on the situation, such as when to refer a customer to someone else.
• Behaviour: how people act with others, such as handling an upset customer.
• Transfer: doing all of it on the job, with real customers and real consequences.

Training at work reaches a lot of adults. A 2016 Pew survey of US adults found that "63% of those who are working" had taken a course or extra training in the previous 12 months to improve their job skills, and more of them said that training took place at a work-related venue (75%) than online (55%).

Do adults need different teaching methods?

Adults bring different experience, responsibilities and reasons for learning. Prior knowledge, access needs and the job help determine the design; age alone does not select a method. In their review of ten learning techniques, Dunlosky and colleagues rated practice testing and distributed practice highly across ages and abilities, including evidence with middle-aged and older adults. That supports trying these techniques across age groups, rather than assuming all adults learn identically.

Learning styles do not change that. Pashler and colleagues found "ample evidence that children and adults will, if asked, express preferences" for how they learn, but no adequate evidence that teaching to those preferences helps. Our guide to Kolb's experiential learning theory sets out that evidence. Knowles' assumptions about adult learners are a fair starting point for relevance, and the debate about how far they hold sits in our comparison of pedagogy and andragogy.

Experience counts, but in a specific way. An adult who has worked in retail for ten years brings useful material: real customers, real mistakes, real cases. The same adult starting a technical job is still a novice in that domain, and novices need worked examples and close feedback before they work alone.

Lesson, practice and assessment

In my terms, a lesson is where people read or watch, a practice activity is where they do something and get feedback, and an assessment is where someone checks they can do it. Most courses have far too much of the first. The first question I ask a training provider converting a unit into a course is where the learner actually practises.

I also looked at a local sample of 104 Coursebox course requests from a 90-day export ending 7 October 2026. These were English-browser requests with no file upload, one prompt per person, classified in two independent passes and reconciled. Some exported prompts stopped at 1,500 characters. This is a sample of briefs, not every request or an inspection of the resulting courses.

The same classification identified practice beyond reading or recall in 29 of the 104 requests, including eight of the 27 workplace-course requests. One workplace request named a role play or simulation and one named spaced review. Truncation and classification judgement limit these estimates. The briefs do not show what the finished courses contained. The practical use is to name the required attempt and feedback in the brief rather than assume a topic list will produce them.

What the research says about teaching adults at work

The clearest evidence for doing over listening comes from university classrooms, not workplaces, so read it as a guide to design. Freeman and colleagues ran a meta-analysis of 225 studies of university science, technology, engineering and maths courses, comparing traditional lecturing with active learning, where students take part "through activities and/or discussion in class, as opposed to passively listening to an expert".

The same paper found that "average examination scores improved by about 6% in active learning sections". The measured effect on exams and concept inventories was 0.47 standard deviations. That figure is an effect size, not a percentage, so it does not mean "47% better".

The workplace evidence comes from a meta-analysis of organisational training. Arthur and colleagues pooled 397 effects from 162 sources on training in organisations and found what they call "a medium to large effect size for organizational training".

Two findings from that paper shape how to choose a method. First, effectiveness varied with "the specified training delivery method, the skill or task being trained, and the criterion" used to measure it, so the method has to fit the skill. Second, lectures did better than their reputation. In the authors' words, lectures, "contrary to their poor public image, appeared to be quite effective in training several types of skills and tasks". The lesson for a trainer is to fix the lecture, not ban it: keep the explanation short and follow it with recall and practice.

There is a reason trainers drift back to lecturing anyway. In a randomised experiment with 149 Harvard physics students, Deslauriers and colleagues found that students in active sessions "scored almost half an SD (0.46) higher on the examination" but rated how much they felt they had learned more than half a standard deviation lower. The authors also cite reports that "one-third of instructors who try active teaching eventually revert to passive lectures, many citing student complaints as the reason".

The 10 adult learning methods at a glance

These are the ten methods this guide covers, the goal each fits best, and what each takes to run. Experiential learning is not on the list as a method of its own: it is a cycle that sits behind several of them, simulation and role play in particular.

