Here we go again. A powerful new technology arrives, and learning and development’s (L&D’s) first question is often: “How can this help us build training faster?”

I understand the instinct. Artificial intelligence (AI) can draft content, create questions, summarize documents and build course assets at remarkable speed. Those capabilities are useful. But if all we do is use AI to create more training, faster, we will have missed its most important contribution.

The better question is: How can AI help people perform better while they are doing the work?

The Role of AI in Performance Support

AI can improve how information is found, filtered, adapted, maintained and delivered. It can make a digital coach more conversational, respond to the performer’s role and situation, help diagnose problems, surface relevant examples and guide increasingly complex decisions.

Dr. Conrad Gottfredson and I have described AI not as a replacement for performance support, but as something that can accelerate workflow analysis and strengthen digital coaching. In many ways, this brings us closer to Gloria Gery’s original electronic performance support system (EPSS) vision. Her vision was not a static job aid. It was an integrated environment capable of supporting complex work, accepting input, offering situational guidance, connecting resources and enabling increasingly self-reliant performance.

Generative AI, intelligent agents, advanced search, analytics and multimodal interfaces make those capabilities far more achievable than they were in 1991. But — and this is an important “but” — an AI chatbot is not automatically a digital coach.

A Digital Coach Needs to Understand the Work

A general-purpose model may produce an impressive answer without understanding how work is actually performed in your organization. It may not know the approved process, the performer’s role, the current policy, the exception path or the consequence of being wrong. It may answer the question while missing the performance need underneath it.

A digital coach must be grounded in the workflow. That means it needs more than content. It needs a structured understanding of workflows, job tasks, steps, decisions, supporting knowledge, resources, dependencies, role differences, common problems and the critical impact of failure.

Without that structure, AI may make access faster while leaving relevance, accuracy and actionability to chance. This is why Rapid Workflow Analysis and, what Gottfredson and I are now describing as task intelligence, are both so important.

AI Should Start With the Workflow

AI adoption should not begin with the tool. It should begin with understanding the work at the job-task level. Once that work is visible, an organization can make better decisions about which tasks should be automated, which should be augmented by AI and which require human expertise, accountability, empathy, creativity or judgment.

The 5 Moments of Need gives us another useful lens:

• At New, AI can create personalized explanations, examples and practice tied to the work a person will perform.
• At More, it can provide additional depth based on role, experience and emerging needs.
• At Apply, an AI-enabled digital coach can clarify a step, locate an example, adapt guidance or surface the precise supporting knowledge required.
• At Solve, AI can help compare symptoms, walk through diagnostic logic, retrieve lessons learned and determine when to escalate.
• At Change, it can identify affected tasks, help update resources, explain what changed and reinforce the new way of working when an old habit reappears.

That is an amazing opportunity. It is also where methodology protects us from technology-first thinking. The purpose is not to put AI everywhere. It is to create the right balance of automation, augmentation, learning, support and human performance.

The Instructor and Classroom Still Matter

So, where does this leave the instructor? In a far more important role than “the person who knows the content.” When information is instantly available, the instructor’s value moves toward the deeply human dimensions of performance: helping people practice, make judgments, recognize context, question an AI recommendation, understand consequences, collaborate, reflect and recover from mistakes. The instructor becomes a performance coach.

The classroom changes again as well. Physical and virtual classrooms become performance laboratories, places to rehearse high-risk work, test decisions, work through ambiguity, evaluate AI output, receive feedback and experience productive failure before the consequences are real. The digital coach and AI should be present because learners must practice performing with the tools they will use on the job.

This does not make instructors, classrooms or training obsolete. It puts each in its proper place. The evolution is not classroom to eLearning to digital coach to AI, with each technology replacing the one before it. It is the evolution from an event-first model to an ecosystem designed around performance:

• Targeted training develops what must be internalized.
• The instructor builds judgment and confidence.
• The classroom provides practice and feedback.
• The digital coach supports performance in the workflow.
• AI makes that support more adaptive, scalable and intelligent.

Gery gave us the vision. The 5 Moments of Need gave us the framework. EnABLE gave us the design discipline. The digital coach gave the vision a practical home in the workflow. AI can now supercharge all of it, but only if we continue to put performance first.

Where Do We Go From Here?

1. Stop asking, “Where can we use AI?” Start asking, “Where does performance need intelligence?”

Pick a critical workflow and identify the tasks where people struggle to find information, make decisions, diagnose problems or adapt to change. Those are the places to explore AI. Not just the places where AI happens to be available.

2. Give AI a workflow to work with.

Before building an AI assistant or digital coach, make the work visible. Map the critical tasks, decisions, resources, standards, exceptions, roles and consequences of failure. AI without workflow context is just a very fast generalist. Give it task intelligence, and it can become a performance partner.

3. Experiment with one task, and decide what AI should not do.

Take one high-value task and deliberately determine: What should be automated? What should be augmented? What should remain human? Then prototype the support and put it in the hands of performers. The goal isn’t to prove that AI can do something. The goal is to prove that it can improve performance.