Zip Enterprise AI Index: AI's biggest buyers are buying more software too

In March 2026, a thesis nicknamed the “SaaSpocalypse” spread quickly. AI agents would soon do most of what business software does, the argument went, and companies would cancel software seats and move that budget to model providers. A related prediction held that AI would also replace much of the outside expertise companies buy from consultancies and analyst firms.

If either prediction were playing out, companies’ approved purchases should already tell the story. In this second edition of the Zip Enterprise AI Index, we tested both ideas against a year of approved purchase requests.

AI spend did grow fast, rising from 0.4% to 2.7% of approved purchasing in a year. The more telling part is what happened to software and outside-help budgets along the way.

AI spend kept climbing, and almost every AI buyer bought more

AI was 0.4% of approved purchasing in the 12 months ending September 2025. A year later it was 2.7%. Counting only companies active in both years, the share rose from 0.4% to 2.1%, and the typical company’s AI spend grew 202%.

This wasn’t growth at just a handful of companies. 94% of companies that bought AI in 2025 increased their AI spend in 2026. Edition 1 found that Anthropic, Cursor, and OpenAI, the three most widely adopted AI vendors, accounted for most of the AI-vendor spend. With AI budgets growing this quickly and this broadly, the next question is where the money is coming from.

The fastest AI growers spent more on software

To test the SaaSpocalypse thesis, we split the companies that bought AI in both years into two halves by how fast their AI spend grew. If AI were replacing software, the faster half should be the one cutting it. The typical faster AI grower raised its Software/IT spend, excluding AI, 12% year over year. The typical slower grower cut it 3%. Ranking companies instead by the share of their 2025 purchasing that went to AI gives the same answer: the typical AI-heavy company raised its software spend 17%, against 5% for the rest.

Renewals and vendor counts tell a similar story. When faster AI growers renewed a software vendor they already used, the typical renewal came in 6% higher, while slower growers renewed at the same amount as the year before. Faster growers bought again from 67% of the prior year’s software vendors, compared with 52% for slower growers. They also worked with 8% more software vendors than a year earlier, while slower growers’ vendor lists shrank 10%. Ranked by 2025 AI share, the direction holds: +4% for AI-heavy companies, -3% for the rest.

Why software stays in the stack

Instead of consolidating their software stacks, the fastest AI growers are expanding them. Purchase data cannot show why, but one explanation fits the pattern. AI can replicate some features of a software product, yet it still relies on underlying systems to store data, enforce permissions and workflow rules, and carry out actions reliably. That holds even when AI reaches applications through tools like MCP or a CLI, because a system has to sit on the other end. 

There’s also some overlap between the categories. Software vendors are adding AI to their own products, and this analysis counts those features as software and leaves them out of AI spend. AI spend here is therefore likely an undercount, and some of the software growth above may be AI add-ons.

Faster AI growers are still buying outside help

A similar narrative surrounds outside expertise. If AI can draft, analyze and write code, companies should need fewer consultants, contractors and outsourced teams. The headline numbers appear to agree at first: the typical company’s approved AI spend grew 202% year over year, while its external labor (consultants, contractors, staffing and outsourced services) grew 6%. Split the companies by AI growth, though, and the picture gets clearer. Faster AI growers raised external labor 12%, against 2% for slower growers, and 7 in 10 companies growing AI held or grew it.

The split is clearest in business consulting. Among companies that bought business consulting the year before, 6 in 10 cut it. But 8 in 10 slower AI growers cut it, while the typical faster AI grower raised its consulting spend 7%. Outside professional services followed the same pattern (+20% for faster growers, -6% for slower). Purchase data cannot show motive. One likely reason is that many companies feel pressure to adopt AI without knowing where to start, which opens the door for advisory firms, including a new wave of AI-native services firms.

At the companies moving fastest on AI, AI and outside expertise are growing together. For them, AI looks like a reason to buy outside help.

What this signals

Taken together, the data points away from a zero-sum trade between AI and the rest of the budget. The companies buying AI most aggressively this year added software vendors, paid more at renewal and kept spending on outside labor, which suggests AI budgets are being added on top of software and services, at least for now.

For procurement teams, that argues for planning next year’s AI budget as new spend rather than money recovered from SaaS renewals, and for asking at each renewal how much of an increase is an AI add-on that a standalone tool may already cover.

Methodology

This report analyzes approved purchase requests from a fixed cohort of Zip customer accounts, comparing the 12 months ending September 2026 with the 12 months before. Requests are dated by when they were fully approved and include approximately $18B in approved purchase request value. Every figure is a commitment approved in Zip, not a payment, and purchases made outside Zip are not included.

Year-over-year figures use only companies with purchasing in both years, and group comparisons use only companies with at least 10 approved requests in each year. A vendor was classified as an AI vendor when it was on a reviewed list of vendors whose primary product is AI, such as Anthropic, OpenAI, Cursor, Glean and Perplexity, or when a purchase made through a cloud provider/reseller such as AWS, Azure or Google Cloud was tied to a specific AI vendor, such as Claude on Bedrock or Azure OpenAI. AI features inside other products, such as Microsoft Copilot, and general cloud commitments are not counted as AI, so AI spend is likely undercounted.

Each request was capped at $18.8 million, so exceptionally large requests were counted at that amount rather than removed. Overall figures pool purchase-request value across the cohort. Typical-company figures take the median of each company’s own change, giving each company equal weight.

Companies that bought AI in both periods were split into two equal halves at the median AI growth rate of 202%. Faster AI growers are the half whose AI spend grew more than 202%, and slower AI growers are the rest. As a check, companies were also ranked by how much of their 2025 purchasing went to AI. The Index contains no customer identifiers.

Learn more about Zip's AI for procurement platform.

Data sources: This analysis is based exclusively on approved purchase request data from customers who have consented to its use for benchmarking analyses. The findings are reported using aggregated and de-identified data, with measures designed to prevent the identification of any customer, authorized user, individual, company, or confidential information. No customer-specific data is disclosed in this report.