Ten methods, worked through one training programme

To show where each method fits, here is one programme that uses all ten. The example is hypothetical; the business, the people and the numbers are illustrative.

A group of six optometry practices in North Queensland, at Cairns, Port Douglas, Townsville, Ayr, Mackay and Rockhampton, has a problem with new staff. Lens orders come back from the lab to be remade because a measurement or a field on the order was wrong. Some customers come back unhappy with new progressive lenses because nobody told them what adapting feels like or when to return.

The group has hired 8 new optical dispensing assistants for its February intake: 2 at Cairns, 1 at Port Douglas, 2 at Townsville, 1 at Ayr, 1 at Mackay and 1 at Rockhampton. They come from retail, hospitality and office jobs, and none has dispensed glasses before. Anika, the group's dispensing lead in Townsville, designs and runs the programme. Each practice has a senior dispenser who acts as the new assistant's buddy. Joel is the practice manager at Mackay, and Tahlia is the new assistant there.

By the end of week 8, each assistant should be able to read a prescription and enter a lab order that passes the group's 6-point order check:

1. Prescription match: every field on the order (sphere, cylinder, axis, add) matches the optometrist's prescription, including the plus or minus sign.
2. Lens choice: lens type and coatings match what the customer chose and signed for.
3. Monocular PDs: a pupillary distance is recorded for each eye, not one combined figure.
4. Fitting heights: recorded for progressive and bifocal lenses, measured in the customer's chosen frame after it has been adjusted to sit correctly.
5. Frame details: lens width, bridge and temple length entered from the frame itself.
6. Second check: a second person checks the order before it goes to the lab.

They should also be able to take measurements within the group's own training tolerance of 1 mm, recommend lens options that fit how the customer will use them, and run the collection and adaptation conversations to the group's standard. Anything about symptoms or a change in vision goes to the optometrist. The assistants never examine eyes or change a prescription.

The tolerance and number of observations are illustrative training inputs, not a clinical standard. The responsible practice professional sets the actual measurement criteria, supervision and sign-off requirements. The second-person order check remains in place after worked examples fade.

This example shows the methods well because one role needs all four kinds of goal: knowledge (lens types, coatings and prescription fields), procedure (measuring and entering orders), judgement (which lens options fit a customer, and when to refer) and behaviour (the adaptation conversation). It also has a half that can be learned from a screen and a half that cannot.

The programme runs over 8 weeks in seven parts:

In my terms, the fully worked examples and Anika's modelled conversation are the lesson, the rest of parts 2 to 6 is practice, and the order check and the sign-off are the assessment. Most of the programme is practice, which is the point.

1. Worked examples

A worked example shows a full solution step by step for the learner to study before trying one alone. Sweller and colleagues, in their 2019 review of cognitive load research, describe worked examples as providing "a full problem solution that learners must carefully study", and report strong evidence for them with an important limit: they are "less effective for high-expertise learners". Their answer is the "completion strategy": start with problems that are almost fully solved and hand over more of each solution until learners generate most of it themselves. Our guide to cognitive load theory explains why studying examples helps novices and why the help fades.

• The goal: Turn a prescription into a lab order that passes the 6-point check.
• The method in practice: Three fully worked prescriptions become lab orders, one line of reasoning per step. A fourth leaves the last two steps for the assistant, and a fifth is done alone.
• What to check: The fifth order passes the 6-point check without help. If it does not, give one more partly worked example.

Tahlia's fourth example comes back with the cylinder sign dropped, so she gets another partly worked one before the fifth. Once an assistant gets two orders right alone, stop giving worked ones.

2. Retrieval practice

Retrieval practice means answering from memory instead of rereading. Dunlosky and colleagues note that practice testing "is not particularly time intensive relative to other techniques, and it can be implemented with minimal training". Rowland's meta-analysis of the testing effect found larger benefits when the first test asked people to recall an answer than when it asked them to recognise one from a list, so use short answers and "explain why" questions where you can. A related experiment, covered in our guide to cognitive learning theory examples, found that rereading looked better five minutes later and testing looked better a week later.

