From Use Case to Business Case: Why 5,000 Copilot Licenses Are Not an AI Strategy

Everyone thinks they need AI. Everyone is buying AI. Very few are getting real value from it.

Rolling out 5,000 Copilot licenses or adding a chat interface to your intranet is not an AI strategy. In this session, I walk through real use cases from enterprise and scaleup environments: embedded AI in products, GenAI for knowledge work and AgenticAI setups in operations.

Instead of another high-level overview of models and tools, we focus on what actually worked, what failed, and why. I break AI initiatives down into business-case types and connect each of them to concrete examples from more than 20 client projects along the way.

You will see that not every business needs AI. But every AI initiative needs a solid business case: expected outcomes, value mechanics, total cost and a realistic plan for long-term operations.

By the end, you will know where AI genuinely helps, where it adds no value at all, and which factors determine whether an idea even qualifies as an AI use case. We close with a simple, step-by-step method you can use to identify and validate AI opportunities in your own organization so they come with a clear business case from day one.

Key takeaways

This talk shows why buying 5,000 Copilot licenses for $1.8M is not an AI strategy but an expensive, unused button. You will learn to start from your value stream rather than a tool, to distinguish the three types of AI business cases (Internal Capacities, Internal Processes, Customer-Facing) and match the right AI type to each, and to ask the decisive question first: does this problem even need AI? The takeaway is that AI is only ever a means to an end, and a real strategy answers "Which problem am I solving, how, and for whom?