Generative AI is accelerating content creation, but Dr Ravinder Tulsiani argues that speed and volume are not the same as value. For L&D, the real opportunity is to move beyond mere output and focus on capability, judgement and measurable performance change, especially where organisations need stronger human decision-making most.
It’s a moment many of us will recognise: A well-meaning stakeholder asks for a new learning module. The problem is described vaguely, inconsistent performance, missed decisions, something “not landing”. The solution, already half-decided, is training. Perhaps now, with AI involved, it will be faster, smarter, more personalised.
An uncomfortable question lingers: Will this actually change anything that matters?
And yet, somewhere between the request and the build, an uncomfortable question lingers: Will this actually change anything that matters? For years, L&D has lived with this quiet tension. We have been busy, productive, and increasingly sophisticated, while never quite sure whether our work was shifting the outcomes the organisation truly cared about.
The arrival of generative AI hasn’t created this problem, it has simply made it impossible to ignore.
When content stops being proof of value
Generative AI can now do, in minutes, what used to take learning teams weeks:
- Draft course scripts and storyboards
- Generate quizzes, scenarios, and role plays
- Summarise policies into microlearning
- Create personalised learning pathways
- Translate and localise content at scale
This is undeniably useful. But it also forces a reckoning. If a general-purpose model can produce 70–80% of traditional L&D outputs on demand, then content volume, speed, and even polish can no longer serve as credible evidence of value.
The uncomfortable question for our profession is not “will AI replace L&D?”, but “if we can produce ten times more learning assets, will the organisation perform ten times better?”.
If the honest answer is “probably not”, then the issue was never AI. The issue was what we chose to measure, reward, and defend as value in the first place.
The risk of hallucinating our own impact
Much has been written about AI “hallucinations”. Less attention has been paid to a more subtle risk: L&D hallucinating its own impact. For decades, learning functions have been evaluated on what is visible and countable:
- Courses launched
- Completions
- Hours consumed
- Content libraries built
- Engagement dashboards glowing reassuringly green
These signals were always proxies. Imperfect, but tolerated. AI makes the weakness of these proxies impossible to ignore. When content becomes cheap and instantaneous, output can no longer stand in for impact. Velocity no longer implies value.
And yet, many L&D teams feel pressure to demonstrate relevance by “AI-enabling” the learning stack:
- AI search in the LMS
- Auto-generated learning paths
- Coaching bots
- Skills inference engines
Some of these are genuinely helpful. The risk lies in what they can obscure. Dashboards look better. Activity increases. Stakeholders relax. The function appears modern and responsive. But the harder questions remain unanswered:
- Are managers having better performance conversations?
- Are decisions improving in messy, high-stakes contexts?
- Are people reaching competence faster in roles that matter?
- Is the organisation executing strategy more effectively?
If we cannot answer these with confidence, AI has not made L&D more strategic. It has simply made the illusion of value more convincing.
Most organisations don’t have a content problem
One of the most persistent myths in corporate learning is that performance gaps are primarily information gaps. Decades of research on learning transfer tell a different story. What happens after training, opportunity to apply, manager support, incentives, work design, feedback, has far more influence on performance than the quality of the content itself.
Most organisations are not starved of information. They are starved of:
- Practice in realistic conditions
- Feedback on judgement, not just knowledge
- Reinforcement in the flow of work
- Permission and support to apply learning under pressure
When L&D uses AI to accelerate content production without addressing these conditions, the likely result is not better performance, but more noise. More assets. More choice. The same friction where learning meets reality.
In this scenario, the learning function can look increasingly productive while the organisation itself becomes more fragile.
