The new AI in L&D report from Donald H Taylor and Eglė Vinauskaitė, The Transformation Triangle, makes an argument that lands hard if your job is building learning. Producing content can no longer be the reason an L&D function exists. AI generates it fast, cheap, and often well enough. Business units make their own. The thing L&D was organised around for decades has stopped being scarce.
That’s an uncomfortable read for a company that builds custom content for a living. We’ve made our peace with it. We think the report is right.
Content is cheap now, so the job has to change
The report’s core move is to separate what L&D does from what L&D is organised around. Most teams are still organised around content: a request comes in, a programme goes out, success gets measured in completion rates and feedback scores. That logic was built for a slower world.
The functions pulling ahead organise around a question the business cares about instead: where is performance about to break, and what capability gap is behind it. They spot the gap forming rather than waiting for someone to report it. The report describes three ways to do this: treating skills as a tracked business asset, surfacing the expertise already sitting in your workforce, or fixing the conditions around performance rather than just the people. Different routes, one shared idea. When content is cheap, the value moves to reading what the business needs before it asks.
We’ve been making a version of this argument ourselves. Faster content won’t move L&D out of the cost center. The “cost center” label sticks because L&D usually reports activity (courses built, hours delivered, satisfaction scores) while the business buys outcomes. Producing those courses faster doesn’t change what L&D is measured on. Changing what L&D does (from producing content to spotting performance problems before they cost the business) is what moves it off the cost line.
You can’t do the new work while production owns your week
Most teams want to read those signals and sit in performance conversations. They can’t, because content production still takes all their time.
The report calls the forces that hold L&D in place “drags,” and a lot of it is mechanical. Teams stitch together a text tool, a video tool, and a translation platform that were never built to work together. People become the glue between them. Every handoff is a place where work gets lost, duplicated, or checked again by hand. When production is that heavy, there’s no capacity left for the higher-value work, and the team stays busy making things.
So the question isn’t whether to abandon content. It’s whether your content engine runs cleanly enough to free your best people for the work that changes how the business sees you.
Content becomes part of the system, not the output
Content doesn’t go away. Its role changes, from the thing you deliver to one part of a larger system.
That only works if the production layer underneath is solid. Most teams don’t have a clean one. Content moves through tools that don’t connect, gets reformatted by hand at every step, and arrives without anyone able to say where a given line came from. A reviewer either trusts it or checks the whole thing against the source again.
A connected setup removes that overhead. The same source produces every format you need, review happens inside the workflow rather than after it, and each AI-generated asset traces back to the material it drew from, so a reviewer confirms the link instead of reconstructing the context. The reviewing job shrinks to the parts that need real judgement. The hours that used to go into reformatting and re-verifying go back to your instructional designers and SMEs, who now have time for the higher-value work the rest of this depends on.
Your experts hold knowledge no course captures
One of the report’s three routes maps closely to work we already do. It calls it the Enablement Partner: organising L&D around the expertise already in your workforce, then surfacing it and moving it to where it’s needed.
Most of what makes someone good at their job never makes it into a course. It stays in the head of the senior person who knows why one approach wins and another stalls. A generic AI tool can’t reach that knowledge, because it was trained on the internet, not on how your organisation works. Connect AI to your own material instead (your manuals, your SOPs, your approved documentation) and tag each source so every output can be traced back to it. Now the content it produces sounds like your organisation rather than a generic version of it. The same setup that makes content yours also pulls expert knowledge out of people’s heads and makes it reusable. For a lot of organisations, that matters more than the content itself.
It’s a systems problem: fix the setup around the work, not just the work
The report makes one point we’d underline. Changing L&D isn’t mainly about mindset or trying harder. It’s a systems problem. The L&D functions that pull this off look at their own organisation honestly, pick the direction that fits it, and put the structure in place to hold that direction steady.
That matches what we keep seeing. Adding another AI tool to a broken setup doesn’t fix the setup. Standardise before you automate. Put validation where quality actually breaks, not at the end. Measure the quality of the finished work and whether it changed anything, not how fast it’s been produced. Bring IT into governance early instead of letting shadow tools create risk you’ll meet later in a compliance review. None of it is glamorous, but it’s the difference between a setup you can scale and one that breaks the moment you run more than a few projects through it.
The report ends on a useful question. If you left tomorrow, would the function keep running on its new “organising logic,” or drift back to its old “content-focused identity” and start producing content on request again? The report’s point is that when the change lives in one person and doesn’t get built into how the team works, the old way reasserts itself the moment that person leaves. Building it in is the harder job.