• The goal: Recall lens types, coatings and prescription fields without notes.
• The method in practice: Each induction morning opens with 6 questions answered from memory, at least 3 of them situations, with the answer and the reason shown after each.
• What to check: Which questions most people miss. Those go into the day 2 review.

The AI quiz generator can draft questions from a topic or document. Check the actual question type and learner view: instant grading does not by itself show that a useful explanation follows each answer. Add or edit the correction where needed, then inspect it after submission.

3. Spaced practice

Spaced practice spreads short reviews over days and weeks instead of one block. Dunlosky and colleagues rated it highly because "it works across students of different ages, with a wide variety of materials", and over long delays. The longer you need people to remember something, the wider the gaps between reviews should be; our guide to spaced repetition covers how to set them.

• The goal: Keep the knowledge from the pre-course and induction for months, not days.
• The method in practice: Short reviews at day 2, day 7, day 21 and day 45 after induction, 6 questions each, mixing old and new questions. Anika assigns each review on its day, and the buddy asks about it at the next shift.
• What to check: Scores on questions first met weeks earlier, not only the newest ones.

4. Deliberate practice with feedback

Deliberate practice is repeating one defined task with immediate feedback on each try. Ericsson and colleagues, in their 1993 paper on expert performance, set out the conditions: people "should receive immediate informative feedback and knowledge of results of their performance" and "repeatedly perform the same or similar tasks". Without feedback, they wrote, "mere repetition of an activity will not automatically lead to improvement". That research was on musicians. Hours alone predict little at work: a later meta-analysis summarised by the Association for Psychological Science found deliberate practice explained less than 1% of the differences in performance in professions.

The feedback itself needs care. Kluger and DeNisi's meta-analysis of 607 feedback effects found that feedback improved performance on average, but in over a third of cases it made performance worse. It worked less well as attention moved "closer to the self and away from the task". Feedback should name the measurement and its cause, not the person.

• The goal: Take monocular PD and fitting-height measurements within the group's 1 mm tolerance.
• The method in practice: In pairs at induction, each assistant takes 10 practice measurements on colleagues. Anika compares each with her reference reading and checks the cause of a mismatch. Eight within the fictional training tolerance prompts the next coaching decision; it does not excuse the other two misses or authorise independent dispensing. Repeat and check any required skill that remains unreliable.
• What to check: How many are within tolerance, and whether the misses share a cause, such as head tilt or a frame that was not adjusted first.

Tahlia's first 10 readings have 6 within tolerance. Anika notices she measures before adjusting the frame, says so, and Tahlia's next round has 9 of 10 within tolerance. The feedback was about the order of two steps, not about Tahlia.

5. Case-based learning

Case-based learning asks people to work through a real situation from the job and decide what should have happened. It is popular, and the evidence on what it adds is thinner than its popularity. A review of 104 papers in health professional education by Thistlethwaite and colleagues found that "Overwhelmingly, students enjoy CBL and think that it enhances their learning", but the data taken together did not show whether it beats other activities. Use cases as a vehicle for practice and feedback, not as a method that works by itself. Our guide to scenario-based learning covers how to write them.

This is also where workplace experience pays off. The best cases come from the business's own real mistakes, and experienced staff can bring cases of their own.

• The goal: Spot where an order went wrong and which check would have caught it.
• The method in practice: One real, anonymised remake from the group's practices each Thursday call: the order as entered, what came back, and the question "at which of the 6 points would this have been caught?"
• What to check: Whether assistants name the point and the check, rather than saying "be more careful".

6. Problem-based learning

In problem-based learning, learners investigate a problem with guidance and identify what they need to learn to address it. Dochy and colleagues' review of 43 tertiary-education studies reported more consistent skill than knowledge benefits. It supports considering this approach for relevant skills, not claiming it teaches only judgement or cannot teach facts. In the example below, Anika guides the investigation and supplies the approved procedure before staff perform the task.

• The goal: Work out what to check when a customer reports a problem with new lenses.
• The method in practice: On induction day 2, two groups of 4 get a customer whose new progressives "blur when reading" and list what they would check, such as the fitting height, a frame sitting low or the lens type ordered. Then Anika teaches the fitting-height rule.
• What to check: Each group's list, compared with the 6-point check. Anything clinical goes on the list as "refer to the optometrist".

