Teaching Excel alongside Copilot: Learn, Confirm, Extend

I’ve had the good fortune to work with clients who are at the forefront of Copilot adoption. They want their people to understand what the technology can do, and they’re willing to rethink how work gets done. At the same time, they recognize that their teams still need a solid foundation in Excel.

That creates an interesting training challenge. If you spend all your time on the fundamentals, people may leave without seeing how Copilot fits into their work. If you jump straight into asking Copilot to build reports and write formulas, they may produce something impressive without being able to explain or check it.

So how do you strike the balance? I’ve been thinking about it as a simple progression: learn the foundation, confirm your understanding with Copilot, then extend what you can do with it. Each stage gives Copilot a different role, while keeping the learner responsible for the work.

1. Learn the foundation without Copilot

Start with the skill you want someone to develop. That might be summarizing transactions, writing a lookup formula, shaping data in Power Query, or choosing a useful chart. Before bringing Copilot into the exercise, give learners a manageable example they can work through and explain for themselves.

This doesn’t mean memorizing every Excel function or becoming an expert before touching AI. It means developing enough understanding to recognize a sensible approach, predict what a result should look like, and notice when something is wrong. Someone learning a lookup needs to understand what identifies a match. Someone building a summary needs to know which records belong in the total.

Consider this small, fictional dataset. The question is: What are actual sales for the East region in January, including returns? Here, Month is a simple text label so we can concentrate on the filtering logic.

The answer is 2,000: 1,200 + 900 − 100. A learner should be able to identify those three records before writing a formula. The West transaction, February transaction, and budget amount are excluded for different reasons; the return stays in because the question asks for actual sales including returns.

From there, they can use SUMIFS with criteria for Region, Month, and Scenario, and compare the result with the total they already worked out. The foundation is the relationship between the question, the data, and the calculation. Knowing the function’s name is only part of that.

2. Confirm understanding with Copilot

Once someone has learned the concept, they need opportunities to use it. This is where Copilot can become a useful practice partner. Instead of searching for another exercise that happens to cover the right skill at the right level, a learner can ask for a fresh example built around the concept they just studied.

The key is to preserve the work of thinking. Ask Copilot for a problem, attempt it before looking at a solution, and then compare approaches. If the first response reveals the answer too soon, set that example aside and request another. The value comes from making a decision and explaining it before receiving feedback.

After working through the exercise, the learner can ask Copilot to question their reasoning. That gives them something more useful than a quick “correct”: a chance to explain an assumption, defend an exclusion, or find a mistake.

The word confirm needs some care here. Copilot’s approval is not proof of mastery. It can misunderstand the task or produce incorrect feedback; Microsoft advises reviewing and verifying its output. In this example, the learner should still check the qualifying rows and total independently. In a training program, an instructor’s checked answer or another known test case can provide that reference.

A stronger indication of understanding is whether the learner can solve a fresh version, explain the result, and catch an error. Copilot helps supply practice and questions. The evidence comes from what the learner can actually do.

3. Extend what you can do with Copilot

With the core logic in place, the next question becomes more interesting: what could you do with this skill that would otherwise take more time or be harder to explore? For our sales example, that might mean comparing every region, adding a budget variance, or investigating how returns change the picture.

At this stage, Copilot can help propose approaches, draft formulas, or suggest a layout. The learner now has a basis for judging those suggestions. If a regional summary includes budget rows in actual sales, they can recognize the problem. If a variance calculation uses an unexpected sign convention, they know to ask what the number is meant to communicate.

Keep this exploratory work in a practice copy. Ask for an explanation of proposed changes, inspect the formulas, and check totals against the source. Copilot features and access vary, so a conversational plan may be the right starting point even when direct workbook editing is available.

Extension also brings new foundations into view. A learner who can use SUMIFS may still need to learn how a PivotTable groups records, how real dates differ from month labels, or how to handle duplicate transactions. When that happens, return to a small example, learn the new idea, and work through the progression again.

How I’d build this into team training

I’d organize a session around one meaningful task and carry it through all three stages. For example, learners could first build and explain a small sales total, then use Copilot to generate and discuss a variation, and finally explore a summary with additional regions or reporting questions. Keeping the same underlying task makes the role of Copilot easier to see.

The checkpoints matter more than giving every stage equal time. A beginner may need most of the session for the foundation. An experienced Excel user may be ready to spend more time on checking AI suggestions and exploring extensions. A quick independent attempt can help establish the starting point.

This follows a familiar teaching rhythm: explanation, supported practice, and application in a new situation. The framework gives us a practical way to decide where Copilot belongs in that rhythm. It also changes what we look for after training: whether people can explain a result, handle a variation, and recognize when they need help.

I’d carry the same approach into follow-up work, too. Give people a small task from their role, a clear way to check it, and a place to discuss what happened. I’ve outlined a broader approach in this 30-day plan for teams after Copilot training.

Give Copilot a clear role in the learning process

There’s plenty to teach people about writing prompts, but prompts make more sense when we know what kind of learning we’re asking them to support. During practice, we may want Copilot to hold back an answer. During application, we may want it to suggest several approaches. In both cases, the learner needs a way to judge the response.

For your next training exercise, choose one skill and write down what learners should be able to explain without AI. Then decide how Copilot could help them practice it and what a worthwhile extension would look like. That’s a manageable place to start balancing the foundations with the possibilities.