Software vendors that charge for AI grew 10 points faster, KeyBanc survey finds
KeyBanc's 2026 AI x Software Benchmark finds vendors that price AI separately are outgrowing bundlers, while more than half say customers are already rebuilding their product with AI coding tools.

Software companies that charge separately for their AI features grew 27% in 2025, ten points ahead of companies that bundle AI into the base price, according to KeyBanc Capital Markets' 17th annual AI x Software Benchmark, released Nov. 13 and broken down in further detail by SaaStr on Nov. 19. The same survey found that a majority of vendors say their own customers are already using AI coding tools to build alternative versions of what they sell.
Charging for AI tracks with faster growth
KeyBanc surveyed 81 private B2B software companies between July and September, most with between $10 million and $50 million in annual recurring revenue, according to SaaStr's write-up of the data. Of those, 41% charge separately for AI today, 29% are testing or piloting a model, and 30% still don't charge for it at all.
Where vendors did describe their pricing model, the most common choice by far was bundling: of 69 respondents, 31 fold AI into the core product price, 16 sell it as an add-on or separate product, 12 use a hybrid of the two, four charge by usage, and one uses outcome-based pricing. Where AI does carry its own price, the median premium is 20%, though a quarter of vendors charge less than 4% more and a quarter charge more than 26% more for it. That spread is wide enough that KeyBanc's analysts flagged it in the report itself, concluding there is no consensus yet on an AI line item's fair value, according to SaaStr's summary.
Growth splits along the same lines. Companies charging for AI grew a median 27% in 2025 and are planning 30% for 2026. Companies still testing monetization grew 24% and expect that to ease to 22%. Companies not charging separately for AI grew 17% and expect 21% next year — still the slowest cohort of the three. KeyBanc's press release ties a wider gap to the product itself: AI-native companies, selling AI as their primary product, grew a median 34% in 2025 versus 22% for companies embedding AI into existing products, but ran a median EBITDA margin of -12% against -1% for the embedded-AI group.
Customers are already building what you sell
KeyBanc's press release puts a number on the competitive threat directly: 57% of respondents say their customers are building alternative products with AI coding tools, while only 20% of vendors offer their own customers a fully autonomous agent. SaaStr's more granular cut of the same question found 43% of vendors specifically saying customers are using AI coding tools to rebuild their product or its features — rising to 55% at companies under 100 employees and 48% in vertical software, against 39% in horizontal and infrastructure software.
Asked to name their single biggest AI-related threat, vendors ranked incumbent competitors shipping AI features faster first, AI simply reducing the need for the product second, and customers building it themselves third — ahead of AI-native startups and hyperscalers. Revenue churn points the same direction: the median company lost 10 points of revenue in 2025, seven from customers leaving outright, and churn ran more than twice as high on 12-month contracts (9%) as on contracts of 36 months or longer (4%).
The vendors that are hard to replace hold proprietary data, run deep inside workflows, or own integrations nobody wants to rebuild.
That line, from SaaStr's Jason Lemkin, matches what KeyBanc tells its own investor clients to check for: 64% of vendors in the survey say they build on proprietary or permissioned data, and the firm's diligence guidance to investors centers on how hard that core functionality would be to copy, not on the AI feature list itself.
What a software marketer should do differently
None of this is really an AI-pricing problem in isolation — it's a packaging and positioning problem that AI has made urgent. The data suggests four concrete moves:
- Price AI as a distinct line item, even a modest one. Vendors that do are outgrowing bundlers by ten points, and a visible price signals the feature is worth something rather than a checkbox.
- Audit renewal messaging for short-contract accounts first. Churn concentrates in 12-month deals, which is exactly where a customer deciding to build instead of buy does the least damage to them and the most to you.
- Lead pricing and positioning pages with what can't be rebuilt in a weekend — proprietary data, workflow depth, integrations — rather than with the AI feature itself, which a customer with an AI coding tool can increasingly approximate.
- Treat "why not build this ourselves" as a standard objection in sales and marketing collateral, not an edge case. A plurality of vendors already rank it above AI-native startups as a competitive threat.
If your product's AI features are still bundled into the base price with no separate line item, that is itself a positioning signal worth revisiting — not because every vendor should charge more, but because an unpriced feature reads as unproven.
SearchGoogle tells CMOs to kill the generic 'Top 10' post, and SEOs push back on the metrics, not the content
Sienna Mcpherson · 5 min readSalesforce's CEO declared SEO dead to sell an AI-visibility tool, and his own company's guide disagrees
Sienna Mcpherson · 4 min read
Tools & Analytics