How to Regain Control of AI Work in 30 Minutes
How to Regain Control of AI Work in 30 Minutes
Say you lead marketing at a company getting ready to launch a new product, and for the past three weeks AI agents have done a lot of the preparation. You open the workspace on Monday morning and realize you can't say what state most of it is in.
A competitor research project finished last Tuesday and nobody has read the report. The launch article is waiting on the product lead. An email sequence hasn't moved in four days. There's a pricing-page project from before the team changed the pricing. The FAQ task has been quiet since Friday, and someone on your team started a second one because the first looked stuck. And the customer-interview summary you approved last week never reached the sales lead who asked for it.
Each of those items is a loose end. Some are finished and some aren't. What they have in common is that nobody has settled what each one is now, or who decides what happens to it next.
The tempting move is to start something new. A fresh project feels like progress and gives everyone something to watch. But it doesn't tell you what happened to the work already in motion, and it adds one more thing you'll have to check later.
Try a half-hour session we call the Loose-End Reset instead. It has three parts: Close, Clarify and Commit. It won't clear your backlog. What it gives you is a queue you can explain and one next step with an owner.
Why AI work gets away from you
AI agents can turn out plans, drafts and analyses faster than one manager can read them and decide what to do with each. So the list grows, and every time you look at an item you have to work out again where it stands.
In the launch workspace, the queue looks busy while useful work sits still. The competitor report is done and unused. The second FAQ task is spending effort on something already underway. The pricing project will keep producing drafts for a decision the team has already made.
Each one keeps a small claim on your attention. In a 2009 study published in Organizational Behavior and Human Decision Processes, Sophie Leroy ran two experiments on what happens when people switch between work tasks. She found it was hard for people to pull their attention away from an unfinished task, and they did worse on the next one as a result. Finishing the first task wasn't enough on its own. People could only let it go when they had finished it under time pressure.
Leroy was studying task switching, not AI. But her findings fit what the launch workspace feels like on Monday. Every time you move on to something, the items you couldn't settle come with you, and finishing more of them won't necessarily leave them behind. So the reset starts by deciding what each one is.
Close, Clarify, Commit
The reset runs in three blocks of about ten minutes. First you close whatever you can settle now. Then you work out what state the rest is in, who owns it and what exactly it's waiting for. Last, you pick one next action with an owner and a finish line, and hold off on a second fix until that one shows a result.
The order matters. If you commit before you close, you're adding work on top of a state you don't understand. If you close without clarifying, you can mistake quiet work for abandoned work, which is how the second FAQ task got started. And if you clarify without committing, you leave with a tidier description of the same stuck workspace.
Some decisions need more than ten minutes, and they should get it. The clock is there to keep you moving through the list, not to rush a call about money, safety or a customer.
Close: settle what each item is
Closing a loose end means deciding what it is now: finished, worth one more pass, no longer needed or fine as it is. It rarely means deleting it or starting over, because useful work may already be sitting inside.
Start with the items closest to being useful.
Read the competitor report. If it does the job, accept it. If one section misses (say it compares your product with last year's version of the competitor's), ask for that section to be redone and keep the rest. Naming one gap protects the parts that are already right.
The interview summary was approved and then went nowhere. That counts as a loose end too, and in some ways it matters more than a task nobody has started, because the value already exists and nobody is using it. Closing it means getting it to the sales lead.
The pricing project exists to support a decision that's already been made. Cancel it. If any of its drafts are worth keeping for reference, archive them once you've looked.
Then there's the FAQ task. It's been quiet since Friday, but when you open it, it has its source material, it's working through the questions and nothing is waiting on you. Leave it alone, and cancel the copy someone started over the weekend. Restarting quiet work is an easy way to make a workspace messier: you end up with two versions, conflicting drafts and less idea of which one is current.
Clarify: name what each item is waiting for
The launch article and the email sequence can't be closed in ten minutes. For items like these, swap "stuck" and "waiting" for a description someone could act on.
Four states cover most of it. Work is healthy when it has a clear outcome, an owner, the inputs it needs and a next event you're expecting. The FAQ task is healthy, even though it looked stuck.
The email sequence is blocked. Something specific stops it: a customer list in a spreadsheet nobody shared. Other blockers are as concrete, like an approval, a missing file or a person who's away.
The launch article is awaiting a decision. The product lead has what they need to approve or reject its two claims about new features, and it can't move until they do.
Work is unclear when you can't say what it's for, who owns it or what happens next. If a lot of your list lands there, that's a management problem to sort out before you ask any agent for more.
