One Concept, One Example, One Question

The concept

Before I delegate a piece of work to AI, the first question I ask is: can I describe the outcome clearly?

This is the first of four questions in my Practical Delegation Test. A request such as “look into this” or “write something useful” may be enough to start a conversation, but it leaves the AI to guess what the output is for and how someone will use it.

Outcome clarity means being able to explain who will use the result and what they need from it. You also need to know what would make it acceptable or unusable. A prompt can record those choices, but the person delegating is responsible for making them.

For a small, reversible task, the question may be easy to answer; a briefing that will guide a decision needs more thought.

The example

Imagine a manager with a collection of customer comments and a meeting coming up. They haven't decided whether the meeting needs diagnosis, options, or a decision, but ask an AI assistant to “write a briefing on the main issues”.

The assistant can produce something fluent, but it has to invent the quality bar. It may organise the comments by topic or foreground the most emotionally striking examples, perhaps with suggested actions attached. The result could look useful while quietly choosing a purpose for the meeting.

The manager could first ask for two outlines: one for identifying a problem and another for choosing a response. A rough draft might help too, if the manager treats it as something to react to rather than a note to circulate. AI can help the manager settle the meeting's purpose before producing the finished briefing.

Suppose the meeting is meant to decide which part of the customer journey to investigate next. The manager can now ask for a two-page internal note for the service leadership team, identifying the two problems mentioned most often in the supplied comments. It should show the evidence for each and flag gaps that prevent a firm conclusion. It shouldn't recommend a policy change or present the comments as representative of all customers.

If the same comments were being used to justify a public commitment, that would be a different task with a different evidence standard. For the internal note, the manager can check whether it answers the meeting's question without claiming more than the comments support.

The question

Ask yourself:

For one task you are considering handing to AI, can you describe who will use the result, what they need from it, and what would make it unusable?

If you can't answer that yet, use an outline or rough draft to work out what you need before asking AI for a finished result.

Related

• The More Work AI Does, the More Judgment Matters explores why deciding what work to ask for matters more as AI makes execution cheaper.

Support

Everything I publish here is free to read, and always will be.

I maintain this newsletter and my website in my own time. If something here has been useful to you, and you'd like to help cover the costs behind that work, you can support me through Ko-fi.

Generative AI is a powerful tool. I openly use it to conduct research, challenge my thinking, generate drafts, and review my work. AI never owns what I think, or my personal judgement. Everything I publish is a reflection of my own thoughts and opinions.