Leave it to a theater producer to hit the nail on the head with a fitting title. As the ATD Forum delved into another member roundtable discussion, Elliott Masie used the words “seductive” and “scary” to describe the artificial intelligence (AI) phenomenon as we advance in the evolving world of organizational learning. Along with co-presenter Bob Gerard of Accenture, he shared ideas, experiences, and suggestions.
The presenters emphasized the need for a deeper understanding of generative AI’s impact and encouraged an ethical and creative approach to its use. The discussion highlighted the importance of connecting with other Forum members interested in the topic and sharing relevant resources to foster a collaborative and informed community on the subject. They also discussed the transformation of education models. This discussion focused on Bloom’s research on the 2 Sigma Problem related to the effectiveness of one-to-one learning experiences over one-to-many—and the possibility that generative AI could scale the effectiveness of the one-to-one into a one-to-many.
The conversation delved into the phenomenon of AI washing in HR and learning tech, especially with vendors, cautioning us to be aware of exaggerated claims and emphasizing the importance of authenticity and scrutiny when evaluating AI-powered solutions. The presenters shared experiences with vendors making misleading and exaggerated claims and highlighted the need to focus on performance needs rather than simply adopting AI for the sake of it. Additionally, they predicted a shift in the use of the term AI in the future, drawing parallels to the evolution of web technologies. They also emphasized the importance of language alignment across learning and HR organizations in order to work more effectively with the technologists.
Masie and Gerard also discussed the practical steps for leveraging AI in learning and development, emphasizing the need for a mindset that thinks about augmentation and experimentation, but with a healthy dose of caution, both for security and in vendor selection. Protected rights, attributions and sources, data currency, tagging, and security remain issues of concern. We must also recognize the ethics of AI usage—and incorporate ethics into the organizational model.
One takeaway from the discussion was that the learning function needs a seat at the table to collaboratively make decisions about generative AI usage within our organizations. The big decisions cannot be driven just by technologists. And like it or not, employees are off and running with it.
Another takeaway was to focus on performance—and how to use the technology to augment tasks related to performance, enabling better and faster performance. Some areas to consider for AI application include skill assessments, data collection and analysis, using predictive data for task assignments, and personalizing learning, especially when the learner is in the zone between confidence and competence. AI could also help shift the moment of learning. Currently, most learning occurs either too early or too late for practical application. Generative AI options can enable learning at the optimal time. Still another arena for potential use is onboarding.
One suggestion was to become fluent in the language of generative AI and savvy enough to push back on the marketing spin, especially when the technology’s promises might be exaggerated. Be able to stop a pitch to ask deep questions about specifics related to learning and performance.
Some suggestions for those just beginning to use generative AI:
- Experiment—play with it, but do not lock in your strategy.
- Focus on performance and tasks to help augment your current work.
- Start official use with a pilot—and learn from it, then make iterative improvements.
In general, Masie is currently considering these ideas for using generative AI capabilities in the learning arena:
- Interactive Learning Tutor or Coach
- Personalization: Learning Agenda for One
- Assessment Engine: Laddered and Diagnostic
- Workflow Support: Nudging by Demand
- Simulations via AI: “Failing Safely” to Success
- Predictive Data for Assigning Learning and Tasks
- Generative Content, Context, and Stories
- Linking Learners for Collaboration
- Tracking Outcomes and Career Impacts
- AI for UX Design: Adapting to Differences
- Graphics, Infographics, and Job Aides for One
- Globalization of Learning: Language, Culture, and Multi-Track
A recent Talent Development Leader article shared this advice:
“As your comfort with AI tools grows, it’s important to approach AI implementation thoughtfully, with a clear strategy and a focus on ethical considerations and data privacy. Define specific goals and objectives you want to achieve with AI in your training programs, such as improving learner engagement, increasing training efficiency, or enhancing personalization.
“Begin with a small, manageable pilot project (such as a specific use case or module) to test the AI solution’s feasibility and effectiveness. It may be easier to adapt an existing program than to start from scratch.
“Scale gradually. As you gain confidence in AI’s effectiveness, consider expanding its use in other aspects of your learning programs. Remember that AI is a tool to enhance your work, not a replacement for human expertise. The combination of AI-driven efficiency and human guidance can lead to more effective and engaging learning experiences in your organization.”
Some thought leaders mentioned during the session include Sal Khan and John Spencer, who work with K–12 groups, and Ethan Mollick, an associate professor at the Wharton School at the University of Pennsylvania.
Prior to the virtual discussion, the ATD Forum surveyed members. One of the interesting findings was that many organizations do not have an Employee Use Policy, so while experimentation is encouraged, without guardrails, the organization could be at risk.
The list of generative AI tools currently used by attendees included Open AI, Gemini, You, Sal Khan, ChatGPT, MS CoPilot, Perplexity, and Adobe Firefly.
Yes, generative AI is seductive, and it is sometimes scary. As learning leaders, we need to proceed with caution, but we do need to proceed.
Author note: Special thanks to some contributions from Read Assist.ai in drafting this article.
