Management & organisational design
Video · 5 min read
The short answer
Watch: How does AI change the structure of an organisation and what it means to be a manager?
Watch the explanation, then explore the practical implications below.
Managing AI starts with explaining the work clearly
The video uses the idea that everyone becomes a manager to some degree. The practical point is that assigning work to AI requires familiar management skills: explaining the purpose, supplying relevant context, defining a useful result and recognising when the work is off track.
This does not mean every employee needs a management title or that a new organisational chart should come first. Start with the activities people can delegate and the judgement they need to supervise those activities well.
Find the context that colleagues quietly supply
People often complete a vague instruction because they know the customer, remember a recent conversation or understand an unwritten rule. If that context is absent from an AI brief, the output may be plausible but unsuitable for the actual situation.
Review a recent task that required clarification or correction. Write down what the person knew beyond the instruction. Decide what belongs in the brief, what should become approved shared guidance and what still requires a conversation. The goal is to improve how the work is defined, rather than expect the system to infer everything.
Give managers a clearer purpose for the capacity released
Removing time spent collecting updates does not automatically create time for better management. Meetings, targets and expectations can absorb the capacity unless leadership makes an explicit choice about what should change.
Specify the work managers should do more of: resolve delivery problems, coach colleagues, improve a weak handover or address a customer concern earlier. Also identify activities that can stop. Keeping the old report, the old meeting and the new AI summary can leave everyone busier.
An example: project reporting and issue resolution
A manager may currently spend hours asking colleagues for updates, assembling a report and checking whether the figures agree. A connected workflow could prepare a sourced summary and identify missing information for review.
The manager’s attention can then shift towards the unresolved delivery issue the report reveals. That requires a named owner, a decision route and time to act. If leadership still rewards the number of reports produced, the process may simply generate more reporting without improving delivery.
Review responsibilities before changing reporting lines
Look at who owns the outcome, who prepares the evidence, who approves commitments and who handles exceptions. AI may reduce some coordination work, but other responsibilities remain or become more important, including reviewing quality and maintaining the information behind the process.
Test the changed workflow before drawing conclusions about team structure. Observe workload, decision delays, quality and the support people need. Do not assume that removing a task automatically removes the whole purpose of a role.
Train briefing, review and recovery together
Training should show people how to complete a real activity, assess the output, correct an error and recover when a source is missing. Managers need to understand the limits of delegation and how to escalate work that sits outside them.
At CSM, we would map those responsibilities with the people doing the work and agree what success means for the team. We would then use a bounded implementation to test the process and review expectations alongside the technology. The useful question is whether the team can achieve a better result with less unnecessary coordination.
Questions for your leadership team
What context do people supply without being asked?
Identify the customer knowledge, recent events and unwritten rules that a typical instruction leaves out.
What should managers do with the capacity released?
Name the higher-value activity and identify which existing report, meeting or task can stop.
Are responsibilities clear after the workflow changes?
Assign ownership for outcomes, approvals, exceptions and ongoing maintenance.
What are managers being measured on?
Check whether measures reward completed outcomes or simply visible activity.
What evidence should inform changes to team structure?
Review workload, quality and decision delays before altering roles or reporting lines.
Full video transcript
The full video transcript is included below for reference.
Everyone becomes a manager to some degree, because everyone has work they can give to AI. So your job is to give clear instructions. Can you explain what you want? Can you give enough context? Can you recognise when something is going off track?
That doesn’t mean every person suddenly becomes good at managing. In fact, AI can expose how much of our management depends on other people quietly filling in the gaps.
You give someone a vague instruction, but they know you, they know the customer and they know what happened last week. So they work out what you probably meant. If that knowledge isn’t available to the AI, the same instruction can produce something completely reasonable that is completely wrong for your situation.
That’s useful feedback. It tells you where the work depends on knowledge that only exists in someone’s head.
The bigger question is what the organisation does with the extra capacity. I think companies can retreat quite quickly to automation because it’s a familiar playbook. Find a task, reduce the labour involved, move on to the next task.
Some of that will make sense. But if that’s the whole ambition, you miss the possibility of people doing work they couldn’t previously get to. A manager might spend less time collecting updates and more time resolving the problem those updates were supposed to reveal.
You have to make room for that. If you remove the administrative work but leave every expectation and every meeting unchanged, you may not have improved the job very much.
So I’d look at what managers are being held responsible for. Are they being rewarded for supervising activity, or for helping the team produce a better result? As more of the activity can happen through AI, that question becomes much harder to avoid.
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.