AI results depend on three decisions that only the leadership team can make. A company can have tools, training and pilots in place and still see little value if these decisions stay open.

 

In my conversations with leadership teams, I often meet companies with a lot of AI activity: licences, pilots, a training programme, an AI working group. When I ask which business result each activity should deliver, who owns that result and what will change in daily work, the answers differ from person to person. The symptoms are easy to recognise:

 

• A pilot works well technically, but no business owner is responsible for turning it into daily practice.
• Employees have access to AI tools, and their workflows look the same as before.
• Nobody recorded a baseline, so nobody can show whether AI saved time or improved quality.
• Managers are unsure which decisions they may hand to AI, and employees are unsure what is expected of them.
• Some people use unapproved tools because the rules are unclear.

 

Each of these points back to a decision that leadership has not yet made. More tools or more training will not close that gap.

 

Why this happens

AI usually enters a company as a technology topic. The answers to the important questions are also spread across the organisation, there is no common ground. Each function decides its own part, and nobody combines them into one direction with a named owner.

Many leaders also have limited hands-on experience with AI themselves. That makes it harder to judge which use cases are realistic, which risks matter and how much change the teams can do. 

 

 

The three leadership decisions

 

Decision 1: Where can AI create value for us?

Start from the business priorities for the next twelve months, for example shorter delivery times, fewer errors in invoicing or faster answers to customers. Then ask which workflows influence those priorities and where AI could improve speed, quality, cost, risk or customer experience.

• Choose three to five business priorities and ask each business unit to name the workflows that hold them back.
• Measure the baseline before you start. How long does the task take today, how many errors occur and what does it cost per case? If a finance team processes supplier invoices, record how many minutes one invoice takes now and how many need correction. Without these numbers, you cannot show in six months whether AI helped.
• Score each use case on value, feasibility and risk on a simple scale from one to five. Value asks what the improvement is worth, feasibility asks whether data, skills and systems exist, and risk asks what happens if the output is wrong.
• Name one business owner per use case. The owner comes from the business side and answers for the result.
• Write one success measure per use case, for example: monthly report prepared in one day instead of three.

Output: a short list of prioritised use cases, each with an owner, a baseline and a target.

A common mistake is to buy the tool first and search for problems afterwards. This produces pilots without owners, and nobody can say later whether they worked.

 

Decision 2: What has to change across the organisation?

AI changes how work is done, so technology is only one part of the answer. Leaders need to decide what changes around it.

• Workflows and decisions: For each priority use case, write down how the work runs today, how it will run with AI and which decisions stay with a person.
• Roles and skills: Identify which roles change, which skills are needed and how people will learn them. Give people time for this by removing existing tasks first, because AI adoption that adds workload will stall.
• Data, governance and human oversight: Define which data may be used, who reviews the output and who is accountable for the result. Keep the rules short enough that a team leader can explain them in two minutes.
• Collaboration: Bring business, IT, HR, legal and risk together in one working group per use case, with one person who makes the final call.
• Incentives and performance measures: Check whether current targets reward the old way of working. A team measured on the number of reports it produces by hand has little reason to automate them.
• Leadership behaviour: Every member of the leadership team uses AI on at least one recurring task of their own, talks openly about what worked and what did not, and makes decisions visible. Employees watch what leaders do and copy it.

Output: a change plan per use case covering workflow, roles, rules, ownership and measures.

A common mistake is to involve HR, legal and risk only when the pilot is ready to launch. They then either delay the launch or approve it without having shaped it.

 

Decision 3: What do we expect from internal communications?

Communications comes third because it works with what leadership has already decided. Once Decisions 1 and 2 are made, expect your communications team to:

• Explain why the company uses AI and which outcomes it wants.
• Say what will change and what will stay the same, including honest answers on the effect on roles and skills. If something is undecided, say so and say when you will know more.
• Translate the message for each employee group and show where human judgement and accountability remain essential.
• Create feedback loops so leadership hears about resistance, confusion and new risks early.
• Share how specific teams changed a workflow, so that other teams can copy it and AI becomes part of everyday work.

Communications cannot compensate for unclear leadership decisions. If the question about roles is unanswered, even the best-written message will sound evasive. Leaders provide direction, ownership and honest answers first, and communications then makes the change understandable, relevant and actionable.

Output: a short brief for your communications team with the answers to why, what changes, what stays the same and who decides.

 

What leaders can do now

• Set aside one leadership meeting to answer the three decisions for your top three use cases.
• Write down baseline, owner and success measure for each priority use case this month.
• Name the executive sponsor and confirm what that person can decide alone.
• Identify the roles that will change first and talk to the people in them before any announcement.
• Invite HR, legal and risk into the working group now.
• Ask every leadership team member to use AI on one recurring task and report back at the next meeting.
• Schedule the first quarterly review so that decisions to stop, continue or scale have a fixed date.

 

If your leadership team wants to settle these decisions together, this is the work we do in the AI Use Case & Strategy Sprint. Send me a message and we can look at where your team stands.

Which of the three decisions is still open in your organisation? Leave a comment and let us compare notes.