Why this project exists, what the data says, and whether the widely-cited industry numbers hold up.
In 2026, two questions dominate SaaS pricing: is per-seat pricing dying, and how should companies charge for AI? The industry answers with confident statistics. A widely-cited one, from an ICONIQ Capital survey published by BCG, is that 68% of vendors charge separately for AI or restrict it to their premium tiers.
The trouble is these estimates measure different things and don't agree — and the highest, most-repeated figures come from aggregator pages with no stated method. Line up the two sources that do name a method:
| Source | What it measures | Figure |
|---|---|---|
| Kyle Poyar, Growth Unhinged (survey, n=230) | Actually meter AI via credits | ~29% |
| ICONIQ Capital (via BCG, 2025) | Charge separately for AI, or restrict it to premium tiers | 68% |
That is more than a 2× gap between two named sources, driven mostly by definition — one counts a live AI usage meter, the other also counts “AI is only in the premium tier.” And unattributed aggregator pages push the figure as high as 73% with no stated method at all. No single number is settled, and there was no public dataset to check them against.
| Primary pricing model | Companies | Share |
|---|---|---|
| Per-seat | 81 | 42% |
| Hybrid (seat + usage) | 44 | 23% |
| Contact-only (no public price) | 25 | 13% |
| Usage-based | 23 | 12% |
| Flat-rate | 18 | 9% |
Per-seat is still the largest single model, but a majority of companies now price some other way. The detail that matters: the second-place model is hybrid, not pure usage. The market is not swapping seats for meters — it is layering usage charges on top of a seat-based base. More companies hide their price entirely (25) than run a pure meter (23).
| How AI is charged (of the 166 that mention it) | Companies | Share |
|---|---|---|
| Bundled — included in the plan you already buy | 94 | 57% |
| Metered AI credits / tokens | 35 | 21% |
| Only in a more expensive tier | 17 | 10% |
| Sold as a separate add-on | 12 | 7% |
| Unclear from the page | 8 | 5% |
87% mention AI, but only 34% of all 191 companies actually charge extra for it. Bundling (94 companies) outweighs all three paid mechanisms combined (64). And among those who do charge, metering is the clear preference — selling AI as a separate add-on is the least common approach of all.
Single sign-on is a basic security feature. Of the 124 companies that offer it, 86 (69%) require their top paid tier or an Enterprise contract to get it.
| Cheapest plan that unlocks SSO | Companies | Share |
|---|---|---|
| Enterprise / contact-sales tier | 57 | 46% |
| Top paid tier | 29 | 23% |
| Mid tier | 25 | 20% |
| All tiers | 12 | 10% |
This is a floor, not a ceiling. When a page doesn't mention SSO we record it as “not mentioned” rather than assuming it's gated — so the true rate can only be higher than 69%, never lower.
Median company offers 4 tiers, discounts annual billing by 20%, and starts paid plans at a median of $10 per month. 82% publish a real price, yet 85% still keep a “contact sales” tier above it.
This needs a precise answer, not a slogan.
We are not claiming the published reports are lying. We are claiming three specific things:
34%
On a checkable sample of 191 leading SaaS companies, roughly one in three charges extra for AI — not the widely-cited three in four. The most common decision in the market today is to bundle AI into the existing subscription and compete on it, rather than sell it as a separate line item.
And on the other headline question: per-seat pricing is not dying — it is becoming the base layer of a hybrid model. Meanwhile 69% of vendors gate basic security behind their most expensive plan.
What this is useful for. A product manager deciding whether to charge for a new AI feature now has a market answer: most competitors don't, and charging makes you the outlier. A founder setting pricing has benchmarks (4 tiers, 20% annual discount, $10 median entry). A buyer negotiating an Enterprise quote can see that the SSO upcharge is standard practice across 69% of vendors, not a special case. An investor can check whether a company's pricing is normal for its category. None of that was answerable from public sources before.
Every headline number was recomputed through three independent paths — a direct recount from the raw scraped files, the analysis pipeline, and the exported dashboard dataset. All agree exactly.
| Metric | Raw recount | Pipeline | Dashboard CSV | Result |
|---|---|---|---|---|
| Usable companies | 191 | 191 | 191 | PASS |
| Per-seat | 81 | 81 | 81 | PASS |
| Mention AI | 166 | 166 | 166 | PASS |
| Bundle AI | 94 | 94 | 94 | PASS |
| Charge extra for AI | 64 | 64 | 64 | PASS |
| Mention SSO | 124 | 124 | 124 | PASS |
| SSO gated to top tier | 86 | 86 | 86 | PASS |
| Show a public price | 157 | 157 | 157 | PASS |
| Offer a free tier | 109 | 109 | 109 | PASS |
Logic checks also pass: pricing models sum to 191; every company recorded as charging for AI also mentions AI; every SSO-tax record also offers SSO. Reproduce with scripts/aggregate_clean.py → scripts/analyze.py → scripts/verify.py.
Sources for comparison figures: ICONIQ Capital 2025 software survey, as published in BCG, Rethinking B2B Software Pricing in the Era of AI (2025); Kyle Poyar, Growth Unhinged — The 2026 State of B2B Monetization (survey of 230 companies, 2026). Higher unattributed figures (e.g. 73%) circulate via SaaS-management aggregator pages and are noted but not relied on. Primary data: this study, 191 companies, self-collected from live pricing pages, 2026.