Our 2021 SaaS predictions, graded against 2026 evidence

In August 2021 we published ten predictions for the SaaS market. Most "future of SaaS" posts are never checked again. Five years on, we went back to ours, graded each one against public data from 2025 and 2026, and wrote down why the ones we can't grade fail. The short version: two were right, two were partly right, one described what had already happened, and five were too vague to be wrong. Those five turned out to be the most useful part of the exercise, because they show how not to write a prediction.

The scorecard

Prediction 3: the rise of AI. Right, but slower for small businesses

In 2021 we wrote that AI would change how SaaS companies design, market and sell. The direction was right; the pace depends on who you ask.

The US Census Bureau's survey of businesses is the broadest measure we found. From December 2025 to May 2026, the share of US businesses using AI in any business function hovered between 17% and 20%, and was 19.8% on 3 May 2026. It was 37% for firms with 250 or more employees, and under 20% for firms with four or fewer (US Census Bureau). The Information sector was highest at 39.7%, and retail around 14%.

The size split explains why AI adoption figures disagree so much: a survey that mostly asks large companies, or tech companies, will find far higher use than one of all businesses. Both pictures are true at once: AI is everywhere in large software companies and still a minority habit in small businesses. For a small SaaS selling to small businesses, that gap matters more than the headline. Your customers may not be ready for an AI feature you'd assume is expected.

AI also changed SaaS pricing, which our 2021 post hinted at in its description but didn't predict. GitHub moved every Copilot plan to usage-based billing on 1 June 2026, with credits consumed by token usage while base plan prices stayed the same (GitHub). We compared how 12 tools define a credit in credits, seats or usage, and worked out what an AI feature really costs per user.

Prediction 4: micro-SaaS. Partly right, for the wrong reasons

We predicted that micro-SaaS was the future, and described it as low-complexity products on a "freemium" model. We also quoted a forecast of 6 million US micro-SaaS customers by 2022. We couldn't find a source for that number, so we've withdrawn it.

What the evidence supports: small, focused, self-funded software companies can do very well. Plausible Analytics, a self-funded analytics product, passed $1 million in annual recurring revenue in 2022 and now lists more than 22,000 paying subscribers, with a team of ten (Plausible). It charges from the start rather than relying on a free tier, which is the opposite of the freemium model we described.

What it doesn't support: that micro-SaaS is "the future" of SaaS as a whole. ChartMogul's 2025 study of 6,525 software companies found that about a quarter reached $1 million ARR within five years of first revenue (ChartMogul). Small software companies are a real path, not a market shift. Our micro-SaaS guide has the numbers in more detail.

Prediction 7: APIs. Right, with a twist we didn't see

We wrote that APIs were the next big thing in SaaS. Postman's 2025 State of the API report, a survey of more than 5,700 developers, architects and executives, found 82% of organisations had adopted an API-first approach to some degree, and 25% were fully API-first (Postman).

The twist is who calls the APIs. The same report found that only 24% of developers design APIs with AI agents in mind, and that unauthorised or excessive calls from AI agents were their top security worry, at 51%. In 2021 we pictured APIs connecting businesses to each other. In 2026 the new consumer is software acting on a user's behalf. For a small SaaS, that means rate limits, scoped keys and clear errors are no longer optional extras.

Prediction 9: the long tail. Partly right

We predicted more startups would serve niches. That is what happened among small, self-funded companies, and the tools to build them got much cheaper (we compare the main AI builders in build an app without coding in 2026). But the 2021 post's claim that a niche product's revenue "will be more than enough" was too confident. The ChartMogul data above shows most software companies take years to reach a meaningful size, niche or not.

The six we can't grade, and why

• 1 and 2 ("will keep evolving", "balance of power will be disrupted"). No measure and no date: any outcome fits. A prediction that can't be wrong carries no information.
• 5 (software over hardware). Our evidence was a Gartner forecast for 2015, already six years old when we wrote it. We tried to check Gartner's 2026 forecasts, but couldn't open the pages, so we leave it ungraded rather than guess.
• 6 (R&D pays off). Companies that invest in products tend to have better products; that's close to a definition. We also wrote that SaaS sales grew only 1% in 2017, and we can't find where that number came from, so we've withdrawn it.
• 8 (commoditisation). Our own 2021 text said Microsoft and Adobe had already moved to subscriptions. That's a description of the past, not a prediction.
• 10 (big companies partnering with small SaaS). Our example was a partnership between Oracle and HighJump Software. We couldn't find a record of it, so we've withdrawn it.

How to write a prediction that can be graded

The exercise taught us more than the grades. A useful prediction has four parts:

1. A measure: something counted by someone else, such as a government survey or a vendor's published report.
2. A number: "more than 30%", not "more".
3. A date: "by the end of 2028", not "soon".
4. A source you'll check: decide now where you'll look, so you can't pick a friendly source later.

For example, "AI will change SaaS" can't be graded. "The Census Bureau's survey will show more than 30% of US businesses using AI in any business function by the end of 2028" can. We'll hold ourselves to that format from now on.

If you're planning around trends, the free sales forecast calculator and startup runway calculator turn assumptions into numbers you can check later, which is the same discipline on a smaller scale.

Sources

• US Census Bureau: Large firms with at least 20 employees biggest AI users (26 May 2026).
• GitHub: GitHub Copilot is moving to usage-based billing (27 April 2026).
• Plausible Analytics: About (read 29 October 2026).
• ChartMogul and Kyle Poyar: SaaS growth: the odds of making it (15 October 2025).
• Postman: 2025 State of the API Report.
• Our 2021 post, which this page replaces: its claims were checked one by one, and those we couldn't source are marked as withdrawn above.