7. Simulation

A simulation is the whole task done in a safe copy of the job. A large review comes from health professions: Cook and colleagues pooled 609 studies of technology-enhanced simulation with 35,226 practising and student health professionals, and found large effects on knowledge, skills and behaviour. Read that number with care. The large effects were compared with no training at all, and 405 of the 609 studies tested one group before and after, with no comparison group. It shows simulation beats nothing, not that it beats a good workshop.

• The goal: Do a full dispense end to end, from prescription to checked order.
• The method in practice: On the afternoon of induction day 2, each assistant runs a mock dispense in the training room with a colleague as the customer, real frames and the practice system in its training mode. It ends with an order that a second assistant checks against the 6 points.
• What to check: The order passes all 6 points. Time is not scored.

A small police study in the Netherlands tested what happens when practice includes some of the pressure of the real job.

Case study: Police officers training to shoot under pressure in the Netherlands

Serving police officers in the Netherlands, all licensed to carry a handgun on duty and with about 11.5 years' experience on average, took part in extra training at a police academy. The authors cite US reports that officers hit "above 90%" on low-pressure tests but "around or below 50%" when firing in the line of duty.

• What they did: 27 volunteers were split at random into two groups of 13 and 14. Both had four 1-hour sessions with the same exercises in realistic settings such as a police car, a building and a shopping street. The only difference was that one group practised against an experienced firearms instructor who sometimes fired back with coloured soap cartridges.
• Evidence and result: In Nieuwenhuys and Oudejans' 2011 randomised trial, the group that trained under pressure shot more accurately under pressure at the posttest than the control group (P = 0.023), and their drop in accuracy under anxiety had gone. The control group still lost accuracy under anxiety.
• Limitation: The sample was 27 volunteers at one site, and accuracy was measured in a staged test, not in real incidents. At a retention test 4 months later, the control group had also kept its performance under pressure, so a lasting advantage was not shown.
• Lesson: If people must perform under pressure, practise under some of that pressure. Same drills and same hours: only the pressure changed, and at the posttest that group no longer lost accuracy under stress.

8. Role play with behaviour modelling

Behaviour modelling is a structured role play: people watch a model of the behaviour, the key behaviours are named, and then they practise those behaviours and get feedback on them. Taylor and colleagues' meta-analysis of 117 behaviour modelling studies found that what people knew about the training faded over time, while "training effects on skills and job behavior remained stable or even increased". Our guide to social cognitive theory explains the modelling side, and our role play examples give scenarios to start from.

• The goal: Handle a customer who is struggling with new progressive lenses.
• The method in practice: Anika models the adaptation conversation with a customer who says "these progressives make me dizzy on stairs", naming the key behaviours as she goes. Then each assistant practises it twice with a colleague and gets feedback against the same behaviours.
• What to check: Acknowledge the concern, refer symptoms such as dizziness to the optometrist promptly, and check the frame and order only within the assistant's authorised role. Use the practice's approved adaptation explanation and return plan; a generic reassurance or a training script must not delay clinical review.

The police case applies here too. A role play where the colleague pushes back, the way a real customer would, is closer to the job than a calm walkthrough.

9. Peer teaching

Peer teaching asks learners to prepare a topic and teach it to others. The explainer often learns most. Kobayashi's meta-analysis on learning by teaching found the benefits of studying to teach someone face to face "are greater than" those of preparing a video or written explanation instead, though he calls the evidence "still inadequate and subject to some exceptions". These studies were of students. In a sales team, a programme built around structured peer coaching was tested in a field experiment.

• The goal: Understand one lens topic well enough to explain it.
• The method in practice: In week 4, each assistant teaches one lens topic to the group in 5 minutes, such as photochromic lenses or why the order needs two PDs, and takes two questions.
• What to check: Anika checks accuracy, and any gaps go back into the spaced review.

Case study: Sales agents coaching each other at a US call centre

At an inbound sales call centre selling television, phone and internet services across the United States, agents at the 75th percentile brought in about 48% more revenue per call than agents at the 25th. Managers put the gap down to know-how, such as knowing which products to bundle, but agents did not ask each other for it.

