The software market has spent two years treating AI as a rising tide that lifts every boat. That story is ending. According to The Information, the winners and losers of the AI software wave are finally coming into focus, and the gap between them is widening fast.
What stands out here is the shift in question. For most of 2024 and 2025, investors asked which software companies would “add AI.” Now they’re asking a harder one: which companies AI makes stronger, and which ones it quietly makes obsolete. The Information frames this as a sorting process, and it’s the right lens. The market is separating firms that own real distribution and proprietary data from those whose main value was a nice interface sitting on top of a model they don’t control.
Why the split is happening now
Three forces are converging at once.
- Model capability keeps climbing. Features that were a startup’s entire pitch a year ago now ship as a checkbox inside a foundation model. If your product was a thin wrapper, the wrapper is the product, and that’s a problem.
- Incumbents woke up. The big platforms with existing customers, billing relationships, and mountains of usage data can bolt AI onto what people already pay for. Distribution beats novelty most of the time.
- Buyers got disciplined. The experimental budget is drying up. Companies now want AI that shows up in revenue or cost, not a pilot that impresses in a demo and dies in procurement.
Who’s winning
The pattern The Information points to lines up with what’s visible across the market. The likely winners share a few traits:
- They own proprietary data competitors can’t replicate.
- They sit inside a workflow customers can’t easily rip out.
- They charge for outcomes, not seats, so AI that does more work raises their revenue instead of threatening it.
Seat-based software is the quiet risk here. If your pricing assumes a human at every desk, and AI means fewer humans doing the same work, your own product is arguing against your growth.
Who’s exposed
The losers aren’t always the obvious ones. The exposed group includes companies whose moat was complexity. If a tool existed mainly because software used to be hard to build or hard to use, AI erodes that reason to exist. Cheap generation and natural-language interfaces turn yesterday’s differentiation into a commodity.
This is significant because it hits mid-market SaaS hardest, the tier that’s too big to pivot fast and too small to out-distribute the platforms.
The future cast: where this goes by 2027
Expect the sorting to accelerate over the next 18 to 24 months. A few things look likely:
- Pricing models flip. More software moves from per-seat to per-outcome or usage-based, because the seat math stops working.
- Consolidation picks up. Wrapper companies with real customers but no moat become acquisition targets, folded into platforms that need the distribution.
- Data becomes the whole game. The durable question won’t be “do you have AI,” it’ll be “do you have data or a workflow nobody else can copy.”
What to do about it
If you build or buy software, a few practical moves:
- Audit your moat honestly. Ask what your product does that a foundation model plus a weekend can’t. If the answer is thin, the clock’s running.
- Rethink pricing before you’re forced to. Tie revenue to work done, not chairs filled.
- Hoard proprietary data now. The workflows and datasets you own today are the defensibility you’ll trade on tomorrow.
- For buyers, favor outcomes. Pay for tools that move a number you care about, and treat demo dazzle as a red flag, not a reason.
The comfortable phase, where everyone got to claim AI as a tailwind, is over. The Information’s read is that the market is done grading on a curve. The companies that treated AI as a coat of paint are about to find out how thin it is, and the ones that rebuilt underneath it are about to pull away. More detail is in the original report at The Information.