Business leaders must tackle shadow AI before it kills strategic thinking

When I first heard the term “shadow AI,” it felt familiar. This is a problem that IT leaders have been dealing with for years under various guises. Employees have always used unapproved software or free web tools to get their work done faster. In 2012, a French survey of IT managers found that 16% of employees were already using unsanctioned cloud solutions, and by 2015, Gartner estimated that 35% of enterprise IT spending happened outside formal IT budgets.

With the arrival of easy-to-access AI tools, I’m sure a lot of IT leaders would jump at a chance to have shadow usage numbers that low now. However, what used to be shadow IT is now shadow AI; it’s much more prevalent and it could have a much more devastating impact on our ability to work.

In “Shadow AI is on the rise. Here’s how to turn it into a strategic advantage” CIO contributor John Donegan explored the privacy and governance issues behind shadow AI. I’m beginning to notice another risk that goes beyond these issues. I call it the sycophantic spiral. It describes what happens when teams lean too heavily on AI for cognitive tasks, falling prey to over-reliance on every output from their generative tools. Eventually, they begin to lose the ability to think critically and creatively.

What I’m seeing across companies

In the past year, I’ve led consultation and training sessions for more than thirty organizations. Every single one had employees experimenting with AI tools on their own.

Most of what I see isn’t reckless use, it’s people looking for small efficiencies in how they work, such as rephrasing emails or summarizing meeting notes. But I’ve also seen employees paste client data or company IP directly into public chatbots without knowing that those tools might store or learn from what they enter.

IBM described it as the silent spread of shadow IT, where people adopt unsanctioned tools because official systems feel too slow or restrictive. Increasingly, that means employees are quietly using generative AI chatbots like ChatGPT or Claude, coding assistants such as GitHub Copilot, or AI-powered content creation and analysis tools to get work done faster.

Security concerns are serious, but what strikes me most is how these habits affect decision quality. Over time, leaning on AI for answers makes it easier to skip the hard thinking that builds real judgment.

How the sycophantic spiral develops

Recent neuroscience research shows that when we rely heavily on AI tools, our brains show an up to 55 percent reduction in connectivity compared to those who complete tasks independently. When teams use AI to handle reasoning tasks, they often stop doing the slower work of weighing options or checking assumptions. Delegating to machines could come at the expense of our own reasoning circuits staying active.

It’s easy to see how businesses are sleepwalking into this new reality. People still feel productive because the output looks polished, but their brains is actually going through something similar to muscle atrophy.

Over time, the outputs also become more and more similar. Algorithmic curation systems reduce content diversity by creating what researchers call “bounded diversification”. Chatbots will give the illusion of choice within an increasingly narrow information diet. As these systems learn your preferences, they filter out contradictory viewpoints or challenging perspectives. The systems essentially act as ‘yes’ men and are incentivized to agree with you.

I’ve watched this play out in real teams. The more employees rely on AI for analysis, the more predictable they become and the fewer new ideas they bring to the table. The greatest ideas and solutions come from a diversity of perspectives, but these tools reinforce familiar thinking patterns, limiting exposure to competing views.

This is the core of the sycophantic spiral. As AI learns our preferences, it feeds us versions of what we already believe.

How to spot the spiral

You can usually tell when the sycophantic spiral has started to creep in. It could be anything from presentations and reports starting to all share a similar tone, fewer experimental ideas in brainstorming sessions or meetings with less healthy debate. Leaving these unaddressed risks, your organization starts to run on autopilot. While it may feel productive, it could actually mean that you’re losing the creativity of human thinking that gives you a competitive edge.

Fortunately, even if the spiral has started, there is still time to bring things back, and CIOs have a critical role to play.

What can CIOs do?

The first step is to bring Shadow AI usage out into the open. Often, employees will be using these tools to solve real problems, so it’s critical to make sure there’s no shame attached to using these tools. Try starting with an anonymous survey where employees can speak freely, allowing you to understand what tools people are using and why.

From there, you can work out the next steps you need to take. It might be that employees have actually uncovered a useful tool that is worth safely onboarding, or that they can achieve the same results with the tools you already have, but just need more training.

Secondly, ensure that you find ways to bake diversity of thought into your organization. Teams and processes should be built in a way that incorporates different viewpoints into ideation and review to ensure you can keep your competitive edge and encourage creative thinking.

Structured dissent is one of the best ways to encourage diversity of views. In my time at Apple, leaders were encouraged to invite critical feedback from peers and teams. By giving permission to critique, you create a space for areas to improve and unique opinions. If you don’t have a team, a custom AI agent could be this sounding board for you. A similar approach is ‘red teaming’, where you establish a separate group with the sole purpose of challenging assumptions and finding holes in your strategies.

You should also find ways to reward originality. By the nature of business competition, processes are often set up to reward speed rather than creative thinking. While this may see immediate productivity gains, it could harm you in the long term.

A good example of this is Pixar. Pixar has a culture of creativity and innovative thinking, which means that its leadership insists on rewriting and re-storyboarding films until they work, regardless of the cost. Movies like Toy Story 2 and Inside Out were almost completely restructured mid-production. Pixar’s consistent creative success across decades proves that taking time to find creative ideas leads to long-term market recognition and gains.

Finally, make sure you establish guardrails around AI usage. CIO senior writer Grant Gross’ article, “AI could prove CIOs’ worst tech debt yet,” highlights how quickly AI projects can create more work when left unchecked. Take the time to look at where AI can make work more efficient vs where human reasoning is essential. Mapping processes for the highest revenue-generating teams and largest cost centers can give you great starting points for a targeted AI pilot program. Lastly, train staff on the need to validate any AI-generated insights before acting on it. If AI is your brainstorming partner, ask it to use layered abstraction to evaluate the initial ideas and look for feasible alternatives. If you’re building on an established idea, ask AI to use 1st, 2nd and 3rd order impact analysis to hone your idea into a more robust or holistic solution.

We can’t afford to stop at data privacy

Donegan and Gross raise valid concerns about Shadow AI. We should start with a lens of governance, privacy and policy adherence, but CIOs who stop there risk a quiet but deadly flattening of company culture and intellect. Building a COE (center of excellence) or coordinating a group of SMEs (subject matter experts) from around your organization helps with both diversity of thought and line-of-sight to operational opportunities for AI collaboration.

It’s not all bad news. If we address this issue right now with intentional and clear policy, leaders can harness all of the productivity gains afforded by AI, without sacrificing what gives their business a competitive edge, people.

The modern CIO role now extends beyond simply gatekeeping AI; they need to become stewards in a new era of technological collaboration. Because when AI thinks for you, innovation slowly dies. When AI thinks with you, genius awakens.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?