L&D Team of One: The 6-System Stack That Replaces the Department You Don't Have

If your title says "Learning & Development Manager" and your team is just you, this playbook is for you.

L&D at a 5,000-person company has an instructional designer, a video editor, an LMS administrator, a curriculum strategist, and somebody whose entire job is reporting completion rates to a VP. L&D at a 120-person company has you. The work is the same. The math is brutal.

The CEO wants "a training program" by Q3. HR bought an LMS eighteen months ago that nobody has logged into since the rollout. Operations wants onboarding fixed. Sales wants product training. Compliance wants annual refreshers tracked. You have a budget that won't cover an instructional designer for one quarter, let alone the year. And every piece of L&D content you read online is written by someone running a department of twelve at a company with 8,000 employees.

I have spent twenty years in this craft. I have a Master's in Adult and Continuing Education from Rutgers. I was Senior Manager of L&D at Horizon BCBS's Service Division, running learning for thousands of employees. Before that, I was Training & Development Manager at TMP Direct, where I built an enterprise LMS from zero, produced fifty-plus training videos, ran a 70-agent contact center program, and supported training across more than thirty national accounts. I know what the enterprise machine looks like, and I know what it looks like when the machine is just one person with a laptop and a deadline.

This is the playbook I would hand my younger self the day he became a team of one.

Why most L&D advice fails the team of one

The structural problem is simple. The big publishers — Training Industry, ATD, CLO — write for buyers at enterprise. The vendors price for enterprise. The conference circuit speaks to enterprise. When a CLO who hasn't designed a course in ten years writes a thought-leadership piece on "engaging the modern learner," they are not writing it for you. They are writing it for somebody approving a six-figure Cornerstone renewal.

The actual job for the team of one is different in three concrete ways.

It is less production and more curation. You will not out-produce LinkedIn Learning or a vendor's stock library. You will out-curate them, because you actually know the work your people do.

It is less platform and more system. The LMS is a delivery mechanism. The system is what feeds the LMS. If the system is good, the platform is interchangeable. If the system is bad, no platform will save you.

It is less measurement-of-training and more measurement-of-behavior-change. The enterprise L&D team optimizes for completion rates because that is what their dashboards reward. The team of one cannot afford to measure anything that does not change how someone does their job on Monday.

Conventional L&D wisdom is built for teams of twelve. You are a team of one. Translate accordingly.

The 6 systems you actually need

Stop thinking in tools. Tools are interchangeable. Think in systems. A system is a repeatable way to make a decision, do the work, and know whether it worked. Here are the six you need. Build them in this order.

1. The needs analysis system

The single biggest failure mode for a team of one is building courses nobody asked for. Or worse: building the courses the CEO did ask for, when the CEO is wrong about what the company actually needs.

Backward Design, from Wiggins and McTighe's Understanding by Design (ASCD, 2005), starts with Stage 1: identify the desired results. Not the desired training. The desired results. The team-of-one version of Stage 1 is a five-question rubric I run on every request that lands in my inbox:

1. What behavior should change? Name it as a verb a manager could observe on a Tuesday.
2. Who is the actual learner? Not the title — the role, the tenure, the prior knowledge.
3. What is the gap? Is this a skill gap, a knowledge gap, a motivation gap, or a process gap? Training only solves two of those.
4. What does success look like in 90 days? Quantify it. If you cannot quantify it, you cannot evaluate it.
5. Is training the right intervention? If the answer is "we need a job aid, a process change, or a manager conversation," say so out loud before you build a course.

If you cannot answer those five questions, you do not have a training project. You have a request that needs a conversation. Run the conversation. Half the requests die at question five, and that is a win.

At Horizon a VP once walked into my office and told me he needed an e-learning program built for the Service Division. CSAT was sliding and his read was that the reps needed retraining. I asked the five questions anyway — the ones that boring instructional designers always ask before they build anything. By question three it was clear the gap wasn't skill. It was that the reps' QA scorecards rewarded average handle time over resolution quality, and a recent system change had buried the most-used policy lookup behind three extra clicks. Training the reps harder on a metric we were punishing them for hitting was not going to work.

So I deferred the training and recommended fixing the scorecard and the policy-lookup path instead. The VP was not happy. I landed in some real hot water for not giving him the thing he asked for, and I had to defend the call to people above him. The call was right — the data came in two months later and CSAT recovered without a single training session — but the immediate cost was real.

As the L&D of one, this is the trap. You have the latitude to do the right needs analysis and refuse to build the wrong thing. You also pay the political price when you do it. The five-question rubric is the discipline. Knowing you'll sometimes be the only person in the room making the unpopular call is the job.

2. The design system

Backward Design is the framework. You do not need an instructional designer to use it. You need to use it the way it was designed.

Stage 1: Identify the desired results. (You did this in needs analysis.)

