Agentic AI: Moving From Information to Action

As part of their quarterly sessions hosting thought leaders in AI, the ATD Forum recently hosted a roundtable discussion with Debbie Richards, AI strategy advisor, president of creative interactive ideas, and frequent conference presenter, on the future of AI in learning and talent development, focusing on Agentic AI. Before this engaging and informative discussion with more than 80 attendees, the Forum members curated information on this topic from both an internal survey and a user case from a member during an AI Cohort meeting.

The Three Types of AI

According to Richards, AI can be grouped into three categories:

• Informational AI finds and summarizes content based on our command, such as when Google Gemini receives an instruction or command.

• Generative AI augments our personal work and creates new content based on an instructional prompt, like ChatGPT drafting a course outline.

• Agentic AI or the Project Coordinator goes further and can be more strategic. It can understand a goal, create a multiple-step plan, and carry out actions across various systems.

One example made the capabilities of agentic AI very clear. If you want to attend a concert, agentic AI can check your calendar, search ticket sites, make purchases, and block the time on your schedule. For L&D, the possibilities are even more expansive and meaningful. For example, an AI agent could onboard a new employee by pulling information from the LMS, HRIS, and calendar, and adjust as needed.

Pulsing the participants on their thoughts related to using agentic AI, some of the comments included: exciting and scary all at once; optimistic and skeptical; cautiously excited. One even stated that so far, it is creating more work for me because of the learning curve, but hopefully, that will change.

Why Agentic AI Matters for L&D

Richards explained that one of the most valuable uses of agentic AI is in building custom copilots for teams. These copilots can automate repetitive tasks, assist instructional designers with measurable objectives, and support debriefs by analyzing chat transcripts, polls, and feedback for common themes. She outlined two core capabilities: multi-step planning and problem solving, which include adapting when conflicts occur, learning from outcomes, and refining the process.

As I listened, I considered the potential impact this could have for L&D leaders. The real value lies in saving time and directly linking AI to business outcomes. If an agent shortens the cycle time for developing measurable objectives or boosts new-hire productivity, the outcome is a tangible impact that leaders will observe. In my own client work, I have seen how much time is wasted when instructional designers struggle with learning objectives. An agent designed for this purpose could drastically improve both efficiency and quality. The same idea applies to custom upskilling programs, onboarding, and process improvements that enhance organizational performance.

Guardrails and Blueprints

The excitement around AI raises concerns. In a poll, the audience identified data privacy and maintaining the human element as their top priorities. To address this, Richards shared a blueprint for building agents responsibly. The process involves clarifying metrics, users, and goals; defining the role of the agent; mapping workflows; specifying knowledge sources such as policies or documents; connecting the right tools like calendars, LMS, or integrations; and asking guiding questions about what the agent needs to succeed.

She also reminded us to apply the “human, AI, human” philosophy. A human generates the idea, AI enhances it, and a human validates the outcome. L&D leaders need to take this further by prioritizing ethical adoption. Data privacy is essential, but so are issues of bias, accessibility, and inclusivity. Employees should understand how AI is used in their learning journey, and transparency will be key to building trust.

Practical First Steps

Richards encouraged L&D teams to start small. Experiment with a GPT, Gemini, or Copilot. Test it with clear business problems and run it through your Intellectual Property department and security guardrails before expanding. She suggested several use cases, such as automating evaluations, summarizing meetings, or supporting skill gap analysis.

This sage advice highlights the importance of starting with high-impact entry points. Use agentic AI to gather feedback from multiple cohorts or create learning objectives that your team can improve. These small successes boost confidence, decrease resistance, and demonstrate the technology’s value to stakeholders.

The Cost of Falling Behind

Richards closed with a reminder that top talent will gravitate toward organizations that provide modern and personalized tools. That message resonated with me. Personalized learning agents can recommend courses, map career paths, and create simulations for practice. This is not only about efficiency, but also about shaping the employee experience and positioning L&D as a true partner in talent strategy.

A big takeaway is that agentic AI is not a replacement for human expertise. It is a multiplier of it. The future of learning is not just about information or content creation. It is about action. Agentic AI allows us to create ecosystems where humans and AI work together, tackling problems and fostering growth in ways we could not achieve alone.

Another takeaway from this peer-learning session was the extensive sharing that takes place when TD leaders from various organizations interact with a thought leader around a topic of interest. The Forum members asked great questions and provided helpful hints and suggestions, including ways they were using agentic AI, various apps they found helpful, and resources. All of these helped the members move from information to action in their learning journey.