Can Your Team Use What Your AI Produces?

Six · 5 min read · October 29, 2026

Can Your Team Use What Your AI Produces?

Say you lead an account at a design agency. Your client has dropped the second phase of their website project and added a members' area they didn't want when they signed. You've asked for a revised proposal before your next call with them.

The notification says the work is done. You open the project and find a summary, three attachments and a note that the proposal has been updated. Which file do you send? What changed from the version the client signed? Has anyone priced the members' area, or is that number a placeholder? The call is in an hour.

Something was produced, but you can't use it on that call yet. Before you say anything to the client, you have to open each file, compare it with the signed proposal and work out which parts are settled. The AI did the drafting, and everything after the drafting has landed on you.

The time goes to whoever opens it

BetterUp Labs and the Stanford Social Media Lab gave one version of this problem a name. In a September 2025 survey of 1,150 full-time US desk workers, they described AI-generated content that looks good but lacks substance as "workslop". Forty percent of the workers said they'd received some in the previous month. On average, each incident took about two hours to resolve, and the researchers put the cost at $186 per employee per month. They wrote up the findings in Harvard Business Review.

That survey was about work with nothing behind it, and your proposal might be perfectly good. You'd still lose time before the call, because you can't tell which file is current or what changed. Either way, the time comes out of your day, not the day of whoever made the files.

Four questions from the recipient's side

When the work comes back, ask what you'd need to know in that hour before the call.

Can you find the result?

The proposal should be where you'd look for it, with a name that says what it is: the revised proposal, phase two removed, members' area added. Earlier drafts can stay, as long as they're marked as earlier drafts. If the right file is the third attachment down with the same name as the other two, you'll spend the first ten minutes of your hour opening files.

Can you tell what you're looking at?

A proposal ready for the client and an internal draft with notes for your team look the same in a list of files. They do different jobs. The result should say which one it is and whether it's ready to send, so you don't discover on the call that the client is reading your team's notes about their budget.

Can you check what matters?

For this proposal, two things matter: what changed from the signed version, and where the members' area estimate came from. If that material is missing or sits somewhere else, all you can do is reread the proposal and hope it's right.

Say the estimate comes with a note: "based on comparable member portals." One of Microsoft's 2019 Guidelines for Human-AI Interaction is to "make clear why the system did what it did", and it comes with its own warning: the mere presence of an explanation has been shown to increase user trust, and it may cause over-reliance on the system. That note is an explanation, so it's easy to trust. To check the estimate, follow the note to something you can compare, like the scope the client sent or the earlier phases you priced for them.

Can you decide what happens next?

Work you can use comes back with a choice attached. You can send it, send it back with a correction, bring in your delivery lead on the estimate, or hold it until the client confirms the new scope. A green check beside an unread proposal isn't a decision. Microsoft's ninth guideline, "support efficient correction", asks systems to "make it easy to edit, refine, or recover when the AI system is wrong." If the members' area estimate is off, you should be able to say what's wrong with that section and get it fixed without rewriting the proposal yourself.

Write the sentence before the work starts

You can answer most of those questions before anyone drafts the proposal. When you assign the work, write one sentence from your own side: "I can use this when..."

For the revised proposal, it might read: "I can use this when it reflects the client's new scope, separates the terms we've agreed from the choices still open, and shows me where the members' area estimate came from."

Write it first, because once a polished proposal arrives, everyone wants to move on and nobody wants to ask what it should have contained. Put it in the assignment, and it tells whoever does the work what the proposal is for and gives you something to review against.

The proposal also touches price and delivery dates, and those may not be your calls alone. Decide who signs off on each before the work starts. You might own the client conversation while your delivery lead owns the estimate. If the revised proposal moves the launch date, it should be clear that the delivery lead sees it before the client does.

Keep the working notes close

The client will never see how the proposal was put together, but you'll need that material on the call. Say the client asks why the members' area costs more than the phase they dropped. The answer is in the working notes: what the estimate was based on, which features it assumed and which parts are still guesses. Those notes don't belong in the proposal. They should still be a click away from it, not buried in a thread you can't find while the client waits.

Running it on Supanova

Say the revised proposal ran on Supanova, our autonomous workforce. Your "I can use this when" sentence goes in the brief, along with the client's change request in their own words, so the atoms work from what the client asked for rather than your summary of it. When the proposal is finished, it comes back to you for review.

That's when you ask the four questions. Can you find the client version, see what changed and follow the members' area estimate back to where it came from? If the estimate doesn't hold up, you find out before the client does. You still know this client better than anyone, including whether they'll accept a higher price for the members' area they asked for, and that judgment is yours to bring to the review.

Try it on work you've already received

Pick one recent piece of AI work and hand it to the person it was meant for. Ask them to find it, tell you what it's for and say what they'd do next. Watch where they hesitate. Whatever stopped them belongs in the next assignment.

Give your workforce its first assignment at Supanova.

Sources

AI reviewAI workAI deliverablesHuman-AI collaboration