One Concept, One Example, One Question

The concept

The fourth and final question in my Practical Delegation Test is: Where does the uncertainty go? The first three clarify the outcome and make checking possible. They also identify what someone must still decide. Even then however, some things remain uncertain.

It may be an untested assumption, an unconfirmed fact or a case the check never reached. AI can make work resting on any of them read as fluent and finished. Unless I say what remains open, the recipient has little reason to look further.

Passing work on silently moves the checking to the recipient, along with effort I thought I'd saved. Review is reasonable when the recipient knows what's open and has the time, context and authority to challenge it. Where the result lands changes the stakes: an internal reviewer can ask for clarification, while a customer may act on it.

Before I pass anything on, I can settle the uncertainty, name it with an owner for the next check, or hold the work. I ask which route fits the work and its destination.

The example

Imagine an organiser preparing an email for a community event next week. It needs to go today so people can plan their journeys; two attendees need step-free access. AI drafts clear arrival instructions but says the meeting room has step-free access, drawing on an old venue description in the plan.

The organiser calls the new venue. The entrance is step-free, but the lift to the room is being repaired. The venue will confirm tomorrow whether it will work on the event day.

The organiser could hold the email until tomorrow, leaving attendees waiting for travel details. Or they could send it today with the access limit stated plainly. Forwarding the original draft as ready to send would turn saved drafting time into a communications colleague's unplanned checking work, or leave the two affected attendees planning around an unsupported assurance.

The organiser chooses to send the confirmed details today and tell attendees that room access is still being checked. Those two attendees may delay deciding whether to come. The note to the colleague says: “The entrance is step-free; we haven't confirmed whether the lift will work on the event day. Say we'll update everyone tomorrow. I'll call the two attendees who need step-free access today and check again with the venue tomorrow.” They make those calls. If the lift stays out, they must arrange another accessible room or change the plan, then tell attendees.

The task was clear and the check cheap, but the organiser couldn't settle the lift question before the email had to go. Naming it in the handoff let them keep tomorrow's check while giving attendees the travel details they needed. AI saved drafting time, but the organiser still makes the calls and follows up. Because the access decision stays with the organiser, my Delegating Work to AI guide treats this as human-led work with AI support. A phone call found this uncertainty; an untested assumption could go unchallenged.

The question

For one piece of AI-assisted work you're about to pass on, what remains uncertain, and who will deal with it or find out first?

Related

• The earlier issues cover outcome clarity, verification cost, and retained human judgment.

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.