If you lead teams through AI adoption, they pick up their AI habits from you: how you check AI output, whether you ask why, and what you share with other units. So the work starts closer to home - with how you use AI yourself.
Christoph Volbers trains Lufthansa leaders on AI at the Lufthansa Innovation Hub, a unit at the edge of Lufthansa Group that brings new skills into the company. His sessions often open with a question: what are the different kinds of AI, from simple rule-based systems to today’s agentic AI, and how do they differ? Very few leaders have a quick answer to this.
For the hub however, this gap matters, because teams follow their leaders. Christoph points to research from BCG showing that a team's progress with AI closely matches its leader's view of AI. Across the group, gaps in that view lead to three problems. Leaders let workslop through: AI-assisted work that looks finished until a colleague has to redo it. Teams stop asking why the AI produced an answer, and the work turns into a black box. Units run their own pilots and learn the same lessons alone. Here is how the hub tackles each one, in steps any innovation leader can take with their own teams.
1. Learn enough about AI to catch workslop
To answer the opening question, Christoph walks through one simple task: filtering spam out of an inbox, and shows how each kind of AI would handle it.
With rule-based AI, someone writes a list of 50 words, and the system flags any email that contains one of them. With machine learning, the system gets 10,000 example emails and a list of possible warning signs, and works out which signs matter. With deep learning, the system gets a million examples and finds the warning signs on its own. With generative and agentic AI, a user simply asks for an inbox free of spam, and the AI works out the rest. At each level, the machine takes on more of the work.
Each kind of AI has its own strengths and weaknesses. Rule-based AI dates back to the 1970s, so most leaders already use at least one kind of AI in their daily work.
An employee finishes a task with AI, feels done, and hands it over. The colleague who receives it finds a bad quotation and a broken link, and has to redo the work. The task looked polished, yet it created more work in total. A leader without a working knowledge of AI struggles to spot this kind of workslop, so the error passes; and each pass erodes trust inside the team. For that reason, Christoph advises: treat AI literacy as a leadership skill, built through the leader's own daily use of AI.
Knowing the tools also helps leaders see a hidden risk. Christoph describes two teams, one working with AI and one without. The real test comes when the AI is taken away: only then does a leader see how well the team performs, and how long it keeps going, on its own.
2. Put the leader's own beliefs on the table
Teams mirror their leader, so the question that matters most for Christoph is: “As a leader, what does your team learn from you while observing you about AI?”
A leader's beliefs about AI shape how that leader uses it, and the team copies what it sees. For that reason, the hub's training starts with the leaders themselves, by asking them to name their beliefs.
Each leader completes one sentence: "My current belief about AI is..." A common answer is that AI is mainly a technical topic. Leaders then talk through their answers in pairs for two minutes. Two minutes is enough to show how deep the topic runs, and Christoph encourages pairs to keep talking over the next break.
Next, leaders share their statements with their own teams. Christoph sees this as equally important, because the team can then see where its leader stands and push back.
From there, leaders plan how the team will learn together. The hub built a list of about 40 team activities, from a one-hour prompt battle for beginners to an ethics quarterly for teams with experience behind them. Leaders use the list to plan the coming year, and Christoph wants that plan to stay playful: "We don't learn if we don't have fun."
3. Make asking why a team routine
As AI output spreads through a team's work, people lose track of where results come from and why the team does what it does. When the team lacks answers and feels unable to challenge the output, trust between colleagues drops, and over time the team culture suffers.
To stop the black box from forming, the hub asks leaders to build AI diligence: team routines where colleagues question each other's AI work. When an AI result comes in, the team asks where it came from, why the team is acting on it, and how it fits with the rest of the work. Over time, asking why becomes normal, and people speak up instead of staying quiet. Christoph gives the job of building that habit to each individual leader, customized to their team activities.
Across Lufthansa Group, many units run their own AI initiatives and lighthouse projects. These projects are useful for trying new things. To use AI at scale, though, the group needs shared standards; however, today one team often learns a lesson while another unit spends time learning the same lesson again.
Standards start with leaders telling each other what their pilots taught them. A leader who has solved a problem passes the solution to peers in other units.
The hub is still working out how to set shared standards. For now, the work falls to leaders: learning the tools, sharing their beliefs, and getting their teams to ask, "Why are we doing this?"
