Skip to main content
leadership cohort three roles of AI Pn

​
Start here: what is your time worth?

Before you open any AI tool, ask yourself one question: is this a task that needs my judgement, or just my time? As an executive, your highest-value work, the decisions, relationships, and judgement that only you can provide, is worth several hundred dollars an hour. The administrative tasks in your calendar and inbox are worth a fraction of that. AI gives you the chance to permanently reassign the $15/hr work so you can spend more time on the $500/hr work.
Before you use AI for any task, ask: should I be doing this at all? If yes, AI helps you do it faster. If no, AI may be able to own it entirely. That is where the real value sits.

​
The three roles

Most leaders use AI for one thing and get inconsistent results. The problem is not the tool. It is not knowing which mode to put it in.
Role + format + word limit = you are 80% of the way to a useful output.

​
The Assistant role: reclaim your day

The Assistant role is the fastest win available to most executives. It covers work that is important enough to do well but does not need your full attention: drafts, summaries, rewrites, and explanations. The key is treating AI like a capable new team member in their first week. They are fast and willing, but they need context. Tell them who you are, what you are working on, and what good looks like. Output quality rises sharply when you do.

Where it saves the most time

Email drafts, meeting summaries, document first drafts, and plain-language explanations of dense reports or contracts. Most executives find this alone saves 30 to 45 minutes a day.

Where to stay alert

Always review Assistant outputs before sending, especially anything going externally. AI can miss context you have not provided, or get details wrong with confidence. Your review is the quality check.

​
The Thinker role: analyse at scale

The Thinker role is where AI creates genuine advantage for leaders. It is not about speed. It is about depth and breadth of analysis that no individual could replicate alone. Think about the research that currently sits undone because no one has time to do it properly: competitive landscaping, vendor contract reviews, candidate screening, annual report analysis. The Thinker role handles all of it.

What Thinker does well

Synthesises large volumes of text into clear conclusions. Identifies patterns across multiple documents. Runs structured analysis against a defined framework. Holds an entire document in memory and answers questions against it accurately.

What still needs you

The right question. The relevant context. A check on the conclusions before they drive decisions. AI does not know what it does not know. A confident-sounding answer is not always a correct one.
For competitive analysis and market research, use ChatGPT’s Deep Research mode. It searches across sources, synthesises findings, and produces a structured report. Use the Prompt Builder to structure your brief before you run it.

​
The Creator role: prototype before you commit

The Creator role has moved furthest and fastest in the last 18 months. AI can now generate images, animate a still photo into video, write and run code, and draft entire presentation decks, all from a plain-language instruction. For executives, the value is not replacing your creative team. It is the ability to prototype an idea before committing resources. Generate a concept, test how it looks, refine the brief. All before a designer or developer touches it.

What to use it for

Concept visuals and mock social tiles. Video from existing images. Presentation structure and first drafts. Data visualisation from a spreadsheet. These are starting points for your team, not finished products.

Before you publish anything

Review all Creator outputs against your brand guidelines, legal obligations, and factual accuracy. This applies especially to anything customer-facing or externally distributed.

​
How to structure a prompt: the GEMS format

The biggest driver of poor AI output is an under-specified prompt. GEMS is a four-part format for giving AI the right information upfront, the same way you would brief a team member on a task.
1

Goal: what do you want it to do?

Be specific about the task. Name the role you want AI to take on, for example “Act as a senior strategy consultant”, and state the exact output you need. Vague goals produce vague outputs.
2

Environment: what does it need to know?

Provide the relevant background. Include what is in scope, what is out of scope, and any constraints. The more relevant context you provide, the more calibrated the response.
3

Method: what should the response look like?

Specify format, tone, length, and structure. Dot points or prose? Table or narrative? Formal or direct? If you do not specify, AI will guess, and it may guess wrong.
4

Sources: what should it draw from?

Tell AI what to reference: a file you have uploaded, a specific source, or its own knowledge. If you are uploading a document, say so explicitly and name what you want done with it.
You do not need all four elements every time. For simple tasks, Goal and Environment are usually enough. For high-stakes outputs, a board presentation or client proposal, use all four. The more you invest in the prompt, the less editing you do at the end.

See the full prompting guide

Covers all nine approaches including iterative refinement, scenario planning, chain of thought, compare and contrast, and more. Use it as a reference when you are preparing for a specific task.

​
A prompting approach worth knowing now

There are several approaches that work well for executive tasks. Here is the one that gets the fastest results for most leaders in this cohort. Ask AI to play a role before you ask your question. Assign AI a specific professional persona at the start of your prompt. This shifts the lens through which it responds and typically produces more specific, contextually relevant output than a generic question. 
Act as a CFO who has led two successful ASX listings. I am preparing for a board 
conversation about capital allocation. What questions should I expect, and how 
should I frame the growth investment case?
This approach works well for preparing for high-stakes conversations, stress-testing your own reasoning, and getting a domain-specific perspective quickly.

See more prompt templates

Covers three scenarios for each of the thinker, assistant, creator roles we can assign AI. Use it as a reference when you are preparing for a specific task.

​
Giving AI context that carries across sessions

A one-off prompt is the minimum viable use of AI. The real unlock is giving AI context that persists, so it already knows who you are, how you work, and what matters to you. There are three ways to do this:

Custom instructions

A standing brief you write once. AI reads it before every conversation. Include your role, your organisation, your communication preferences, and any standing rules.

Memory

AI tools like ChatGPT can remember facts and preferences across conversations. Over time, the tool builds up a picture of how you work and applies it without being asked.

Connectors

Connect AI to your existing tools: SharePoint, Google Drive, Outlook, Teams, and more. This allows AI to access current, organisation-specific information rather than relying only on its training data.
Think of it like onboarding a new team member. You would not expect full performance on day one without briefing them on the context, access, and working style they need. AI is the same.

​
Putting the roles together in a workflow

The biggest shift for most executives is moving from using AI as a one-off question tool to embedding it in actual workflows. The difference in value is significant. Here is how the three roles work together on a common executive task: preparing a briefing document from a complex data source.
1

Thinker: analyse the source material

Upload your data, report, or transcript. Ask AI to extract the key insights, identify patterns, and flag anything that warrants attention.
2

Assistant: draft the document

Ask AI to produce a first draft of the briefing using the analysis from step one. Specify the format, audience, and tone.
3

Assistant: review and refine

Return the edited draft to AI and ask it to check for consistency, flag anything missing, and suggest improvements.
4

You: add your judgement and final sign-off

Download the draft and edit it. Add the strategic context, the nuance, and the insights that only you hold. This is the step AI cannot replace.

Map your own workflow

Use the Workflow Discovery Tool to identify which role handles which step in one of your own recurring tasks.

​
Quick checkpoint

You are done with this section when you can:

Name the three roles

Describe what AI does in each mode and give a real example from your own work

Build a GEMS prompt

Write a prompt using Goal, Environment, Method, and Source for a task you need to complete this week

Design a workflow

Map which role handles which step in one of your recurring tasks, and identify where your review sits

Set up context

Write a custom instruction or add a memory to one of your AI tools so it knows who you are

Ready to apply this?

Head to the Week 2 Challenge to put the three roles into practice using your own work.