The human capabilities AI does not commoditise
To understand where L&D can still create defensible value, we need to be honest about what AI is not good at. AI excels at scaling information, drafts, and options. It is far weaker at the operational capabilities that determine whether strategy succeeds or quietly fails in the middle of the organisation:
- Judgement under uncertainty, when no prompt contains the full context
- Navigating trade-offs across competing stakeholder demands
- Ethical reasoning and accountability: Deciding what should be done, not just what can be done
- Leadership courage in moments where rules, data, and precedent run out
- Trust-building: The relational capital that enables execution
These are often labelled “soft skills”, which does them a disservice. They are not soft at all. They are the capabilities that keep organisations coherent under pressure. When these capabilities are weak, organisations become brittle: highly informed, poorly prepared.
If L&D remains centred on content production, albeit faster with AI, it risks strengthening the wrong muscles.
From learning supply to capability stewardship
What is required is not a rebrand, but a shift in operating model. The most useful way I have found to frame this shift is from learning supply to capability stewardship.
- A supplier provides content
- A steward owns outcomes
Capability stewardship means being willing to sit with the business and own a small number of difficult questions:
- Which capabilities will most strongly determine performance over the next three to five years?
- What evidence would convince us those capabilities are actually strengthening?
- Where is performance breaking down for reasons training alone cannot fix: workflow, decision rights, incentives, manager habits?
- How can AI remove friction, so people spend more time in high-effort practice rather than passive consumption?
These are not easy questions. But they are the questions senior leaders are already asking, whether L&D is in the room or not.
What this looks like in real L&D work
For practitioners, capability stewardship does not mean abandoning learning design. It means redirecting it.
Practice before catalogues
High-impact capability is built through practice, feedback, and reflection, particularly in complex, ambiguous situations. Instead of using AI as an answer engine, it can be used as a sparring partner:
- Role-playing difficult conversations
- Pressure-testing decisions against counterarguments
- Exploring realistic scenario variations
The goal is not perfect answers, but better judgement.
Managers as the learning environment
If transfer is the problem, managers are part of the solution. Rather than overwhelming them with more leadership content, L&D can equip managers with:
- Observation guides
- Coaching prompts
- Clear descriptions of “good enough” performance
- Simple practice routines embedded in team meetings
This is often where learning strategies succeed or fail, not in the LMS, but in day-to-day management behaviour.
Learning to supervise AI
As AI becomes embedded in work, a new human capability emerges: AI supervision. People need support to:
- Detect unreliable outputs
- Validate recommendations against context and policy
- Escalate risks
- Remain accountable for decisions
This is a significant, underexplored opportunity for L&D to add value beyond content.
Measuring what leaders actually recognise
If L&D wants to be taken seriously in the age of AI, it must stop measuring what is easy and start measuring what leaders already care about. A defensible scorecard does not need to be perfect. It needs to be credible. Examples include:
- Time-to-competence in critical roles
- Changes in manager effectiveness (quality of coaching and performance conversations)
- Reductions in error, rework, or quality incidents linked to decision-making
- Bench readiness and internal mobility
The aim is not to claim sole attribution, but responsible stewardship: a clear line of sight between learning effort and organisational movement.
A quieter but important risk
Alongside the rush to adopt AI tools sits another, quieter illusion: that selecting the “right” platform will future-proof learning. Markets consolidate. Vendors disappear. Regulation tightens. Budgets fluctuate. Resilience does not come from tools. It comes from clarity.
L&D functions anchored on clear human capabilities, portable data models, and internal judgement can move across tools without losing themselves. Those anchored on platforms are one procurement decision away from irrelevance.
The invitation
The age of AI is forcing a choice. L&D can double down on appearance, more content, more features, more “AI-powered” labels, or it can use this moment to tell the truth:
- Information access is not capability
- Content output is not performance change
- Automation does not reduce the need for judgement; it increases its value
For practitioners, this is not a threat. It is an invitation, to step out of the role of course builder and into the role of capability steward. To stop defending activity and start owning outcomes. To help organisations become not just more efficient, but more resilient.
The learning functions that thrive will not be those that produce the most assets, but those that can clearly name, and measurably strengthen, the human capabilities no model can replace.
Dr. Ravinder Tulsiani is Principal, Interim Enterprise Capability & L&D Advisor at Catalyst Learning
