The day the tools changed hands and changed shape

Thursday's productivity news circled a single question: how much control do you really keep over the tools you build your work on. A collaboration platform changed owners, a file service began piping its contents into an outside AI, and a spreadsheet lifted its size ceiling — each story asks, in its own way, whether the tool you commit to today will still be yours to run on your terms tomorrow.

That is the quiet risk in every software decision. You are not only buying features. You are betting on continuity, on governance, and on the tool growing with you rather than against you. The items below are worth a few minutes for anyone who is about to sign a contract or expand a rollout.

A collaboration tool changes owners

When you adopt a tool, you are making a bet on the people who decide its roadmap, its pricing, and the fate of your data. An acquisition quietly rewrites that bet. The product you chose may be run by a different company with different priorities within a quarter, and you will not have been asked.

That is what makes the Miro news relevant beyond the deal pages. Bending Spoons is buying the collaboration tools maker Miro for $1.36 billion, a figure that sits far below the $17.5 billion the company was valued at in late 2021 [1]. The valuation story will get the headlines, but the operational question for a customer is simpler: who signs my renewal next year, and do they intend to keep investing in the parts of the product I rely on.

There is no way to prevent a vendor from being acquired. What you can do is make sure a change of ownership never traps you. That means knowing, before you commit, exactly what you could carry out of the tool if you had to. This is why we keep returning to the data you should be able to export on any Tuesday: the test of a healthy dependency is whether leaving is possible, not whether you expect to.

A file service starts feeding an outside AI

The second thread is about where your documents go once you let an AI help with them. Connecting a file store to an AI assistant is convenient — the assistant can answer from the context of your actual work rather than from generic knowledge. It also means your files, or representations of them, now travel to a second system with its own retention and access rules.

Dropbox has made this kind of link explicit. Working with OpenAI, the company says it is "making it easier to bring the context behind your work directly into your AI workflows" [2]. Read that as a feature and it is genuinely useful; read it as a data-governance decision and it deserves the same scrutiny you would give any new place your files can end up.

The practical lesson is not to avoid these connections. It is to treat each one as a deliberate choice about which systems may see which data, and to keep a record of what is connected to what. Integrations that quietly multiply are how a business loses track of its own information. We have written before about why your tools do not talk to each other, and the flip side applies here: when they do start talking, you want to have decided the terms rather than discovered them later.

A spreadsheet raises its ceiling

Most businesses run far more on spreadsheets than they admit. The spreadsheet is the first tool for almost every job, and it keeps working right up until the day the dataset is too large, the formulas too tangled, or the shared file too contested. A higher capacity ceiling delays that day without removing it.

Google has made a larger cell limit in Sheets generally available, framing it as part of an ongoing effort to keep the tool "a powerful, responsive, and scalable spreadsheet tool for your needs" [3]. For a team that keeps bumping into limits, this is welcome headroom. It is also worth reading as a signal: if you are pushing a spreadsheet hard enough to care about its maximum size, you are probably past the point where a spreadsheet is the right home for that data.

Headroom buys time, not a different outcome. A spreadsheet has no record of who changed what, no enforced structure, and no permissions beyond sharing. Those gaps are exactly what you feel when a spreadsheet stops being enough, and a doubled cell count does not close any of them.

Governance moves into the admin console

As more work passes through AI features, the question of who can share what, and with whom, stops being a niche concern. The useful development is when that control becomes granular and lives where an administrator can actually set it, rather than being a single on-or-off switch.

Google has done exactly that for one of its AI surfaces: administrators can now enable external sharing for Gemini Notebook "using granular controls in the Admin console" [4]. The detail that matters is the move from a blunt toggle to fine control. A blunt toggle forces a business to choose between no external collaboration and unmanaged external collaboration. Granular control lets you say yes to the specific cases you trust and no to the rest.

This is the unglamorous side of enterprise software, and it is the part that protects you. If you cannot see and set who shares your AI-assisted work outside the company, you do not really control it. That is close kin to the audit trail nobody thinks about: the controls you never notice are the ones doing the work when something goes wrong.

Checking what the AI wrote

The more an assistant generates on your behalf, the more your real job becomes reviewing its output rather than producing it. That shift is easy to underrate. GitHub notes that checking agent-generated code "usually means hopping between tabs" [5] — a small friction that, repeated hundreds of times a day, quietly determines whether a team actually reviews what the AI produced or just accepts it.

The broader point outlives the specific tool. Any workflow that lets an AI draft, write, or act needs an obvious, low-effort place to inspect the result before it ships. If reviewing the output is harder than generating it, people will stop reviewing. This is the practical edge of the question we raised in what AI should and should not do in your business: the boundary is only real if checking the work is easy.

Reliability is a feature, and vendors are starting to say so

Finally, a reminder that availability is part of what you buy. GitHub's monthly report states plainly that "in August, we experienced five incidents that resulted in degraded performance across GitHub services" [6]. The number is less important than the habit: a vendor that publishes its incidents gives you something to judge, which is more than you get from a vendor that stays quiet.

When you choose a tool your business depends on, ask whether it reports its downtime at all. A public, honest incident record is not a weakness. It is evidence that the people running the service treat your uptime as their responsibility — which, on a day when tools are changing hands and changing shape, is exactly the kind of commitment worth paying for.