For anything that matters, try finishing this sentence:
This work exists to help [person] decide or do [outcome]. It's [state] because [exact condition]. [Owner] will look at it again when [event] happens.
"The launch article is blocked" doesn't help anyone. "The launch article is waiting on the product lead, because its two claims need checking against the latest release notes, and the content owner will review it once the corrected links are in" tells you who has to act and when you'll find out.
Naming the spreadsheet does the same for the email sequence. It stops being a mystery and becomes a small job for whoever owns that list. Nobody had written down who that was, or when someone should step in if the list never turned up. Those are the details that go missing when work stalls, and writing them back down is most of the clarify step.
Commit: choose one next step
A good commitment is the smallest action that will tell you something new. It needs an owner, a piece of work it should change and a way to tell whether it worked.
In the launch workspace, the best candidate is the article. The product lead approves or rejects the two claims, the content owner updates the links, and by Wednesday you know whether the article can go out.
Other good commitments look like this:
- approve or decline the oldest decision that matters, after reading what it's based on;
- ask for one section to be revised against one source;
- share the missing input that's holding up work you already know is useful;
- turn one accepted result into the next piece of work it was meant to feed;
- close one project that's no longer worth the attention;
- run one small test task before switching a bigger workflow back on.
Weak commitments sound like "get the workspace moving," "restart the agents," "make everything better" or "start three new projects while we sort out the old ones." None of them names an owner or a result you could check.
Give the commitment a stopping rule as well. Don't start a second fix until the first has produced its result or clearly failed. Otherwise, when things improve, you won't know which change did it.
Check whether it worked
The reset worked if one useful result comes out of it for a reasonable amount of review. Activity is a poor signal. Logins, task counts, agent messages and finished drafts can all go up while nobody can say what the work is for or who decides next.
So check the work against what it was meant to do:
- Did the product lead approve the claims, and did the article go out?
- Did the revision to the competitor report fix the comparison without losing the rest?
- Did the sales lead get the interview summary, and use it?
- Did anything new turn up that now needs clarifying?
Making the plan also helps with the pull of everything still open. In a 2011 paper in the Journal of Personality and Social Psychology, E.J. Masicampo and Roy Baumeister reported several studies in which unfinished goals caused intrusive thoughts during an unrelated reading task and hurt performance on an anagram task. When people made specific plans for those goals, the effects went away, and how much they dropped depended on how earnest the plans were. So end the reset on a step someone means to take, like the product lead's call on the two claims, not just a tidier list.
Running the reset
Have your open work in front of you: the projects, the tasks and anything finished that's waiting for someone.
0 to 10 minutes: Close
- Read finished work that hasn't been accepted yet, starting with the oldest.
- Accept what does its job.
- Ask for one focused revision where a named gap stops the work from being used.
- Get accepted work to the person who needs it.
- Cancel work whose purpose has gone away.
- Find duplicates and keep the original.
- Leave healthy work alone.
10 to 20 minutes: Clarify
- Label each important item healthy, blocked, awaiting a decision or unclear.
- Name what it's for and who it's for.
- Name its current owner.
- Replace "stuck" with the exact approval, input, file, permission, budget limit, dependency or failure.
- Say what event should prompt the next look.
20 to 30 minutes: Commit
- Pick the action most likely to tell you something useful.
- Define the one piece of work or change of state it should produce.
- Name one owner and one reviewer.
- Say what a good result looks like.
- Don't start a second fix until this one comes back.
When the work runs on Supanova
Say the launch work ran on Supanova atoms. When a project finishes, the work comes back to you for review. The competitor research would have been in front of you last Tuesday as finished work to read, not a report someone had to remember to open.
The calls in the reset are still yours. Deciding that the pricing project no longer matters, or that the launch article is the one to move first, depends on what you know about the business, and that's where your half hour goes.
Before you start anything new this week, take thirty minutes with the work you already have. Close what you can, name what the rest is waiting for and pick one thing to move. If you'd like an autonomous workforce doing the launch work while you make those calls, you can start at supanova.team.
Sources
- Sophie Leroy, "Why is it so hard to do my work? The challenge of attention residue when switching between work tasks," Organizational Behavior and Human Decision Processes 109(2), 168-181 (2009). Record on IDEAS/RePEc; publisher page.
- E.J. Masicampo and Roy F. Baumeister, "Consider it done! Plan making can eliminate the cognitive effects of unfulfilled goals," Journal of Personality and Social Psychology 101(4), 667-683 (2011). PubMed.