What Should a Transformation Leader Do When Their Team Is Using AI Behind Their Back?

Stop calling it "behind your back" and start calling it what it is: your team solving problems faster than your organization can approve tools. They are using ChatGPT for code reviews. They are using Claude for documentation. They are using Gemini for data analysis. And they did not tell you because they assumed you would say no. The question is not whether this is a problem. The question is what you do now that you know.

Stanford's Human-Centered AI Institute reported in 2025 that AI adoption is diffusing across workforces at unprecedented speed. For perhaps the first time in history, the technology driving automation can be used by anyone without IT provisioning a license or security signing off on an integration. That speed is creating a gap: teams are adopting AI faster than organizations can govern it. And transformation leaders are finding out about that adoption weeks or months after it started.

Here is the practitioner moment: You are the transformation director. You just found out from a casual Slack conversation that your engineering team has been using AI coding assistants for three months. No one asked permission. No one ran it through procurement. No one checked whether it complies with your data governance policies. Your first instinct is to shut it down. Your second instinct is to ask why no one told you. Neither instinct solves the actual problem, which is: your team found a tool that makes them more productive, and your organization made it so hard to get approval that they went around you.

This is where Leading Self in the Age of AI becomes the only pillar that matters. Shadow AI adoption is not a compliance failure. It is a signal that your organization's approval process is slower than the pace of change. You can shut down the tools. You will. But unless you also fix the process that made your team feel like they had to hide this from you, they will just find another workaround. The technology is not the problem. The trust gap is the problem.

Deloitte's 2025 research found that 93% of AI transformation budgets go to technology and only 7% to people. That imbalance shows up here. Organizations invest in buying AI tools but not in teaching people how to evaluate them, adopt them safely, or raise their hand when they find something useful. So people do not raise their hands. They just start using the tools quietly and hope no one notices. That is not malice. That is your approval process being a bigger blocker than the problem they were trying to solve.

So what do you actually do? Not next quarter. This week.

WHAT TO DO MONDAY MORNING

1. Do not shut down the tools. Ask what problem they solve first. Your instinct is to say "stop using that until we approve it." Resist. Schedule 30 minutes with the team lead who has been using the AI tool. Say this: "I found out you have been using ChatGPT for code reviews. I am not mad. I am curious. What problem does it solve that our current process does not?" Then listen. Really listen. Do not interrupt with compliance concerns. Do not explain why they should have asked first. Just listen. Because here is the truth: if the AI tool is solving a real problem, shutting it down just sends the problem back to your team with no solution. If the AI tool is not solving a real problem, you will hear that too. Either way, you need to understand what they were trying to fix before you decide what to do next.
   

2. Turn the underground experiment into an official pilot. Here is what you say to the team: "You have been using this tool for three months. That means you have three months of data on what works and what does not. I want to turn that into an official pilot. We will run it for 60 more days with proper tracking: what use cases work, what use cases do not, what risks showed up, what guardrails we need. At the end of 60 days, we will decide whether to adopt it enterprise-wide, restrict it to specific use cases, or shut it down. But we are doing this transparently now, not underground." That does three things: it legitimizes the work they already did, it gives you visibility into risks you did not know existed, and it shows the rest of the organization that when someone finds a useful tool, you do not punish them for finding it.
   

3. Fix the approval process so this does not happen again. Right now, your team went underground because they assumed the official process would take too long or say no by default. Fix that. This week, sit down with whoever owns your technology approval process -- probably IT, security, and procurement. Show them what just happened. Tell them: "Our team adopted an AI tool without approval because they believed our process would block them. That means our process is broken. We need a fast-track approval path for low-risk AI tools. I am proposing this: if a tool does not touch customer data, does not integrate with our systems, and costs less than $50 per user per month, teams can adopt it with notification, not approval. We get visibility. They get velocity. If that is not acceptable, tell me what is. Because the alternative is more shadow adoption and less visibility." If IT pushes back, show them the data: shadow adoption is already happening. The question is whether you see it or not.

You cannot stop your team from finding better tools. But you can decide whether they trust you enough to tell you when they do. Written by Transformation Leader. Published at t4leader.com.