Agentic AI at Work: What L&D Teams Need to Know Now

Not long ago, "AI at work" meant a chatbot answering HR FAQs or a tool that suggested your next compliance course. That was helpful, but it was passive. It was merely AI that waited to be asked.

That era is ending fast...

A new category of AI has entered the chat: agentic AI at work. These are systems that don't just respond to prompts. They set goals, make decisions, take actions, and complete multi-step tasks largely on their own. They can book meetings, write and send emails, run code, pull data from multiple systems, and hand off work to other AI agents, all with minimal human input.

This isn't a future prediction. Deloitte projected that 1 in 4 companies already using generative AI would launch agentic AI pilots or proofs of concept in 2025 alone, with adoption expected to reach 50% by 2027. And Microsoft's 2025 Work Trend Index describes this as the dawn of the "Frontier Firm"- organizations redesigning work around human-plus-AI teams, with 81% of leaders expecting AI agents to be integrated into their strategy within the next 12-18 months.

For Learning & Development (L&D) teams, this shift raises urgent questions: What do employees actually need to know to use agentic AI at work? What new skills does this demand? And how should L&D change its approach in order to keep up?

Let's dig in.

What Is Agentic AI, Exactly?

Before you can train employees to work with agentic AI, it helps to understand what makes it different from the AI tools most workers have used so far.

Generative AI (think ChatGPT, Copilot, Gemini) produces outputs such as text, images and summaries in response to human prompts. You ask, it answers. The human remains in the driver's seat at every step.

Agentic AI goes further. These systems can:

• Define and pursue goals across multiple steps without constant human direction
• Use tools (web browsers, code executors, APIs, databases) to take real-world actions
• Coordinate with other AI agents to complete complex, interdependent tasks
• Adapt their approach based on what they learn along the way

A simple example: instead of asking an AI to "draft a project update email," an agentic system might automatically pull the latest project data from your PM tool, summarize progress against goals, draft the update, cross-reference the stakeholder list, and send it. All this would be triggered by a calendar event, with no human touching it along the way.

Harvard Business Review describes AI agents as "fast becoming digital teammates- an emerging category of talent" rather than just tools. That framing matters enormously for L&D, because teammates require a fundamentally different kind of training than software does.

The Gap Between Enthusiasm and Readiness

The uncomfortable reality is employees are excited about agentic AI at work, but organizations aren't ready to support them. A recent EY survey of more than 1,100 U.S. desk workers found that 84% of employees are eager to embrace agentic AI in their roles, anticipating positive impacts on productivity, efficiency, and work experience. But beneath that enthusiasm, there's real anxiety:

• 54% of employees feel they are falling behind their peers in agentic AI use at work
• 61% feel overwhelmed by the constant influx of new agentic AI information
• 56% worry about their own job security working alongside AI agents
• And 73% of employees remain unaware of how AI agents will actually impact their work- despite their employers already planning to deploy them (Salesforce, 2025)

That last number is striking. Nearly three-quarters of employees don't have a clear picture of what's coming while their organizations are actively moving forward. That's not just a communication problem. It's a training problem, and it lands squarely on L&D's plate.

Meanwhile, the scale of change ahead is significant. According to the World Economic Forum's Future of Jobs Report 2025, nearly 39% of current skillsets will be overhauled or become outdated between 2025 and 2030. And McKinsey reports that by 2030, 59% of the world's workforce will require training or retraining at a scale that has no real precedent in modern L&D.

The message is clear: L&D teams that treat agentic AI at work as a minor update to existing AI training are underestimating the moment.

What Employees Actually Need to Learn

So what, specifically, does agentic AI demand from your workforce? The honest answer is: more than just tech skills.

1. AI Fluency- Not Just AI Literacy

AI literacy is the ability to understand what AI is and recognize where it's being used. It was the right starting point for generative AI, but agentic AI requires something deeper: AI fluency, the ability to actively work with, direct, evaluate, and improve AI-driven workflows.

McKinsey's analysis suggests that by 2025, the percentage of work requiring "no or minimal support" from Generative AI would drop to just 10%, while workflows requiring "moderate to significant support" will surge to 56%. If most work will involve meaningful AI collaboration, employees can no longer afford to be passive users.

Training for AI fluency means teaching employees to:

• Break complex goals into tasks that can be delegated to AI agents
• Evaluate AI-generated outputs critically, and know when to override them
• Craft clear, specific instructions (prompts) that get useful autonomous actions, not just responses
• Understand the boundaries of what an AI agent can and cannot reliably do in their context

2. Human Oversight and Judgment

This is the big one, and it's often skipped in AI training programs that focus on capabilities rather than responsibilities.

Agentic AI operates with significant autonomy, which means employees need to know when and how to intervene. An agent executing a multi-step workflow might make a decision in step three that seems reasonable in isolation, but creates downstream problems you would have caught if you were watching closely. Employees need training in:

• Setting appropriate guardrails and decision checkpoints before deploying an agent on a task
• Recognizing the warning signs that an AI agent is going off-course
• Understanding how to audit what an AI agent did and why
• Knowing when a task genuinely requires human judgment and confidently reclaiming it

EY's research confirms that generational differences are already creating friction here. Gen Z people managers, for instance, show more hesitancy than Baby Boomers when adopting agentic AI tools introduced by their organizations. Rather than assuming younger workers are naturally AI-ready, L&D teams need to address the very different anxieties and confidence gaps that exist across generations.

