How to Choose AI Support for Programme Leadership Work

Someone in the corridor says the team should “just use AI more”. A senior stakeholder asks what the AI plan is. A vendor deck promises that status reporting will write itself.

None of that answers the only useful question on a stressed programme:

Which piece of real work is a sensible place to try AI this week?

That is a different question from “which AI product is best?” It is also a different question from “how do we transform the organisation with AI?” For hands-on programme and transformation leaders the useful answer is usually narrower and more practical.

Why starting with a tool is often the wrong first move

Product rankings and model comparisons feel productive. They rarely fix the thing that is eating leadership time.

If priorities thrash every Monday, a better model will not name the trade-off. If the decision owner is unclear, a clever prompt will not create authority. If the status pack rebuilds every Thursday because nobody trusts the outline, another writing tool will not fix the outline.

The first job is not to choose software. It is to choose a piece of work that is:

• real
• recurring enough to matter
• safe enough to try
• small enough to finish

Write one sentence before you open a tool:

“The work under pressure is ____, because ____ keeps getting rebuilt, delayed or argued every week.”

If you cannot complete that sentence, pause the tool hunt. Clarify the work first.

Four checks before you try AI on programme work

Use these as judgement checks, not as a form to fill in.

1. Can you name the output you need this week?

A decision paper. A trade-off view. A steering narrative. A stakeholder message. A risk table with owners and dates. If you only have a vague sense that “AI should help with the programme”, you do not yet have a starting point.

2. What information would go into the tool?

Keep sensitive names, commercial terms, personal data and restricted programme detail out of consumer tools and unapproved systems. If the only version of the work is full of protected information and you have no redacted alternative, AI is not the starting point yet. Redact first, or improve the structure offline.

3. Which tools are you actually allowed to use?

Enterprise Microsoft. Google. A short approved list. Consumer tools for low-risk drafting. Or none for this kind of work. If you do not know, do not invent permission. Use non-sensitive material only, or improve process while you find out who owns the rule.

4. Can you act on the result?

You can rewrite your own paper. You can prepare a conversation you own. You cannot use AI to manufacture a decision that someone else must make, or to bypass a governance step that still needs a human call.

If any of those four checks fails, that is useful information. It usually means the next move is process, ownership or policy clarity, not a prompt library.

Good starting points in programme and transformation work

These are the kinds of tasks where a careful AI assist often helps, provided the information is suitable and you stay inside the tools you are allowed to use:

• Preparing a decision paper from notes you already trust, so the options and ask are clearer before the meeting.
• Structuring competing priorities into one page that shows what is in, what waits and what risk each choice creates.
• Summarising known evidence into a short brief for a steering conversation, when the sources are already public or approved for that use.
• Preparing for a stakeholder conversation by turning messy notes into a calm agenda, likely concerns and one clear ask.
• Improving the structure of a plan by testing whether sections, owners and decision points are explicit.
• Finding missing decisions or ownership by reading a draft pack and listing gaps a human still needs to close.

In each case the AI is helping with structure, language or gap-finding. You still own the judgement.

When AI is not the right starting point yet

AI is a poor first move when the real problem is human or organisational:

• Nobody owns the outcome. The team is drafting while the decision still has no owner.
• The information cannot safely be used. There is no redacted version and the live material is too sensitive.
• Nobody has authority to act. A polished draft will not replace the missing approval.
• The team does not agree on the outcome. AI will only accelerate disagreement into better-formatted disagreement.
• The evidence does not exist yet. You cannot summarise what has not been collected, tested or decided.
• AI is blocked for this content. Then improve templates, ownership and decision structure offline first.

An honest “not yet” saves more time than a clever trial that creates policy risk or false confidence.

A 25-minute experiment for this week

Pick one real piece of work that is already on your desk. Prefer something you own, with low sensitivity, that needs better structure rather than new facts.

1. Redact names, codes, commercial figures and anything that should not leave your control. Use notes you would be comfortable putting in an ordinary draft email.
2. Write the outcome in one line. Example: “A one-page trade-off view that names three options and the risk of delay on each.”
3. Choose the tool you are allowed to use for this kind of drafting. If none is allowed, stop here and improve the outline on paper or in your normal office tools.
4. Ask for structure, not magic. Give the tool your redacted notes and ask for a draft outline, options table or agenda. Tell it what must stay empty for a human to fill.
5. Run a human quality check. Read the result and mark:
   • what is accurate
   • what is invented or too confident
   • what still needs an owner, date or decision
   • what you would not send as written
6. Decide the next step. Keep the structure and rewrite in your own words, or drop AI for this task and fix ownership or process instead.

Observe only two things by the end of the week:

• Did the structure save real time on a live item?
• Did anything unsafe, vague or over-confident nearly leave your desk?

That is enough evidence to decide whether to continue, narrow the use or stop.

If you want a personalised starting recommendation

When you already know the kind of pressure you are under and want a recommendation shaped to your situation, use free Where AI Could Help.

It asks seven short questions in the browser and gives you:

• one piece of programme or transformation work worth improving first
• one practical first action for the next 20 minutes
• a prompt or a non-AI alternative when that is safer
• a simple structure, quality checks and a small week experiment

It does not search tool directories, recommend named software products, compare vendors, verify prices or do live market research. The article above still stands on its own if you never open the tool.

If the free result is complete and the situation remains genuinely complex, for example mixed tools, high sensitivity or unclear approval ownership, you can later ask about a Leadership AI Opportunity Review through the contact page. Do not send confidential programme files unprompted. The free plan should still be usable while you decide.

What to do next

1. Name one piece of work under pressure this week.
2. Run the four checks on information, permission and authority.
3. Try the 25-minute experiment on a redacted version, or improve the process if AI is not yet safe.
4. If you want a recommendation shaped to your situation, use Where AI Could Help.

AI can support programme leadership work. It cannot replace the decision about which pressure is real, which constraints bind you and which next step is safe enough to try.