Enterprise L&D teams keep recycling the same instructional design methods from a decade ago. ADDIE templates. Generic needs analyses. Click-through modules with a quiz bolted on the end.

Then they sit in a review meeting wondering why the capability gap hasn't closed.

I've watched this pattern repeat across lots of organisations. The methods aren't ancient history, they're still being taught, still being certified, still shaping how teams plan their year. And they're still failing to move the metrics that matter.

This isn't a nostalgia problem. It's a diagnosis problem. Most instructional design methods were built to answer a different question than the one enterprises are actually asking.

Why do instructional design methods keep failing?

They fail because they optimise for content production speed, not behaviour change. A course gets built, published, and rated highly by learners. Nobody performs any differently three months later.

That's the gap nobody wants to name in the steering committee meeting.

Traditional frameworks treat "analysis" as a box-ticking exercise before the real work — building — begins. Teams rush the diagnosis to get to production. However, the diagnosis is the real work. Skip it, and you're designing a solution to a problem you never actually confirmed exists.

I see this constantly: a stakeholder says "we need training on X," and the instructional designer starts storyboarding within the week. Nobody asked whether the gap is a skill gap, a motivation gap, or a broken process that no course can fix.

The result is training that satisfies stakeholders — glossy, on-brand, well-reviewed — but does nothing for the business metric it was supposed to move. That's not education. That's theatre with a completion certificate.

Is fast AI content the answer, or is that the wrong question?

Faster content is not the fix — it's the same failure at higher speed. Enterprises now face a false choice: slow, rigorous frameworks versus fast, AI-generated modules. Neither addresses the actual weak point, which sits in analysis and evaluation, not build.

AI has genuinely changed production speed. A module that took three weeks now takes three days. That's real, and I won't pretend otherwise.

But speed in the build phase doesn't fix a design that was never grounded in a validated performance gap. You just get shallow content faster, at scale, with a nicer interface.

The industry conversation around 2026 is starting to catch up to this. Commentary from the eLearning Academy has framed it plainly: AI dominated instructional design in 2025, but the real question is whether it actually changed outcomes — or just changed how fast we produce the same mediocre thing.

That's the honest version of the debate most vendors won't say out loud.

The fix isn't choosing between rigorous-and-slow or fast-and-shallow. The fix is rethinking what happens before you ever open an authoring tool — and what happens after the course launches.

What does a better instructional design approach look like?

A better approach starts with a tight performance-gap diagnosis, designs explicitly for transfer, and builds evaluation in from day one — not as an afterthought once the launch date has passed.

Three shifts separate teams who move metrics from teams who produce content:

Start with the gap, not the request. Before any design work begins, confirm the gap is actually a capability gap. Interview performers. Look at the data. If it's a process or incentive problem, no course fixes it. Design for transfer, not completion. Learning that doesn't survive contact with the job is wasted effort. Build in leader reinforcement — practice prompts, follow-up conversations, real work applications — as part of the design, not a nice-to-have appendix. Build evaluation in from day one. Decide what "success" looks like — a specific business metric — before you write a single learning objective. The Kirkpatrick Model has argued for decades that evaluation should be planned at the start of design, not measured after the fact as an afterthought. Most teams still ignore this.

This isn't a radical framework. It's a discipline most teams abandoned under delivery pressure.

Do instructional design methods need replacing, or just used properly?

They don't need replacing. They need teams willing to question whether the framework fits the actual business problem before they start building.

ADDIE isn't broken. Needs analysis isn't broken. What's broken is the habit of running through the motions without asking whether this specific problem needs this specific method.

Furthermore, the rush to adopt AI tools has made this worse, not better. Speed has become a proxy for progress. Meanwhile, the actual diagnostic work — the unglamorous, slower part — gets skipped entirely.

If your instructional design methods are producing content but not results, the framework isn't your problem. Your discipline is.

Three things to do this week:

  • Audit your last three courses against one metric: did performance actually change, or did learners just rate it well?
  • Before your next project brief, force a written performance-gap diagnosis — no build work starts until it's signed off.
  • Define your evaluation metric on day one of design, not after launch.

If your L&D function is producing content but not results, it's time to rebuild the approach — from Discovery through Ongoing Refinement. Explore how at https://calebfoster.ai.