Where AI value sits in the tools you buy

Monday's announcements circle a single idea — the worth of AI inside business software depends less on the feature you can see and more on the data underneath it, and on the way that software is priced. From a company staking its future on one connected database to a spreadsheet assistant that reads the cells around your error, the day is about where the value actually sits.

That is a useful lens when you are choosing tools, because it moves the question away from feature lists. A demo can show you what a product does on a clean screen. It cannot show you whether the product will still be useful once your own messy data flows through it, or whether the price you pay this year still makes sense the year after. The four items below each speak to one of those two questions: what the data underneath is worth, and what you are really paying for.

The moat is the data, not the feature

The most instructive announcement of the day is a claim about strategy rather than a shipped button. Rippling described its approach to AI as resting on one foundation: "Rippling's whole AI bet comes down to one thing: a single connected database under every product it sells." [4]

It is worth slowing down on why that matters, because it is easy to read as a technical detail and miss the point. AI is only as good as what it can see. If your payroll tool holds one version of an employee record, your HR tool holds another, and your device-management tool holds a third, then any assistant you point at that mess has to guess at the parts it cannot reach. It will be confidently wrong in exactly the places where the data disagrees. When the same record sits under every product instead, the assistant reads one truth and its answers get sharper without anyone doing extra work.

This is the real reason connected data is described as a moat. It is hard to copy, not because the idea is secret, but because most software estates grew one tool at a time and never joined up. A rival can ship the same feature in an afternoon. It cannot retrofit a shared record across systems that were never built to share one. For a business choosing tools, the lesson is to ask where each product keeps its copy of the truth, and what happens when two products describe the same customer or employee differently. We wrote about that gap in why your tools do not talk to each other, and about the discipline of one clean record in what a customer record should contain. 360REV is built around a single record precisely so that the AI and the reporting sit on the same facts rather than reconciling several.

AI where the work already is

The second item is smaller in ambition and closer to daily life. Google announced a Gemini capability inside Sheets aimed at formula errors: "We're excited to introduce a new Gemini in Sheets capability that enables you to diagnose and fix formula errors in one click." [2]

The concept here is that AI is most useful when it meets people inside the work they are already doing, rather than in a separate window they have to remember to open. A spreadsheet error is a good example. The person hitting it is usually mid-task, not looking for an assistant, and the cost of the error is a wrong number that quietly flows into a report. Putting the help at the point of failure removes the step where someone copies the formula out, describes the problem elsewhere, and pastes an answer back.

There is a trade-off to name, and it is not a criticism of the tool. When a fix arrives in one click, the person applying it may never learn why the formula broke. That is fine for a throwady calculation and less fine for a model that other decisions depend on. The judgement a business has to keep is knowing which errors it wants understood and which it is happy to have quietly corrected. That line — between help you accept without thinking and help you should inspect — is the same one we drew in what AI should and should not do in your business. Convenience is real value. It is worth being deliberate about where you spend it.

What you are really paying for

The third item is about money rather than features, and it frames the whole day. Writing about software valuations, SaaStr set the scene bluntly: "Four months ago the markets decided B2B software was dead." [3] The piece goes on to weigh whether names like Salesforce, HubSpot and Adobe have been marked down too far, but the part that matters for a buyer is the reason given for the fall — the fear that AI would make per-seat pricing obsolete.

Per-seat pricing is the model most buyers know: you pay for each person who logs in. It made sense when the value of software was a human sitting at a screen using it. The worry the market is expressing is what happens when much of the work is done by an assistant rather than a seat. If one person plus an AI does the work of five, do you still pay for five seats, or for the work that got done. Nobody has settled that question yet, which is exactly why it is showing up in valuations.

You do not need a view on the stock market to take something from this. When you choose a tool, look at what its price is attached to. A price tied to seats rewards you for keeping headcount low and punishes you for adding casual users who would benefit from access. A price tied to usage or outcomes behaves differently as you grow. Neither is wrong, but they age differently, and the one that suits you this year may not suit you in two. We pulled that apart in how pricing pages fail and in what a subscription plan is really selling. The market's nervousness is just this question, asked with more zeros.

Tools everyone on your team can use

The last item is the quietest and the easiest to skip, which is why it is worth ending on. GitHub published an update on accessibility in open source, framed as a progress report: "Learn about the progress we've made toward our accessibility goals and how you can help make open source more inclusive." [1]

Accessibility is usually filed under compliance, which undersells it. For a business choosing tools, it is a plain question of who on your team can actually use the thing you bought. A tool that a colleague with a screen reader cannot navigate, or that fails at high zoom, or that cannot be driven from a keyboard, is a tool you have paid for and then locked part of your staff out of. That is a cost, even when it never appears on an invoice.

The reason it belongs in the same briefing as connected data and pricing is that all three are about value that a feature demo hides. A demo runs on clean data, at a comfortable seat count, on the reviewer's own screen. It does not show the reconciliation you will do later, the bill you will get as you grow, or the colleague who cannot get in at all. Reading an accessibility update as a buying signal — is this vendor treating access as a goal it reports on — tells you something a feature tour will not.

The thread, in one line

Four announcements, one idea underneath them. The value of business software is drifting away from the visible feature and towards the things a demo cannot show you: the data the tool sits on, the price it is attached to, and whether everyone you employ can use it. When you next sit through a product tour, those are the three questions to bring with you, because they are the ones that decide whether a good demo turns into a good decision.