The Tool Zoo and Other Wrong Turns in AI Adoption

In short: Many teams collect AI tools, each introduced for a good reason, and lose a shared direction along the way. The tool zoo is the visible symptom of a missing strategy for AI adoption. Choosing a few workflows and offering one good official tool gets more impact from fewer licenses.

One tool for copy, one for images, one for analytics, one for automation. Each one makes sense on its own, and each one is introduced for good reasons, usually in the name of faster AI adoption. A year later, the team has fifteen logins and no shared answer to the question of what it’s all actually for. I call this the tool zoo. Lots of cages, but no zookeeper.

In practice

I once asked a team to write down which tool they used for what. The result was a list of more than twenty entries, more than half of which overlapped in function. Nobody had planned it that way. It had simply grown over two years, one tool at a time, each one the answer to a single, specific problem.

When activity replaces direction

The tool zoo is rarely the real problem. It is just the most visible symptom.

Behind it there is almost always a missing AI strategy. Nobody has a picture of which decisions systems should support in the future and which they shouldn’t, so in the end every team member decides for themselves what to try out. That creates a lot of motion but little direction. I often see it when two people on the same team solve the same task with different tools, without knowing about each other.

For the organization, that means paying twice, inconsistent results and, in the end, the question of why so many new tools haven’t produced a noticeable jump in impact. The answer rarely lies in the quality of the individual tools. It lies in the fact that nobody decided how they should work together.

The fear nobody talks about

A second, deeper pattern is the fear of losing power, which teams rarely talk about openly. People who have spent years building expertise in a particular discipline at first see an AI that does the same task in minutes as a threat. I’ve watched experienced colleagues deliberately hold back in how they used systems, simply because they sensed that their core expertise was losing value.

More training won’t make that fear go away. Training teaches people how to operate a tool, not which new job replaces the old one. Only once that question is answered, once it’s clear what role someone actually plays in the new setup, does the fear lose its footing.

Change takes more than a rollout

And third, AI adoption is often treated like an IT rollout. Buy a system, schedule a training, done. But people rarely change their behavior because a tool is available. They change it because they understand why the work has to change, and because they see a role for themselves in that change. I’ve seen rollouts that ran flawlessly on the technical side and were still barely used three months later, because this second part was missing entirely.

In practice, it pays to answer three rarely asked questions before every new tool rollout. Which decision should the system prepare going forward, what role does a human play in it, and who on the team actually loses something from this change that deserves to be taken seriously? These three questions cost an afternoon. Skipping them often costs a year of detours.

What matters most to me is that none of these three patterns is a technical problem. They are strategic and cultural questions that happen to show up dressed as a tool problem. That’s why hunting for the next, better tool doesn’t help. What helps is asking up front which operating model you actually want to build with AI, and then sticking with that answer, even when it gets uncomfortable. Once you’ve settled that, you’ll quickly notice: the tool zoo thins out on its own.

What does the tool zoo have to do with shadow AI?

Next to the officially rolled-out tools, a second zoo grows in almost every company, and nobody sees it. McKinsey found that three times as many employees use generative AI for a large share of their work as their leaders assume. The German Economic Institute (IW) reports that German companies mostly use free AI tools. A big part of AI adoption happens without anyone steering it.

A ban rarely helps. If you prohibit tools that make people’s work easier and offer no alternative, usage goes underground. Data keeps flowing out, and now even fewer people know about it. What works better is an official tool good enough that nobody wants to work around it, plus a short rule on which data may go into which tool. Add a simple way to propose new use cases.

My book describes a fictional example. The mid-sized manufacturer Overbeck gives employees four weeks to report, without consequences, any tools they use outside the official ones. 63 reports come in, and IT adds five of those applications to the official offering. Since then, the managing director asks one question about every new AI idea: Which of our three workflows does this improve? If the answer is none, the idea goes on next year’s list.

Frequently asked questions

What is the tool zoo?

That’s my term for a landscape of individual AI tools, each introduced for good reasons, that don’t serve a shared goal. The result is a long list of logins and little impact from AI adoption.

Sources: McKinsey (2025): Superagency in the workplace · Engels, Scheufen & Schmitz (2025): Künstliche Intelligenz als Wettbewerbsfaktor für die deutsche Wirtschaft, IW-Report 33

Similar Posts