• What they did: Each agent was paired with a randomly chosen partner. In the structured meetings group, each agent filled in a short worksheet about their most difficult call and how it could have gone better, met their partner to discuss it, recorded the partner's advice, and handed the worksheet in by Wednesday each week. Pairs who handed it in got a free catered lunch. Another group was paid prizes for the pair's joint improvement instead, and a third group got both.
• Evidence and result: In Sandvik and colleagues' pre-registered field experiment, 653 agents were randomised by manager, with a further 83 in a control office. Structured meetings brought 24% more revenue per call than the control groups during the four weeks. Twenty weeks after the meetings ended, the structured meetings group and the group that got both were still 18% to 21% ahead. The prizes-only group no longer showed a statistically reliable advantage over controls at that follow-up. The biggest gains went to low performers paired with high performers, and the firm later used the same protocol to pair experienced agents with new recruits.
• Limitation: This was one firm and one job with fast feedback on sales, and the authors question how far it generalises. Agents knew they were in a study, and the treatment bundled pairing, a worksheet and a lunch, so which part did the work is unknown. The figures come from the working paper.
• Lesson: Take the awkwardness out of asking. Pair a stronger and a weaker performer, give them one real recent piece of work to talk about, set a deadline, and make the meeting easy to have. Paying pairs for results did not, on its own, produce a lasting gain.

10. On-the-job practice

On-the-job practice is doing real tasks at work with someone checking, and it is where every other method gets tested. Most of the evidence for on-the-job practice comes from studies of training in organisations. A later review of transfer by Blume and colleagues found that a supportive work environment was one of the things linked to people using their training on the job. Salas and colleagues' summary of the science of workplace training says "what occurs before and after training" weighs as much in a programme's success as the training itself. Our guide to training someone at work covers the buddy's side, and our comparison of coaching, training and mentoring covers where a coach fits.

• The goal: Dispense to real customers without the buddy's corrections.
• The method in practice: In weeks 2 to 6, each assistant dispenses to real customers at their own practice. The buddy second-checks every order, and the assistant keeps a short log of anything new each day.
• What to check: The week 8 sign-off: 3 observed dispenses against the observation checklist, with orders passing the 6-point check without the buddy's corrections.

Joel signs Tahlia off at Mackay in week 8 after her 3 observed dispenses, with Anika watching one of them on video.

Case study: Customer service agents practising on live calls in the Netherlands

The in-house call centre of a multinational mobile network operator in the Netherlands was missing its targets, and management wanted to cut the average time agents took to handle a call.

• What they did: Agents took 10 half-day sessions over five consecutive days. Half the sessions were coached technique training or group discussion. In the other half, real customer calls were routed to the training room, and agents handled them or listened to colleagues' calls while the coach and their own team leader supervised.
• Evidence and result: In De Grip and Sauermann's randomised field experiment, 37 of 86 eligible agents were picked at random for the training, and the rest were trained later. Across 74 agents and 1,859 agent-weeks of the firm's own data, the training improved performance, measured as average handling time, by about 10%, or 8.8% after allowing for a time trend. Call quality did not drop. Untrained teammates improved a little too, though that result was statistically weak.
• Limitation: Only experienced agents were eligible. The programme mixed coaching, discussion and live calls, so the result cannot be credited to live practice alone. By about week eight the effect estimate was no longer statistically reliable; that does not prove the improvement had disappeared.
• Lesson: This programme combined real work with coaching and team-leader support. Plan a later work check rather than assume the initial gain will last. The trial did not isolate the effect of follow-up.

Our guide to observation checklists shows how to write and score one.

How to choose a method for the goal

The best learning methods for adults are the ones that fit the goal. Arthur and colleagues found training effectiveness varied with the skill or task being trained. I apply that by matching each goal type to the methods that fit it. The mapping is a design judgement built on the evidence in each section, not a rule from one study.

Most programmes need several rows, because most jobs mix goal types. Start by working out what people cannot yet do; our training needs analysis questionnaire helps with that. Then pick the methods row by row, and put most of the time into practice. Lessons are the quickest part to write, which is one reason courses end up with too many of them.

