What Makes AI Training Stick? A Learning Design Framework From a Former Chief Learning Officer

Austin's first pillar comes down to one idea: people need a judgment-free place to learn, and a group to learn alongside.

That starts with a corporate sandbox. Teams need a "corporate sandbox and safe places where they can do this," Austin said. One example would be guided practice built in 30-minute daily increments, so learnings and conversations can build on each other.

For teams that want to build this shared practice, see our team AI training.

The sandbox only works if it fosters a learning community. In Austin's view, even people who'd rather learn alone still need to return to a group to see how others apply what they learned.

Some of that pull is simply social. People respond to "a little bit of peer pressure, a little bit of competitiveness," Austin said, and a healthy dose of both keeps a cohort moving instead of stalling out after week one.

The clearest proof came from a small-business owner cohort Austin facilitated. Participants arrived afraid of AI, running their own small businesses with little background in the tools at all.

By the end, what they credited for getting them through wasn't just the curriculum; it was having peers "to bounce ideas off, and to demonstrate, too, to encourage and support their learning."

"You're gonna get the behavior you pay attention to," said Austin. Groups where sharing your AI experiments is normal produce more of that behavior. Individual learners or groups left alone don't.

This is also part of the key idea on AI fluency: skills only compound when people keep practicing together after the course ends.

Training alone fails, Austin said, if "nothing has changed in the environment" around it. That’s the gap between adoption and AI transformation: training changes what people know, while transformation changes how the work gets done.

She breaks the fix into three parts:

• Leadership messaging: Leaders have to be seen using AI, not just talking about it. "Maybe they have an agent that creates charts and summaries for them that they use in a meeting," Austin offered as an example. Visible use solidifies the message.
• Direct communication: That visibility only lands if it's paired with conversations about what AI means for people's jobs. People might quietly worry that trying something new puts their role at risk. Communicate directly about how these tools can empower them and how AI integration and innovation support the individual performer and the mission of the organization.
• Process redesign: Austin said doing process redesign at the team level first is "one of the most important ways you can empower the team members to not feel afraid." People end up contributing to the conversations around building the change. This collaboration welcomes people to AI integration and helps them overcome the resistance and anxiety of feeling that AI is being done “to them”.

None of this holds if leaders are asked to carry it without support. Here’s Austin's warning: organizations tell frontline leaders to support a rollout, but they might not give them the proper resources to do it.

Give leaders governance support, budget, and visibility, not just a mandate to make it happen.

The resourcing model Austin points to is cascaded learning. Frontline employees get baseline training plus reference material, such as "e-manuals and electronic support."

Leaders get a separate track built around what leading the change actually requires: how to model it, how to talk about it, and a peer group of other leaders to compare notes with. It's close to what a formal AI champion program model looks like: the people responsible for sustaining the change get enhanced support to “champion” it across their organization.

The last resource is structural: build touchpoints into what is being developed. A place to share prompts and assistants people have built, a fast approval process, and AI as a standing item in staff meetings and comms, not a one-time announcement.

Austin's framework reminds us that the fundamentals of how people learn and change together just need intentionality to be useful in the age of AI. Give people a real sandbox and a community to practice in, and the skills compound on their own.

The same is true of change management. It succeeds or fails based on whether leaders are willing to be seen learning in public. A leader who shares their own AI experiment, imperfect prompts, and so on, drives adoption more than any policy memo ever could.

Build the sandbox, support the community, and pay attention to the right things, and the training will stick.

To see how this framework played out against a real department with reservations about AI and jobs, read the companion case study: how Harvard's School of Public Health piloted AI fluency, with Kirstin Austin's fellow presenter, Bryn Austin.

Kirstin Austin is an Impact Partner at Lead with AI and a retired federal government executive who most recently served as Chief Learning Officer at the IRS. This article is based on her July 31, 2026 Lead with AI PRO session, "Designed to Stick," co-presented with Bryn Austin.