What today's tools let you see, not just what they do

The announcements that landed today have less to do with new features than with an older question: when a tool takes on more of your work, how much of that work can you still see. A business choosing software this year is really choosing how much visibility it keeps — over its support queue, its code, its books, and the tasks it now hands to software instead of people.

That is a useful lens because it cuts against the instinct to buy on capability alone. A tool that does more but shows you less can leave you worse off than the one it replaced. The items below are ordered by how much they matter to that decision.

When you hand off support, measure what you handed off

Most support metrics were designed for a queue worked by people. First response time, tickets closed per agent, average handle time — these describe how fast humans move through a list. When a large share of conversations is answered without a person touching them, those same numbers still report, but they report on a shrinking slice of what actually happened. Intercom's argument today is that the measurement gap grows exactly as the automation does: "As your Agent takes on more, the gaps in what you can see get bigger." [1]

The practical response is to change what you count. Instead of speed-of-first-reply, measure whether the customer's problem was resolved and whether they came back about the same thing. Instead of tickets-per-agent, watch the escalation rate and what triggers it — the conversations software handed to a person are a map of its limits. And keep a sample of automated answers under human review, because a metric that only aggregates cannot tell you a confident wrong answer from a correct one. If you are picking a support tool this year, ask what it lets you inspect after the fact, not only how much it can close. We wrote separately about what good support data looks like, and the short version is that the useful record is the one that survives the handoff.

Close the easy doors before you add anything clever

Most projects are not compromised by a novel attack. They are compromised through a default someone never changed — a setting that was there to be turned on and never was. GitHub's post today is a plain checklist of six settings a maintainer can enable, and it is careful not to oversell them: the settings will not make a project unhackable, but, in the post's own words, "What they will do is close the easy doors." [2]

That framing is worth carrying into any tool you evaluate. Ask what a product's defaults are, and who has to act to make them safe. A tool that ships open and asks you to lock it down puts the burden on your memory; a tool that ships closed and asks you to open what you need puts the burden on an explicit decision. Neither is wrong, but the difference determines how many easy doors sit open across your account a year from now. A checklist like this one is cheap to work through and belongs on the calendar of anyone who owns a repository, not filed as reading for later.

Financial numbers are worth more where the decision is made

Every small business owner and advisor knows the tax of moving figures by hand. You are writing a report or a board slide, the numbers you need live in the accounting system, and you copy them across — losing a few minutes each time and introducing the chance of a stale figure that no longer matches the books. Xero framed its Microsoft 365 work today around exactly that friction, noting that "As a small business owner or an advisor, you likely often find yourself creating a presentation or report where you need to include key financial performance data." [3]

The general principle underneath the specific integration is that data is most valuable at the point where a decision is written, not in the system that happens to store it. Any time you paste a number from one tool into another, you have created a copy that can drift, and a task that has to be repeated every time the source changes. When you assess a tool, look at where its numbers can go without a human retyping them — the value is not the figure, it is the figure arriving current in the place you were already working. This is the same reason we keep arguing that the numbers that change a decision have to be trustworthy at the moment you read them, or they are worse than no number at all.

Fewer moving parts can be the goal, not more

The prevailing story about automation is that you will run a fleet of narrow specialists, each doing one job. SaaStr reported the opposite pattern from its own stack today, opening with the received wisdom before contradicting it: "Everyone keeps telling you the future is 100 specialized agents." [4] In their case the pieces are collapsing into fewer, broader ones — including finance work moving onto automatic.

For a buyer, the count is not a vanity number; it is a visibility cost. Every task you delegate to software is another surface you have to watch, another place a quiet failure can hide, another set of outputs to reconcile. More narrow tools can mean more seams between them, and seams are where work falls through. The question to ask of any automation is not how many jobs it can take, but whether you can still tell, at a glance, what it did and whether it did it correctly. That is the line we drew in decisions automation should never make: handing off a task is fine, handing off the ability to check it is not.

A growth rate means little without the multiple beside it

Finally, a reminder that a single number rarely settles anything. SaaStr looked at public software valuations today and used Figma as the anchor, recounting that "Figma went public on July 31, 2025 at $33 a share, popped 250% on day one to close at $115.50, and got valued near $68 billion." [5] The wider point in the piece is that companies growing at similar rates can trade at very different multiples — the market is pricing something beyond the headline growth figure.

This matters to a business choosing a vendor as much as to an investor. When you evaluate the company behind a tool, a growth rate on its own tells you how fast it is adding customers, not whether it can keep them, fund the product, or still be there in three years. Read the number next to its context — retention, margin, the multiple the market assigns — the same way you would read your own metrics. A figure without the second number beside it is a story, not evidence.

The thread across all five is the same. The tools worth choosing this year are the ones that take work off your plate without taking the work off your screen. Capability is easy to demonstrate in a demo; visibility is what you rely on every ordinary Tuesday after the demo is over.