People & leadership development
Video · 5 min read
The short answer
Watch: What capabilities should people develop to become effective leaders in an AI-enabled business?
Watch the explanation, then explore the practical implications below.
Why leadership development needs more than tool training
Knowing how to use an interface is helpful, but it does not establish whether an output is commercially sound or appropriate for the customer. Leaders need enough understanding to set direction, evaluate the work and decide what deserves attention.
The four capabilities in the video provide a useful development framework. They are complementary: experimentation without expertise can miss mistakes, while expertise without practical use can leave a leader unable to assess where AI genuinely helps.
Deep knowledge: recognise what does not hold up
Domain expertise helps a person spot an implausible assumption, an omitted constraint or an answer that sounds right but does not fit the situation. Preserve access to the underlying work: source documents, customer evidence, calculations and operational consequences.
Develop this through review of real examples. Ask the learner to explain the problem and form an initial view before comparing it with AI output. Have an experienced colleague explain which differences matter and why. The aim is to strengthen judgement, rather than train people to accept polished text.
Wide knowledge: understand the surrounding workflow
A leader does not need to master every discipline. They do need to understand the dependencies around their own area. A commercial decision may depend on delivery capacity, data quality, permissions and the way an integration updates information.
Use cross-functional walkthroughs of a real job. Ask colleagues to explain their inputs, decisions and failure points. This helps a leader write a more useful brief and recognise when a request needs technical or operational expertise beyond their own.
Taste: explain what makes an answer good
In this context, taste means judgement about fit and quality. Several outputs may be factually acceptable, while only one is clear, relevant and appropriate for the customer. Make the criteria explicit: accuracy, usefulness, tone, completeness and ease of action.
Compare two drafts and discuss the trade-offs. Replace “this feels better” with an explanation of what improves the outcome. Keep approved examples with the context that makes them appropriate. Avoid turning personal preference into a universal rule without examining its purpose.
Agency: take bounded action and learn from it
Agency is the willingness to move from discussion to a sensible test. Choose a task with clear boundaries, an owner and a way to assess the result. Use approved information and keep external commitments under the normal approval process.
After the test, record what helped, what failed and what to change. Reward accurate reporting of weaknesses as well as useful results. Otherwise, people may feel pressure to demonstrate enthusiasm while leaving practical problems unresolved.
A practical learning exercise for a leadership team
Choose a familiar proposal or operational decision. Ask each participant to form a view, then use AI to prepare an alternative or critique. Compare both against the original evidence. Discuss one factual issue, one missing consideration and one judgement about fit.
Finish by changing the brief or review criteria based on what the team learned. Repeat with a different case to see whether the learning transfers. Completion speed alone should not determine success; assess whether people can explain and defend their choices.
How CSM would approach capability development
We would connect training to the workflow being implemented. People should practise completing the task, checking evidence, correcting an error and escalating uncertainty. Managers need to understand their own review responsibilities as well as the user interface.
Make learning part of the process after launch. Review recurring mistakes, update approved guidance and preserve supervised practice for junior colleagues. Productivity gains are more sustainable when the team is also developing the expertise needed to judge future work.
Questions for your leadership team
Where do we need deep expertise to judge AI output?
Identify the tasks where a plausible answer could conceal a consequential mistake.
Which neighbouring disciplines do leaders need to understand better?
Map the operational, commercial and technical dependencies around a real workflow.
Can we explain our quality standards?
Compare examples and document why one is more useful or appropriate.
Do people have room to run sensible tests?
Agree a bounded task, approved information, a reviewer and a learning goal.
Are junior colleagues still developing judgement?
Give them opportunities to form an answer, inspect evidence and discuss differences with an experienced person.
Full video transcript
The full video transcript is included below for reference.
I think there are four things that are particularly useful. They’re not the only things, but they give you a good starting point. One is deep knowledge.
You need to understand something well enough to recognise whether the work is any good. An experienced person can look at a recommendation and notice that an important assumption is wrong before anyone else has finished reading the introduction. That becomes valuable when producing an answer is easy.
You can generate a great deal of material very quickly. Someone still needs to know which parts deserve attention and which parts don’t hold up. The second is wide knowledge.
As people become able to work across more areas, it helps to understand a little about the work around your own. A commercial person doesn’t need to become a software engineer, but understanding how a system is put together can help them ask for something useful and recognise where the difficulty might be. The third is taste.
Can you choose between several answers that are all technically acceptable? Which one fits the customer? Which one sounds like us?
Which one is clear enough that someone will actually use it? And you can develop that by making the judgement visible. Don’t just say, “This version is better.” Explain what makes it better.
That helps other people learn, and it gives the AI more useful direction next time. The fourth is agency. The willingness to take action when you don’t have complete certainty.
To try something sensible, look honestly at the result and change your approach. I think the difficult part for companies is developing all of that together. You want people experimenting, but you also want them learning the work deeply enough to judge what they’re doing.
If a junior person only sees the finished AI answer, they may get through the task without understanding it. So give them a chance to form a view, compare it with the system’s answer and discuss the difference with someone experienced. Otherwise, you can become more productive today while weakening the expertise you’ll need tomorrow.
Put these questions to work in your business
Bring a real decision or workflow to a conversation with CSM. We can explore the evidence, responsibilities and practical steps needed to make AI useful for your team.