Agents reshape the work while data moves between tools

Two threads ran side by side on Tuesday. One is about who does the work as software agents take on more of it, from writing code to answering support tickets; the other is about whether your data stays intact and usable as it moves between the tools that do that work.

For a business choosing software, both threads land in the same place. The tool you pick changes what your people spend their time on, and it changes how hard it is to leave later. The announcements below are worth reading in that light, not as product news but as signals about where the trade-offs are moving.

The developer's job shifts from writing to directing

GitHub framed the change plainly: the role is moving from coder to orchestrator. In its own words, "Developers are owning more of the delivery system around code, not just code itself." [1] That is a short sentence with a large implication. If agents can draft, test, and ship more of the routine work, the scarce skill stops being the typing and becomes the judgement: deciding what should be built, reviewing what an agent produced, and owning the system that moves it to production.

This matters when you choose tools because it changes what you are buying. A tool that only speeds up the keystrokes helps a shrinking part of the job. A tool that helps a person direct, review, and correct the work helps the part that is growing. The same reasoning applies well beyond engineering. Any team being sold "automation" should ask what the human is left holding, because that is the part the tool has to support. We have written before about what AI should and should not do in your business, and the orchestrator framing is the same argument from the other side: the machine does more of the doing, so the person has to do more of the deciding.

Deflection is a tempting number and the wrong one

Pylon's founders used their SaaStr AI Day session to push back on a metric that a lot of support teams quote. The description of the talk is direct: "Marty Kausas and Advith Chelikani on why deflection rate is the wrong number in CX." [2] The headline fact around the session is that the team "Deflected 50% of Tickets Without Increasing Headcount" [2] — a result most buyers would want. The useful part is the caution attached to it.

Deflection rate measures how many tickets never reached a human. It feels like progress because it falls when automation works. But it says nothing about whether the customer got what they needed. A reply that closes a ticket and a reply that solves a problem both count the same way. If you are shopping for a support or inbox tool, the lesson is to look past the metric a vendor leads with and ask what it leaves out. A high deflection figure sitting next to a rising number of repeat contacts is not a win. This is the same discipline we described in what good support data looks like: measure the outcome you actually want, not the one that is easiest to report.

Spreadsheets should cross tools without breaking

Two of Tuesday's announcements were about something less dramatic and more immediately practical: moving data between tools without losing its shape. Google said of its Sheets update, "We're introducing two key improvements in Google Sheets that preserve formatting and linked data when converting files from Microsoft Excel." [3] The change brings Excel tables across as Sheets tables and carries linked pivot tables with them, rather than flattening the file into static values.

That sounds minor until you have lived through the opposite. When a spreadsheet moves between tools and loses its formulas, its tables, or its links, the work of rebuilding it falls on a person, and the risk of a silent error rises. Preserving structure on import is a portability feature even when nobody calls it that. The question for a buyer is not only "can I get my data in" but "does it arrive still usable." This is the practical edge of a principle we keep returning to in the data you should be able to export on any Tuesday: data you cannot move cleanly is data you do not fully control.

The second Sheets change points the same direction for larger datasets. Google is "introducing two usability enhancements to Connected Sheets to improve data presentation and give more flexibility when analyzing BigQuery data," [4] including list parameters that let a query reference more than one value. The theme across both is reducing the friction of working across a database and a spreadsheet. When tools make that crossing easier, you keep more of your options; when they make it harder, you are quietly locked in.

Where the AI market is actually growing

Stripe published its reading of demand, and the framing is a reminder that the AI story is now a business story with geography. The company wrote, "We analyzed Stripe data to understand where global demand is the strongest, and how companies can build to best capture that demand." [5] For a business choosing tools, a payments processor's view of where money is moving is a different signal from a vendor's own growth chart. It describes the market a tool is being built for, not the tool itself.

The practical takeaway is modest but real. If you are evaluating an AI-heavy product, it helps to know whether its vendor is building for the places your customers are, and whether the demand it is chasing overlaps with yours. A product racing to serve a market you are not in may make choices that do not fit you.

A large acquisition is a signal about your vendor's stability

SaaStr's look at Palo Alto Networks is a security story on its surface, but the number that matters to any software buyer is the consolidation underneath it. The company reported $11.4 billion in revenue with 60% ARR growth and 120% net revenue retention, alongside a $25 billion acquisition. On the market's response, SaaStr noted, "The stock is up roughly 100% over the past twelve months and hit an all-time high of $368.80 in July." [6]

The reason this belongs in a tools briefing is that acquisitions reshape the products you depend on. When a vendor doubles in size by buying another, your contract, your roadmap, and your support path can all change without you choosing any of it. A 120% net revenue retention figure tells you existing customers are spending more over time, which is a sign of a healthy product — and also a sign of how much a buyer's spend tends to grow once committed. Neither is bad. Both are worth pricing in before you sign.

The thread, pulled together

The day's stories rhyme. Agents are moving the human up the stack, from doing to directing, which changes what a good tool has to support [1]. The metrics vendors lead with can flatter the wrong outcome, so a buyer has to read past them [2]. And the quiet features — importing a spreadsheet without breaking it, querying a database from a sheet — decide whether your data stays yours as you move [3][4]. For how we think about picking software that respects both your time and your data, see choose software worth using.

360REV is built so your records, conversations, and documents stay exportable and connected rather than trapped in separate tools. That is the whole point of a system that is meant to last past the day you first set it up.