Stage 2: Determine acceptable evidence. Before you write a single slide, decide how a learner will demonstrate they got it. Write the assessment first. Yes, first. If the assessment is a multiple-choice quiz, you are probably training the wrong thing — most workplace performance does not look like multiple-choice. Scenarios, role-plays, work samples, and supervisor checklists are the evidence formats that map to actual behavior.

Stage 3: Plan the learning experience. Now, and only now, do you design the content. The content exists to produce the evidence. If a module does not move the learner toward the evidence, it does not belong.

This is the order. Most team-of-one L&D builds in the wrong order — content first, assessment last, outcomes never. That is why the courses get built, get launched, get completed, and change nothing.

ADDIE still works as a project-management spine: Analyze, Design, Develop, Implement, Evaluate. SAM (Successive Approximation Model) works when the stakeholder cannot tell you what they want until they see a prototype, which is more often than they admit. Use ADDIE when the scope is clear. Use SAM when it is not. Both are fine. Backward Design sits inside either one as the design logic.

3. The production system

This is where AI changes the math, and it is the only place AI changes the math. The production system for a team of one in 2026 looks like this:

• A generator that takes your outcomes and produces a course outline.
• A generator that takes the outline and drafts lessons.
• A generator that takes the lessons and drafts assessments aligned to the outcomes.
• A template library — visual, structural, voice — that you reuse across every build.
• A recording set-up that is one camera, one mic, one light, and one quiet room. Stop trying to film in a green-screen studio you do not own.

You write the outcomes. AI drafts the body. You edit. The edit is where your craft lives. This is not a workflow that replaces an instructional designer — it is a workflow that lets one well-trained instructional designer do the production work of three.

What to skip: animated character explainers, voice-over actors, custom motion graphics, anything that takes more than a day per finished hour of content. Your audience is not Pixar's audience. They are adult professionals who need to do their jobs better on Monday. Plain video, clear slides, well-edited script, accurate captions. Done.

4. The delivery system

HR bought an LMS. You did not pick it. You will probably not get to replace it. So the delivery system is a portability system.

Build everything to export as SCORM 1.2 and SCORM 2004. Build it to export to xAPI when the LMS supports it. These are not optional. SCORM 1.2 is the lowest common denominator — almost every LMS speaks it. SCORM 2004 handles sequencing and richer reporting. xAPI (Tin Can) reports learning events from outside the LMS, which matters when the actual learning happens in a CRM, a call-center platform, or a field app rather than inside the course player.

For most SMBs, xAPI is overkill until it isn't. The trigger is when learning happens outside the LMS and you need to track it — a sales rep practicing in a call coaching tool, a tech doing a procedure in a checklist app. Until that is the real use case, SCORM 1.2 + 2004 is enough.

The discipline is this: never build a course that lives only inside the LMS HR picked. The course must export. The course must port. If the LMS gets replaced in eighteen months — and it will — the library survives.

5. The evaluation system

Donald Kirkpatrick's four levels (Kirkpatrick Partners, refined by James and Wendy Kirkpatrick in the New World Kirkpatrick Model) are still the right frame fifty years later. Level 1: Reaction. Level 2: Learning. Level 3: Behavior. Level 4: Results.

Here is the honest team-of-one prescription, which contradicts most of what gets posted on LinkedIn:

Do L1 and L2 for everything. A two-question reaction survey (relevance + confidence) and a real L2 assessment built from your Backward Design Stage 2. These cost you almost nothing and tell you whether the design worked at all.

Do L3 sparingly and well. Behavior change measurement is expensive. It requires manager involvement, observation windows, and a baseline. Pick one or two flagship programs a year where you do L3 properly — supervisor checklists at 30, 60, 90 days — and do not pretend to do it on every course. Pretending is worse than not doing it.

Do L4 only when the CEO actually asks. Business-impact attribution for a training program is one of the hardest measurement problems in the discipline. If you have not nailed L1 and L2, you have no business claiming L4 results. Most L&D teams of one waste enormous time pretending to measure L3 and L4 when L1 is a smiley-face survey and L2 does not exist. Fix that first.

6. The coaching and reinforcement system

Training without reinforcement decays in thirty days. This is the most-cited and least-acted-on finding in the field. Harry and Rosemary Wong's work on the first days of school, extended into workplace induction, makes the same point in a different vocabulary: the structure of what happens after the formal teaching is what determines whether the learning sticks.

The 90-day post-training cadence is cheap, manual, and works. After the course closes:

• Day 7: A single email to the learner. One question: "What is one thing you have used this week?"
• Day 30: A manager check-in prompt. You write a one-page guide for the manager — three questions to ask, two behaviors to observe. The manager runs the conversation. You do not need to be in the room.
• Day 60: A peer-share micro-prompt. One Slack message, one video, one example.
• Day 90: The L3 evidence collection — supervisor checklist, work sample, or short interview.