The planner has the dispensing programme filled in, with blank pages for your own goal.

Where these methods go wrong

Five things commonly go wrong, and the dispensing programme has a fallback for each:

• Worked examples stay too long: Once someone gets two orders right alone, stop giving worked ones. Past that point the examples slow experienced learners down, and the buddy's second check catches what is left.
• Spaced reviews get skipped in busy weeks: Each review is about 5 minutes and tied to a date. If one is missed, it is done the next day, not dropped.
• Feedback turns into praise: Anika and the buddies name the measurement and the cause ("you measured before adjusting the frame"), never the person.
• On-the-job practice becomes shadowing: The buddy uses the same 6-point check and observation checklist as induction, and the assistant's log lists what was new each day.
• The trainer drifts back to talking: Active sessions can feel worse to learners. Hold the line by checking the skill a week later, not by reading the feedback forms.

Building the online part with an AI course tool

On 26 October 2026 I gave the Coursebox course agent a brief for the online part of the dispensing programme. Coursebox is our product, so weigh my account with that in mind. The brief asked for a course that was mostly practice: three worked examples and then two that fade, four orders with a planted error to find, six questions from memory, four dated reviews, a remake case and a role play. It gave the 6-point check in its exact words and order, and asked the agent to mention the induction days, buddy shifts and sign-off, but not to teach measuring through reading pages.

Before planning anything, the agent stopped and asked for the facts it would otherwise have invented. It asked how to handle the missing prescription examples, lens types, coatings and adaptation wording, and what return timing the role play should teach, adding: "I should not invent a clinical or business rule." I told it to use fictional practice prescriptions labelled as such, and gave it the lens types, coatings and the group's adaptation wording. It then laid out an outline of 10 practice pages and 5 lesson pages, with the wrap-up last.

Most of what I checked came back as briefed. The worked examples faded: the fourth told the learner to "Complete steps 5 and 6" and then revealed a "Correct answer and why", and the fifth was built by the learner alone. The in-person parts stayed in person: "Measuring and frame adjusting are in-person skills practised during the Townsville induction with Anika and during buddy shifts; they are not tasks to perform from this page." The four reviews were separate pages, and the two I opened each began with their day, such as "This review is for day 2". The role play was set up as an AI Interactive Scenario that asks learners to "Make decisions across five turns to practise acknowledgement, frame fit, approved wording, return agreement and symptom referral."

The problem I fixed was in the faded example, which learners study as a model. Example 4 had pre-filled step 2 as "Lens choice: Progressive." The customer card listed "Selected coatings: Scratch-resistant and blue-light filter", and point 2 requires both lens type and coatings to match the customer's choice. The half-worked example had omitted the coatings.

In the page editor I changed step 2 to "Lens choice: Progressive, with scratch-resistant and blue-light filter coatings." After I reloaded the learner view, example 4 showed the corrected step, with steps 5 and 6 still left for the learner.

I checked the outline and a sample of pages: examples 1 and 4, the planted-error page introduction, two reviews and the role-play setup. I did not play the role play through. One content defect remained: the examples used frame model, size and colour as the three frame details, while the supplied check required lens width, bridge and temple length. The missing temple-length entry was not fixed. Example 4 also revealed a model answer without checking the learner's selections, so it was a self-check, not an automatically scored assessment. The coating correction did not establish that the full order check was modelled correctly. Anika still assigns the dated reviews; induction, buddy shifts and sign-off remain in person. No learning or dispensing outcome was measured.

Putting adult learning methods to work

The methods that work for adults at work are the ones where people do the task and find out how they did. The best-evidenced of them help learners of different ages, so pick the method from the goal, not from the learner's age or preferred style. Lectures have a place, but keep them short and follow them with recall and practice, and judge the result by a check of the skill rather than by how the session felt.

Your next step: take one course and identify what people need to know and do. Map its explanations, attempts, feedback and assessment to those goals. Where a required task is only described, add a suitable attempt; keep explanations and reference material that help people perform it.