INSIGHTS
From Shadow AI Chaos to 100+ Use Cases in Five Days
A Canadian energy firm's team was using shadow AI with no guardrails. Five days later, they'd contributed over 100 use cases and were building their own AI tools.
The owner called with a problem that had two sides.
Side one: his team was already using AI. Not the tools the company provided — their own workarounds, their own accounts, their own prompts. Classic shadow AI. No governance, no consistency, no visibility into what was being shared with which models.
Side two: when he tried to formalize it, the team pushed back. They'd built their own workflows and didn't want someone coming in to tell them they were doing it wrong. There was real resentment toward the idea of a workshop or any kind of top-down alignment.
The owner was a power user himself. He understood the technology. He just didn't have time to educate 30+ people and didn't want to create a political problem by mandating a specific approach.
We ran the five-day implementation sprint.
Monday and Tuesday were discovery — a series of 45-minute conversations with team members across the company. We learned the industry, learned the workflows, and more importantly, learned what people were already doing with AI and why. This is where most consultants skip straight to the presentation. We don't. You can't solve a problem you haven't listened to.
Day three was fundamentals and use cases. Not a lecture — a working session. We showed them the tools already integrated into their workplace (the ones the company was paying for but nobody was using), demonstrated what was possible, and started building use cases together. The shift happened here. When people see their own work getting easier in real time, the resistance drops.
Day five was the roadmap. ROI formulas for prioritization, a phased plan, and over 100 use cases the team themselves offered to the company as places to start. Not our use cases — theirs. That's the difference between a roadmap that sits in a drawer and one that actually gets executed.
Today, the company is working the roadmap. Team members are creating Gems — small, purpose-built AI tools — for their specific areas of expertise. The people who know the work best are building the tools that improve it. Each Gem is managed by the team expert in that area, not by IT, not by us.
Our feedback score: 85%. The most common comment was about how approachable we made the process of sharing ideas with leadership. That matters more than the score. A team that feels heard is a team that executes.
The sprint cost less than one bad hire. It resolved a shadow AI governance risk, aligned the team, and produced a roadmap with 100+ concrete next steps. Five days.
