AI is moving into the tools you already use

Three releases this week point the same way. Claude now works inside Google Docs, Sheets, and Slides. Anthropic launched tools that turn a prompt into an editable chart or a short animated explainer. And Microsoft's Copilot started routing each task to whichever of its own models is strongest, sending a spreadsheet job to one and a slide draft to another.

What they share is that AI now lives in the documents and tools where people already work, rather than waiting in a chat window they open on the side. It's becoming an integrated piece of the work.

For your team, that resets two things. First, production. A chart or an explainer that used to be a design ticket is now a prompt inside the document. The work shifts from creating the asset to deciding what it should say and ensuring it holds up. Second, tool choice. When the tool picks the best model on its own, you're choosing a workflow and a set of data connections. The model brand is no longer the deciding factor.

As a marketing leader, start thinking about where the next AI tool sits in your team's actual workflow and what it can see, rather than the model behind it. As production work coalesces into a few tools people use all day, the edge moves to taste, to deciding what's worth putting out at all, and that's the part the tool won't do for you.

The slop grenade

Shopify's CEO Tobi Lutke named a pattern this week that your team is already living. He calls it the slop grenade: a report polished by AI to look finished, handed to a colleague who then burns a couple of hours working out what's actually solid underneath. One survey found roughly 40% of US desk workers getting work like this. FactSet, meanwhile, told financial institutions that an AI answer sounding convincing no longer means it's correct. They advised institutions to start demanding accuracy that they can check.

When anything can be made to read well in seconds, reading well stops proving a person did the thinking. The scarce signal becomes visible judgment. Readers are desperate to hear that signal, especially when making high-stakes buying decisions. Is there a named expert who'll stand behind a claim? Can the reader check a method and trace the numbers to a source?

The practical fix is to make work easy to check before it leaves your team. Put a real person's name on it, the one who'll answer for the claim. Put the source right next to the number, so a reader can confirm it with a click instead of taking it on faith. The harder discipline is reading your own AI-assisted draft with discernment, lest you become the one who threw the slop grenade.