Who gets to act inside your tools

The announcements that landed today share a single concern: who, or what, is now allowed to act inside the tools a business runs on, and whether a team can see and afford what that actor does. Payment platforms, office suites and developer tools all moved in the same direction at once, handing more work to software agents while adding the controls, the governance and the cost visibility a company needs before it trusts them.

That is the thread worth reading for if you are the person who decides what your team buys. Each item below is a small piece of the same shift, and the useful question for each is not "is this impressive" but "does this change the trade-off I am weighing when I choose a tool."

Stripe makes currency conversion something you do in one place

Start with the plainest kind of value, because money is the easiest trade-off to understand. A business that sells across borders takes payment in one currency and pays its costs in another. Somewhere between those two facts sits foreign exchange: the rate you are given, the spread a provider keeps, and the settings and timing you have to manage so you are not quietly losing a percentage point on every transaction. For most small companies this is handled by whoever does the books, in a spreadsheet, after the fact. It is tedious, it is error-prone, and it is rarely measured.

Stripe today described two upgrades aimed at that problem: wider multicurrency settlement and the ability to convert currencies instantly [1]. The company's own framing is that the two changes "make it easy for global businesses to manage FX entirely on Stripe" [1]. The phrase to weigh there is *entirely on Stripe*. The appeal is consolidation — one fewer system, one fewer manual step — and the trade-off is the one that comes with every consolidation: the more of a process you move into a single platform, the more that platform's rates, coverage and reliability become your rates, coverage and reliability. That is not a reason to avoid it. It is the thing to check before you commit, the same way you would with any tool you are letting own a step that used to be yours. We have written before about choosing software worth using, and FX is a clean example: the right test is whether the consolidated version is cheaper and clearer than what you do now, measured, not assumed.

A tool that worked for seven years still got replaced

The sharpest story of the day was not a product launch. SaaStr published an account of a company dropping a tool it had used for seven years, with the detail that an internal AI agent had taken over the work it did. The writer's own words set the scene: "We canceled Notion last week after 7 years" [2].

What makes this worth your attention is not the specific product. It is the mechanism. A tool can be well-liked, well-used and entirely functional, and still lose its place because the job it did got absorbed by something the team already runs. This is the quiet version of a decision every business will face repeatedly over the next few years: not "which tool is better" but "does this job still need a tool of its own."

That is a harder question than it sounds, and it is easy to get wrong in both directions. Replace too eagerly and you lose structure you did not know you depended on. Hold on too long and you pay for overlap. The honest way through it is to be specific about what a tool actually holds for you and what would happen to that if it went away — the same discipline we described in when a spreadsheet stops being enough. The lesson from today's account is that satisfaction is not the thing that keeps a tool in your stack. Fit is. When the surrounding system changes, fit changes with it.

Administration becomes something you can ask about

Google Workspace announced that it is bringing AI help into the place administrators work. In the company's words, "We are introducing new admin assist capabilities, bringing two new Gemini-powered capabilities to the Google Admin Console: the side panel and AI overviews" [3].

The concept underneath this matters more than the feature. Administration — permissions, troubleshooting, the long tail of configuration — is where small businesses lose the most time per person, because it is skilled work that nobody was hired to do. The promise of an assistant inside the admin surface is that you can ask a question in context rather than reading documentation out of context. The trade-off is that an assistant is only as good as your willingness to check what it tells you before you act on it. An AI overview of a permission problem is a starting point, not an authority. The value is real, but it is the value of a faster first draft of an answer, and it should be read that way.

Governance arrives before the automation does

The fourth item is the most telling about where all of this is going. Google Workspace also announced new enterprise security controls for Workspace Studio, which it describes as a place where "users to boost their productivity with custom, no-code agentic automation" [4]. The point of the announcement is the controls themselves: "Today, we are adding a new set of enterprise security controls to enable additional collaboration use cases" [4].

Read those two sentences together and you have the shape of the whole day. First you give people the ability to build automation without writing code. Then you discover that automation anyone can build is automation anyone can misconfigure, and you add the identity and data controls that make it safe to let more people use it. The order is the lesson. Governance is not a feature you bolt on after an automation habit forms; it is the thing that decides whether the habit is safe to form at all. A business adopting any no-code automation should ask for the controls in the same breath as the capability, and should decide in advance which decisions a built automation is never allowed to make. We have argued that case directly in what AI should and should not do in your business, and today's release is a vendor reaching the same conclusion from the other side.

You cannot steer what you cannot see

The last item is about a problem that gets worse as agents do more. GitHub published a piece on using canvases with agentic workflows, and its opening line names the difficulty exactly: "Chat is great for intent, but agent work gets lost in the scroll" [5].

This is a small observation with a large consequence. When you delegate work to an agent through a chat interface, the work happens in a stream that is hard to review, correct or cost out. You can tell it what you want; you cannot easily watch what it did or stop it halfway. The author's answer is to give the work a surface of its own — a canvas — so it becomes "visible, steerable, and cost-efficient" [5]. You do not need GitHub's specific tool to take the point. Any agent you let loose in your business should produce work you can see while it happens, not only a result you accept or reject at the end. Visibility is what turns automation from a bet into a managed process. It is also, not coincidentally, how you keep a cost under control: you cannot trim what you never watched.

What ties the day together

Five announcements, one direction. More of the work inside your tools is being done by software that acts on your behalf, and the serious vendors are shipping the controls, the visibility and the cost transparency alongside the capability rather than after it. For a business choosing tools, that reframes the buying question. The old question was what a tool does. The better one now is what it lets act inside your business, whether you can see that actor at work, and whether the system you choose keeps your own data where you can still get it out and reason about it. Connected tools make this easier and riskier at once, which is the tension we unpacked in why your tools do not talk to each other. The tools announced today are each answering half of that question. Your job is to make sure you are answering the other half before you commit.

How we think about it

At 360REV the rule we hold to is that automation should be visible and reversible, and that no agent decides anything a person would want to review. The news today is a sign that the rest of the market is settling on the same standard.