3. Managing Human-AI Teams

Managers have a particularly new challenge ahead: leading teams where some members are human and some are AI agents. Salesforce's HR research found that 80% of CHROs believe that within five years, most workforces will have humans and AI agents working together. And yet most management training programs have zero content on this.

McKinsey puts it plainly: "Organizations should prepare managers to lead teams that include both humans and AI agents." That requires building new skills around:

• Allocating work appropriately between human and AI team members
• Communicating AI-augmented decisions to human stakeholders transparently
• Building and maintaining team trust in environments where AI agents are involved
• Keeping human employees engaged and valued when significant portions of routine work shift to agents

4. Ethical Reasoning and Accountability

Agentic AI removes humans from many decision points, which raises real questions about who is accountable when something goes wrong. This isn't a theoretical concern. When an AI agent acts on a customer's data, makes a business decision, or communicates externally on the organization's behalf, that carries real consequences.

L&D teams need to incorporate ethics into agentic AI training and not as a compliance checkbox, but as a genuine skill. Employees need to understand:

• What decisions should never be delegated to an AI agent
• How to recognize and flag potential bias or errors in AI-generated outputs
• What organizational policies govern agentic AI use, and what the employee's responsibility is within those policies
• How data privacy and security apply when AI agents are accessing systems on an employee's behalf

5. Role-Specific Application Skills

Broad AI literacy training is a starting point, not a finish line. According to the EY research, the feeling of falling behind is significantly more pronounced among non-people managers (65%) than among people managers (48%), likely because front-line workers often don't have a clear picture of what agentic AI at work means for their specific job.

Effective training has to connect agentic AI to the real workflows of real roles. A marketing coordinator, an operations analyst, a customer service rep, and a supply chain manager will each encounter agentic AI in different ways. L&D teams need to build role-specific scenarios that show employees how agents will interact with their actual tools, their actual tasks, and their actual decisions, and not just hypothetical examples.

👉Learn more: The New Literacy: Training Your Employees for AI Critical Evaluation

What This Means for How L&D Operates

Here's where it gets interesting: agentic AI at work doesn't just change what L&D teams teach. It's also changing how L&D teams work.

Agentic learning assistants can now track course completion, send personalized nudges to learners, provide real-time feedback, and adapt learning paths 24 hours a day, at a scale no L&D team alone can match. This is a genuine opportunity to close the personalization gap that has frustrated L&D practitioners for years.

But it also means L&D professionals need to develop their own fluency with agentic AI. Teams that are still manually administering learning programs, building static courses on 18-month development cycles, and measuring success primarily through completion rates are going to struggle to keep pace.

Smart L&D teams are already shifting toward:

• Skills-based learning architectures that can adapt in real time as AI reshapes role requirements
• Just-in-time training embedded in the workflow, not delivered in a separate course environment
• Scenario-based simulations that give employees practice working alongside AI agents before they encounter them in high-stakes situations
• Shorter, more frequent learning interventions that can be updated rapidly as agentic AI tools evolve- because waiting two years to update a course is no longer an option

McKinsey's guidance is direct: L&D must become "part of the engine of organizational performance, adaptability, and resilience- not just a support function." This is a significant elevation of the L&D role- and it requires L&D professionals to make a compelling case for their seat at the strategic table.

👉Discover More: eLearning for Communication and Customer Service Skills

Practical First Steps for L&D Teams

The scale of this shift can feel overwhelming. Here's a grounded starting point.

Start with a skills audit, not a course catalog. Before building any training, find out what your employees actually know, what they're uncertain about, and what specific agentic AI tools or workflows they'll be encountering in the near term. This audit should be role-by-role; the needs of a finance analyst and a frontline service rep will look very different.

Prioritize awareness and psychology alongside skills. More than half of employees in the EY survey simultaneously worry about job security even as they express enthusiasm for agentic AI. Training that ignores the emotional dimension of this transition will fall flat. People need space to ask hard questions about what AI means for their career, and L&D can lead that conversation.

Build manager training first. Managers are the amplifiers of organizational change. Millennial people managers in particular report high anxiety about managing increasingly AI-reliant teams, but also show high eagerness to learn, with 88% saying most of their agentic AI knowledge is self-taught. Giving managers better tools and frameworks before they have to lead their teams through AI integration makes every downstream training effort more effective.

Use simulation, not just explanation. The most effective way to build confidence with agentic AI is practice in safe environments. Scenario-based simulations, where employees can try delegating tasks to an AI agent, evaluate its outputs, and practice course-correcting, will outperform slide decks and videos every time.

Plan for continuous learning. McKinsey's research found that 75% of U.S. workers expect their roles to shift due to AI in the next five years, but only 45% have received recent upskilling. The gap between expectation and preparation is real. L&D teams need to build learning systems designed for ongoing iteration, not one-time rollouts.

👉Learn More: Upskilling and Reskilling: Strategic Frameworks for Workforce Transformation

The Bottom Line

Agentic AI isn't a new feature in an existing tool. It's a fundamental change in how work gets done, and that demands a fundamental change in how L&D teams think about training.

The good news is employees want this. 86% report that working with AI agents has had a positive impact on team productivity, and 90% of those already using agentic AI express confidence in their abilities. The appetite is there but what's missing is the structure, the clarity, and the training to support it.

L&D teams that act now in building the skills, the programs, and their own internal fluency to lead this transition have a rare opportunity to move from the back seat to the front. The organizations that thrive in the agentic era will be the ones that didn't wait for the future to arrive before they started preparing for it.

The moment is now.