The AI stress-test for software stocks: a simple framework to spot disruption risk

That gap is the plot twist. It says the average stock is doing better than the headline index suggests. It also says leadership is rotating, and the market is asking a blunt question: if artificial intelligence (AI) can do the job, why pay for the seat?

The double hit: substitution fear meets spending fatigue

Software sells a promise: pay now, become more productive later. AI makes that promise both easier to deliver and harder to charge for.

The first hit is substitution. Investors worry that some tools become “good enough” features inside bigger platforms, or tasks handled by AI assistants. The second hit is spending fatigue. The same market that cheered “AI investment” in 2024 now worries about the size of the bill. Reuters notes that big technology firms’ AI-related spending plans, around 600 billion USD, add to investor unease.

Put these together and you get a tough cocktail for tech: pricing pressure on the revenue line, and higher costs in the background. Even professional allocators describe early 2026 as rotation away from AI-related technology and software, with only a few S&P 500 sectors down year to date, including information technology.

This is why “all software is the same” trades stop working. The market starts sorting. And it sorts fast.

Why breadth looks better, even when your watchlist looks worse

The equal-weight index outperformance is not trivia. It is a signal that the market’s centre of gravity is shifting.

The equal-weight version’s lead suggests more stocks participate in gains, even if the biggest names drag on the cap-weighted index.
That can be healthy for long-term investors, because it reduces reliance on a handful of giants to carry returns.

But there is a catch: improved breadth often comes with higher dispersion. Some areas feel calm. Others get repriced in days. Software sits in the second camp right now.

So the useful question becomes practical, not philosophical: how do you stress-test software exposure when AI changes both what the product does and how it gets paid for?

The core idea: observable proof beats narrative

Generative AI shifts software from “sell more seats” to “sell more outcomes”. That changes pricing, competition, and retention. The goal is not to predict winners. It is to spot, early, whether AI is turning into paid value or quietly becoming a cost line.

The five-lens framework we use to screen AI disruption risk:

1) Seat-heavy pricing exposure
Companies that charge mainly per user can face slower seat growth if AI lifts productivity or headcount trims.

2) Bundle-away risk
Tools that solve a narrow task can be replaced more easily, or bundled away by larger suites.

3) “Good enough” substitution risk
In creative and content-related work, AI can make basic output cheap, pushing some users toward free or low-cost alternatives.

4) Small and mid-sized business sensitivity
Budget cuts often hit smaller customers first, which can increase churn, downgrades, and discounting.

5) AI agent disruption risk
Some categories may be reshaped by AI agents, especially automation, customer support, work management, and parts of developer tools.

To make this practical, we have built an internal shortlist that groups software names by disruption risk and the metrics to watch.

These signals map to two camps. “Disrupted” businesses tend to be seat-heavy point tools in workflows an AI agent can complete end-to-end. “Disrupters” more often own distribution, a system of record, or trusted data, where AI can deepen the moat, but only if monetisation keeps up with the AI bill.

Risks worth respecting (because the future enjoys surprises)

First, AI disruption is not one-way. Companies can adapt pricing, rebuild products around outcomes, and re-bundle into stickier suites. A stock can look “disrupted” and still execute brilliantly.

Second, timing is messy. Markets can punish on fear long before the business weakens, or ignore weakness until it is obvious. That is why receipts matter, but also why patience and position sizing matter.

Third, the AI spending cycle can cut both ways. If data centre spending slows, suppliers feel it. If it accelerates without clear payback, valuations can still compress.

Investor playbook: how to use the shortlist as a tool, not a tip sheet

• Treat software like a “pricing model” bet: track how revenue per customer evolves as AI features roll out.

• Watch for paid AI adoption signals in results, not just AI mentions in slides.

• Compare retention and churn trends across tools that sell seats versus tools that sell outcomes.

• Keep a scenario view: if bundling spreads, who loses pricing power first, and who owns distribution?

Conclusion: the quiet skill is knowing what to measure

AI makes software feel less like a product and more like a moving target. That is why 2026 punishes broad labels like “tech” and “software” and rewards specifics.

The equal-weight gap with the S&P500 tells you the real story: markets are rotating, and they are sorting. In software, the sorting question is uncomfortable but useful: is AI a feature upgrade, or a pricing reset?

Our five-lens framework is designed to make that question measurable. Not certain. Not predictive. Just measurable. Because in a receipts market, inspiration is good, but evidence is better.