Training Departments Are Not Sitting Still
Corporate learning has spent years being treated like the office instruction manual, and AI is making that old arrangement look awfully flimsy. RedThread Research's November 19, 2026 study lands on a point that ought to make senior management pay attention: learning and development leaders are not merely waiting around to be handed an AI training assignment. Many are trying to remake their function while the rest of the company is still arguing over who owns the AI strategy.
That is the sensible response, frankly. A vague order to “train everybody on AI” is management's favorite kind of half-baked command: urgent in tone, empty in detail, and usually delivered without budget or a definition of success. RedThread says C-suite leaders broadly want AI adoption, yet “AI fluency” often has no settled meaning. Meanwhile, strategy tends to live with the CIO, a transformation office, or a Chief AI Officer. L&D gets summoned late and told to clean up the human part. Familiar mess.
Forty-Four Moves, Not One Magic Playbook
The study draws on in-depth interviews, research roundtables, and a review of 50 sources. It organizes the findings across five areas: direction and vision, development experience, operations and governance, technology and data, and stakeholder alignment. RedThread cataloged 44 actions, with 22 requiring little or nothing in place before a team can begin.
That 22-of-44 figure is the part worth circling. It cuts against the lazy excuse that an organization must first purchase some giant platform, finish a pristine data project, or wait for corporate AI policy to descend from the mountain. Some work can begin with judgment, visibility, and a willingness to quit acting like a request desk.
- Set direction early: work through how AI will alter work and learning before a formal company plan arrives.
- Get into AI planning: show up where decisions are made rather than hoping for an invitation.
- Raise the bar: avoid treating stakeholder demand as the ceiling for quality.
- Design for judgment: deliberately protect expertise that AI could weaken through overreliance.
- Automate intake: route requests intelligently and challenge whether training is even the right answer.
None of that is glamorous. Good. Most useful organizational work is not glamorous; it is the unsexy discipline of asking better questions before money, time, and employee attention get poured into another forgettable course library.
The Real Prize Is Better Judgment
RedThread's examples are practical rather than theatrical. One talent leader supplied managers with an AI prompt pack structured around a 30/60/90-day framework to support new hires. Another global learning leader saw business teams already creating courses with AI. Her response was not to publish guidelines and pray. The team is building an agent intended to guide non-L&D employees through outcome definition, measurement, and learning-design practices.
That second example carries the sharper warning. Once generative tools make course creation cheap, organizations will produce a lot more learning material. Cheap production is not the same thing as sound learning. It can mean a mountain of polished nonsense delivered at high speed. The function that adds value will be the one that helps people define outcomes, test output, measure whether anything changed, and recognize when the machine is bluffing.
“If L&D's value is the stuff AI can copy now, the function has a problem.”
Dani Johnson, RedThread's co-founder and principal analyst, puts it bluntly, and she is right to. This is not a case for protecting old job descriptions. It is a case for moving up the chain from content production toward design, governance, advisory work, and hard-nosed evaluation.
What Sector Watchers Should Take From It
For investors watching enterprise software and human-capital technology, the implication is mildly uncomfortable for vendors: customers are increasingly building around systems they do not own and cannot wait on vendors to perfect. That does not erase demand for platforms. It does mean the winning products will need to fit into messy real workflows, support oversight, and help users verify outputs rather than merely generate them.
The study is not a forecast of revenue, adoption rates, or productivity gains, and it should not be stretched into one. It is qualitative research, not a claim that every company has its act together. Still, the signal is clear enough. L&D teams that claim a seat in AI decision-making may become far more strategic. The ones that remain course factories are staring at a very unpleasant identity crisis.