This costs you maybe two hours of design per program and ten minutes per learner over ninety days. It is the single highest-leverage thing a team of one can build. Most L&D programs end at "course completed." Yours should end at "behavior observed at day 90."

What AI actually changes about L&D craft

AI changes the production cost of course content. That is the change. It is a real change and it is large. A workflow that took me four weeks at TMP Direct in 2014 — outline, script, slide deck, assessment, recording, edit — now takes me four days, and the edit is where I spend most of those four days, which is where the craft was always supposed to live.

AI does not change methodology. Backward Design is still right. ADDIE is still useful as a project frame. SAM is still useful for rapid prototyping. Kirkpatrick is still the evaluation frame. The Wong and Wong literature on induction and reinforcement is still the strongest case in the field for the 90-day window. None of that is replaced by a generator. A generator that writes a lesson without an outcome to anchor it produces noise that looks like content. The methodology is what stops that.

The L&D pro who treats AI as a craftsmanship multiplier wins. Outcomes first, generator second, edit third, evaluation fourth. The L&D pro who treats AI as a replacement for the methodology loses, because they ship faster and worse and the courses do not change behavior, and within two cycles the CEO stops asking for the training program.

I built LearningByDesign around this premise — AI-augmented production sitting on top of Backward Design, ADDIE, and Kirkpatrick, priced for the team of one and the SMB consultant rather than for the enterprise procurement cycle.

The consultant version of this

If you serve multiple SMB clients as an independent L&D or training consultant, the six systems still apply. What changes is leverage.

Your template library is the moat. Every course you build for one client becomes a structural template for the next. If you are not building a reusable template library, you are billing hourly for production that you will redo from scratch on the next engagement. That is the worst business model in this discipline.

White-label matters more for you than it does for an internal team of one. Your client should see their brand on the course, in the player, on the certificate. Anything else looks unserious.

Billable hours per finished hour of course content must drop. The enterprise instructional design benchmark is 100-200 hours of build per finished hour of content. For an SMB consultant in 2026, an AI-augmented workflow on a strong template library should land at 20-40 hours per finished hour for standard content. If you are still at 80, you are leaving margin on the table and pricing yourself out of the SMB market.

SCORM export must work for twelve different LMS targets. You will not get to pick the LMS. Your portability is your product. If your build only works in one platform, you have built a liability, not a library.

The consultant audience is the most underserved segment in this market right now, and most of the tooling is priced and designed for either solo creators selling Udemy courses or enterprise teams replacing Cornerstone. The middle — a consultant serving five to twenty SMB clients — is where the work is and where the tools are weakest.

The 5 traps L&D-of-one falls into

I have made each of these mistakes. So has every team-of-one I have ever talked to.

One: Building courses nobody asked for. You see a gap, you build a course, nobody takes it. Run the five-question rubric or do not build.

Two: Spending 80% of your time on production and 5% on needs analysis. Production is the visible work. Needs analysis is the leverage. Invert the ratio. The teams that get this right look slower from the outside and ship better outcomes.

Three: Buying or accepting a fancy LMS the company will not use. Adoption is the only LMS metric that matters. The cheapest LMS your people will actually log into beats the most beautiful LMS they will not.

Four: Skipping evaluation because Kirkpatrick is "hard." It is not hard at L1 and L2. It is only hard at L3 and L4, and you do not have to do L3 and L4 on every course. Do L1 and L2 on everything, properly, this quarter. That alone puts you ahead of most L&D shops at any size.

Five: Treating the job as "make training" instead of "make behavior change." This is the trap that swallows a career. If your dashboard is course completions, you will optimize for the wrong thing for a decade. If your dashboard is behaviors observed at day 90, you will get harder questions, fewer projects, and dramatically better outcomes.

Closing

The L&D team of one is not a failure mode. It is the new default for everyone under 200 people, and increasingly for divisions inside companies much larger than that. The enterprise L&D department is being unbundled — partly by budget pressure, partly by AI, partly by the simple fact that most learning that matters in an SMB is closer to the work than a centralized team can ever be.

The play is to pick the methodology that fits the constraint, use AI where it raises craftsmanship and refuse it where it lowers craftsmanship, and stay honest about the difference between training people and changing how they work. Backward Design tells you what to build. ADDIE or SAM gets you through the project. Kirkpatrick tells you whether it worked. AI helps you ship it. The 90-day cadence makes it stick.

That is the stack. Six systems. One person. It is enough.

If you want the tooling layer that fits this stack — outcomes-first authoring, SCORM and xAPI export, evaluation templates built on Kirkpatrick, and a price tag that respects the team-of-one budget — that is what I built at TranscendByDesign. Either way, build the systems. The tools will follow.