Agentic AI in L&D: How to Move from Pilots to Trusted Workflows

AI has moved beyond conference buzz and isolated chatbot experiments. Both our State of Learning Technologies 2026 report and McKinsey's latest AI study show the same pattern: AI use is broadening, but scaling it remains difficult. In our research, AI has become an operational reality for enterprise L&D, with topics like integration, security, and trustworthy governance now acting as the real bottlenecks. McKinsey reaches a similar conclusion at enterprise level: adoption is expanding fast, but most organisations are still in experimentation or pilot mode and have not yet embedded AI deeply enough into workflows to create genuine enterprise-wide value. 

One of the clearest findings in our report is that the nature of the challenge surrounding artificial intelligence has shifted. L&D is no longer asking whether AI belongs in learning. It is instead asking how to connect it, govern it, and prove that it works. More than half of respondents cite integrating AI or new learning technologies effectively as their biggest challenge. At the same time, the learning stack is consolidating around a central LMS backbone, while investment moves toward AI-powered authoring, coaching, analytics, and skills infrastructure. 

McKinsey defines AI agents as systems that can plan and execute multiple steps in a workflow. Their survey shows strong interest, but also a reality check: 62 percent of respondents say their organisations are at least experimenting with AI agents, while 23 percent say they are scaling an agentic AI system somewhere in the enterprise. Even then, scaled use in any single function remains limited.

For L&D, that makes agentic AI interesting for a practical reason: It points to a move away from one-off prompting and towards workflow-based support. That can mean helping teams create content faster, recommend relevant resources, support learners in context, identify skill gaps sooner, or guide practice and feedback more dynamically. In our report, respondents describe the next game-changer not simply as “AI,” but as practical support that helps people learn while working, together with better skills visibility and more adaptive systems. One of the clearest signals in the report is the phrase: “Agentic AI, not just chatbots.” 

This is where the broader McKinsey findings are especially useful. Their strongest message is not just that AI is spreading. It is that organisations seeing the greatest value are more likely to redesign workflows, pursue innovation alongside efficiency, and embed AI into how work actually gets done. That matters for L&D because AI creates the most value when it is not treated as a bolt-on feature, but as part of a broader redesign of content operations, performance support, skills development, and learner guidance.

Our own research supports a similar takeaway. Integration into existing systems, strong training and communication for users, and clear KPIs are all among the most important factors for successful implementation. Engagement is strongest when learning is embedded into daily workflows, not separated from them. That is the real promise of agentic AI for L&D: less friction when it comes to fostering learning within an organisation. 

Agentic AI will not scale in learning because it sounds advanced. It will scale if people trust it. In our report, AI adoption is increasingly tied to questions of data security, reliability, privacy, compliance, and human oversight. The report is explicit: AI adoption is accelerating, but trust and governance will decide who can scale.

The older whitepaper still adds an important perspective here, showing how many organisations hesitate because of legacy systems, lack of expertise, resistance to change, and concerns around confidentiality and compliance. It also argues that strategic AI implementation requires investment in infrastructure, skills, and change management, and not just enthusiasm for the technology itself. Finally, governance shouldn't be viewed as a barrier to adoption. It is part of what makes adoption possible.

The most straightforward advantages of AI in L&D still concern speed: Faster content creation, lower admin load, quicker updates, and better scalability all matter. The whitepaper captures this clearly, especially around AI-enabled content creation, automated LMS processes, and scalable learning delivery. 

But the bigger opportunity also involves capability building. Our 2026 report shows that measurement is moving from activity to outcomes, with productivity, skills improvement, and skills gap analysis becoming more central. It also shows that AI readiness and automation readiness are now the top skills priorities for L&D. In other words, the next phase of AI in learning is about more than just doing existing work faster, but also helping people and organisations to become more capable, adaptable, and measurable over time.

The practical takeaway remains that agentic AI will create only value in L&D when it is tied to real workflows, trusted by users, connected to skills and performance, and positioned as more than a pure efficiency play. For learning leaders, that means focusing less on AI as a feature set and more on the conditions that make it useful at scale.