Copilot Catalyst Lesson 3: Teach People to Fish, Don't Give Them Cool Tips
There’s an unofficial rule of IKEA furniture assembly: everything goes fine until step five. That’s the moment you realize you thought you followed the instructions perfectly and repeated every step exactly as outlined, but somehow the shelf still managed to end up backwards and upside down.
“What? How…? Aghhh!”
Frustrating as this situation is, my rule is always the same: shut my eyes, take a deep breath, and retrace my steps taking apart my hard-won progress until I find the misstep. I then redo it exactly as the instructions say and inevitably, I end up with the correct result.
AI works differently. The phenomenon is called stochasticity. You can feed it the exact same input twice and get two very different outputs. That isn’t a bug. That ‘unpredictability’ is a feature. It’s where the creativity and insightfulness of the model actually come from. The skill users need to employ is in understanding how AI operates and knowing how to adjust when you don’t get the output you thought you were going to get.
In this case, the IKEA trick of going back and repeating a predefined set of steps is not likely going to get you the end result you are looking for.
Tips and Tricks Don’t Scale
raditional software rewards people for memorizing steps. Click here. Navigate to this menu. Use this command. If the instruction set is stable, memorizing the steps works. But generative AI runs on an inherently less predictable model. It is not like any technology we have used before. It evolves constantly, and its outputs vary on purpose. It breaks all the rules of traditional software training.
This is why we spend so much time up front teaching our clients not just what AI is, but what it is not.
You cannot pull the "best" answer out of AI by following a rigid sequence. In fact, AI is not designed to give you the best answer at all. It gives you the most likely answer based on probability and statistics. In some cases when there is a lot of uncertainty in the question you can get some really varied answers.
You only learn to pull better answers by understanding how the model works, how it reasons, and how it responds when you adjust your approach.
But here's the thing: when you're building a Copilot rollout, the temptation to include specific tips is almost irresistible. People want concrete answers. They want the "right way" to do things. When you factor in how quickly changes are being made to Copilot and other AI models, building around example prompts and how-to guides becomes precarious.
Early on, we built around it too, not quite realizing the potential quicksand we were stepping into.
This was exemplified (to my embarrassment) in one of our first training programs, where I felt compelled to include an eight-tip “best practice” guide for Copilot use. Several of the tips were built around the fact that Copilot inside the Microsoft apps and inside Teams had noticeably different interfaces than the web version. I was very proud of this particular insight.
So, I carefully crafted a set of recommendations explaining when and where to use certain features, where you could find them, and why a few of them behaved inconsistently depending on the location. It was a great slide. It would have been insightful, nuanced, and genuinely helpful if Microsoft had not, two days before the presentation, completely revamped and updated all the in-app Copilot UIs silently in the middle of the night. Fortunately, we caught the change in time and avoided the confusion, but it could have derailed a critical first working session and undermined confidence in the entire system.
But if you are not teaching features, what are you teaching? You are teaching people how to fish.
Resilience Compounds. Features Expire.
We wanted people to understand what they could do with AI, not just what the tool could do for them. That meant teaching them how to learn, adapt, and think with this new technology. Yes, we taught them prompting techniques and demonstrated functionality in each application. But our real focus became a small set of skills that do not go out of date when the interface changes.
Exploration. The ability to try different approaches, test variations, and push the boundaries of what the tool can do.
Error recovery. The ability to respond when the answer is off by clarifying, reframing, or breaking the task down.
Judgment. The ability to know when AI is the right tool, when it is not, and how to evaluate its outputs.
Peer-learning. The ability to share patterns and discoveries so the whole team gets smarter over time.
Those skills compound. Someone who understands how to explore, recover, and evaluate will always outrun someone who memorizes a list of prompts. And we saw this immediately. After three weeks in our Copilot Catalyst pilots, we see more than 85 percent of participants feel comfortable knowing not just how, but why and when to use AI. Not because they mastered features but because they learned how to shake off an unexpected failure.
Learning to shake it off is key, but how about learning to reframe that unexpected failure as not a failure at all. It's a cue to reframe your approach and shift your perspective. It’s realizing that backwards IKEA shelf isn't a mistake, it's a sign you might be trying to build the wrong thing entirely.
The Takeaway
If you want AI adoption to stick, resist the temptation to build a feature-first curriculum. You will always be a step behind the next update.
Teach exploration. Teach error recovery. Teach judgment. Teach peer-learning.
Cool tips can wait.
Up Next
Next time we’ll move into Lesson 4: You Get Three Chances to Make a First Impression and go deeper on one of the most overlooked elements of AI adoption: creating the right first impressions. Those early experiences are much more impactful than appear at first glance.
If you are navigating this journey right now, or preparing to roll out Copilot across your organization, feel free to take our lessons learned and apply what fits. These lessons are not about us having all the answers. They are about shining a light on a path that is unfamiliar and filled with new challenges.
And if you want to walk through how this applies to your environment, or simply compare notes on what you are seeing, our team at FlexPoint Consulting is always happy to connect.
Leave your thoughts below. I'd love to hear from you!