One of the best things AI has done for Learning & Development is make practice exciting again.

Prefer to listen? The podcast version of this article is here.

I saw it firsthand at a recent L&D conference. Practice was everywhere. AI roleplays, simulations, avatars and scenarios. Vendors and learning leaders were talking about giving people the opportunity to actually do something, rather than simply consume more content.

After 25 years of building Practica Learning around Deliberate Practice, we think this is a very good thing. But we also think we need to be careful, because we have been here before.

When eLearning emerged, it gave organizations something they desperately wanted: the ability to deliver learning consistently, quickly and at scale. It was a huge advancement. But somewhere along the way, scale became the learning strategy.

Companies accepted one-size-fits-all courses because that was the model vendors could efficiently deliver. Thousands of employees could receive exactly the same content, regardless of their experience, role, customers, culture or individual development needs. The economics were compelling, and financial pressures reinforced the approach, along with the appealing assumption that if you teach or tell people what to do, they will do it.

But knowing what to do and being able to do it are very different things. We see the consequences in poor feedback about learning programs, onboarding that can take months to translate into performance, and bored learners clicking through content on autopilot.

The economics worked beautifully. The learning didn't always work as well.

AI gives us the chance to do something different

We have always known that people develop skills faster when learning is relevant to them. They need to work on the behaviours they personally find difficult, in situations that resemble the ones they actually face. They need to try, receive meaningful feedback, adjust and try again. That is the fundamental idea behind Deliberate Practice.

AI gives us an extraordinary opportunity to make that kind of practice available to far more people, more often and at a much lower marginal cost. But are we already making the same mistake? Instead of generic eLearning courses, we are beginning to see generic AI scenarios.

From the learner perspective, the experience tracks like this:

1. Choose "difficult conversation."
2. Choose "coaching."
3. Choose "sales."
4. Talk to the avatar.
5. Get a score.
6. Receive some feedback.

This simple approach to AI practice scales wonderfully. But whose conversation is it?

• Does the scenario reflect your organization's culture?
• Does it use your language, conversation models and expectations?
• Does the customer behave like your actual customers?
• Does it mirror differences between your markets and regions?
• Does a sales conversation incorporate your go-to-market strategy, the objections your people actually encounter, or your risk and regulatory framework?

Most importantly, does it address the specific gaps you're trying to close?

Practice without clearly defined learning objectives is just activity.

Don't let the vendor's business model become your learning strategy

This is where I see a familiar problem emerging. Customization is expensive or discouraged in some large-scale learning models because customization interferes with the economics of standardization. The most profitable model for a technology provider is often to build something once and sell it many times.

There is nothing inherently wrong with that business model. But it shouldn't determine how your people learn.

If customizing an AI scenario to reflect an organization's actual customers, culture and strategy becomes prohibitively expensive, the natural response is to accept the generic scenario. That is exactly how we ended up with one-size-fits-all eLearning.

There is even a faint echo here of Cory Doctorow's idea of "enshittification": platforms can begin by creating significant value for users, then gradually optimize around the economics of the platform itself. We should watch for that possibility in AI-enabled learning.

Because context isn't a nice-to-have.

Context is where much of the learning lives.

And in many organizations, that context is also a competitive advantage. Your culture is different. Your customers are different. Your strategy is different. Your competitors are different. The way you want a salesperson to handle an objection, a manager to coach an employee or a leader to communicate strategy should reflect those differences.

Why would you train your people to have the same conversations as everyone else?

Start with your organization, not the AI

At Practica, we believe the starting point should be the learner and the behaviour we are trying to change, regardless of whether the practice is ultimately delivered by AI or a human Roleplayer Coach. We recommend creating two distinct practice environments.

The first focuses on client-facing conversations. Build it around your organization's culture, strategy, go-to-market approach, customers, best practices and aspirations. Define the specific skills and behaviours that represent an effective conversation in your organization.

Those behaviours become a Skills List. Every practice interaction is measured against that same Skills List, creating consistency across scenarios while allowing the practice itself to be individualized.

Then create multiple customized scenarios based on real interactions: customer objections, negotiations, market situations, relationship challenges and the conversations your customer-facing people actually encounter.

The second environment focuses on internal conversations. Again, start with the organization's culture and expectations. Create a separate Skills List for conversations such as peer-to-peer interactions, coaching, leadership development, communicating strategy and difficult conversations.

With these two environments clearly addressed, your AI practice stops being an isolated technology experience and becomes part of an organizational learning system.

You can measure the same behaviours across people, scenarios, teams, regions and time. You can identify patterns. You can see where individuals need help and where entire groups may be struggling.

Most importantly, you can provide people with practice that feels like their work because it is their work.

And sometimes AI shouldn't be the end of the practice

There is another important distinction. Human conversations are emotional experiences.

A difficult conversation with an employee, a negotiation with a customer or a coaching conversation with a struggling team member isn't simply a sequence of correct and incorrect responses.

There is tension. Status. Trust. Empathy. Curiosity. Frustration. Confidence. Uncertainty. Effective practice needs to recognize those dynamics.

Feedback needs to go beyond "consider asking more open-ended questions." It should help learners understand the effect their behaviour had on the other person, what they missed and what they could try differently, and then give them the opportunity to try it again.

This is where we believe AI and human practice become particularly powerful together. At Practica, an AI scenario might provide a seven- to ten-minute practice interaction that can be repeated whenever the learner needs it.

That same scenario can then be expanded into a 30- to 60-minute Deliberate Practice session with a highly trained, experienced human Roleplayer Coach. The conversation can become more nuanced. The Roleplayer Coach can respond to subtle changes in behaviour, emotion, tone and body language. They can stop the interaction, coach the participant, rewind the conversation and let them try a different approach.

And because the human and AI practice are built around the same Skills List, they aren't two disconnected learning experiences. They are part of the same development system.

That deeper level of human investment can also be targeted where it creates the greatest value: senior leaders, high-potential employees, people struggling with particular skills, or customer-facing individuals responsible for significant relationships and revenue.

AI provides scale and repetition. Human Roleplayer Coaches provide depth and nuance. The learning architecture connects them.

The technology isn't the methodology

For 25 years, Practica Learning has helped people improve important workplace conversations through individualized Deliberate Practice. For much of that time, part of our job was explaining why practice mattered. We don't have to explain nearly as much anymore. AI has helped resurrect practice as a serious part of the learning offering, and we couldn't be happier about that.

Now let's not repeat what happened with eLearning. Don't start with the technology. Instead, start with:

• the behaviour you want to change
• the skills you want to develop and measure
• the situations your people actually face
• your customers, your strategy, your culture and your learning objectives.

Then decide how AI and human expertise can best support the practice.

Don't let someone else's technology determine how your people learn. Start with your people and what they need to learn.

‍