This document is an approval checklist for a SaaS company whose product includes AI. It takes what G2's 2026 Buyer Behavior Report found about where B2B software purchases stall, who approves them now and what buyers need to see, and turns it into twelve checks to run before the next deal reaches the buyer's security and finance teams.
1. The slow part moved from finding software to getting it approved
G2 surveyed 1,038 people who make or influence B2B software purchases in June 2026, across North America, Europe, the Middle East and Africa, and Asia Pacific, and interviewed 55 software marketers. More than 80% of the buyers had taken software recommendations from an AI chatbot in the last two years. Finding options is no longer where time goes.
Evaluation is. 40% of buyers now say it is the stage that takes them longest, up from 36% a year earlier. Once the shortlist exists, the single biggest delay is the IT security review, named by 39%, followed by budget approval at 32%. Among enterprise buyers, 50% name the security review.
The shortlist also hardens early. 80% of buyers who took recommendations from AI chatbots bought from their first shortlist in at least three of their last five purchases. Among buyers who did not, 65% did. Getting on the list matters, and then everything depends on surviving the review.
2. Finance joined the buying committee and started reversing decisions
Finance involvement in software decisions rose from 31% to 46% in a year, while information security fell from 32% to 25%. Security still causes the longest delay after the shortlist, but finance is now involved in more of these decisions and security in fewer. A deal has to pass both. The question at approval moved from whether the product is safe to whether its cost can run over.
Finance is not only asking. 49% of buyers say their CFO reversed a purchase the team had already approved in the last 12 months. Among companies that give their technical teams a separate budget for AI model usage, that figure is 54%, against 29% of those without one. Where AI spending has its own budget line, finance looks harder.
A reversed deal changes how a buyer buys the next one. 70% of all buyers say the pace of AI is pushing them toward shorter contracts. Among those who saw their CFO undo a deal, 40% strongly want contracts under 12 months, and three in four expect a positive return within six months of signing.
3. AI in the product helps the sale and can also stall it
72% of buyers say AI is a must-have or a differentiator when they choose software. In the same survey, concern about colleagues resisting AI grew from 16% to 29% in a year, the largest shift in the whole study. The person inside the buyer's company who wants your product, your champion, now has to convince coworkers who do not want AI in their day.
What helps that person is plain explanation. 87% of buyers say they are more likely to buy from a vendor that explains how its AI uses data or how its models are trained than from a cheaper competitor that does not.
4. A price the buyer can defend to finance
Budgets are growing and being rearranged at the same time. 62% of buyers expect their software spend to rise in the next 12 months and 51% say AI already raised it. 84% consolidated at least three separate tools into one platform in the last year. That is where much of the room for AI costs comes from.
80% of buyers give their technical teams a budget for AI model usage. Who owns that budget varies from company to company. 49% say a current vendor has offered them a price based on usage or outcomes instead of seats. Preference for outcome-based pricing rose from 11% to 23% in a year, and 52% say a variable price improved their view of the vendor. Buyers welcome a price tied to usage or results. What they reject is a bill they cannot predict.
5. Agents compare vendors, people still sign
More than 60% of buyers use or plan to use AI agents, software that carries out tasks on the buyer's behalf, in the buying process, and another 19% would for some cases. The top uses are comparing total cost of ownership and building shortlists, each at 51%, then researching solutions at 49% and evaluating shortlisted vendors at 46%.
Authority stays with people. 47% would let an agent research and recommend while people make every final decision, and 21% would limit it to research. 9% would let an agent buy within approved limits, and 2% without approval first. For a SaaS vendor this means two readers of the same material: an agent comparing cost and features, and a person deciding whether to trust it.
6. Twelve checks before the next deal reaches security and finance
Score each check yes or no against what you can show or send today.
| Area | Check | What passes |
|---|---|---|
| Security | 1. Security documents sent with the proposal | One file that says where customer data lives, who can reach it and which certifications you hold |
| Security | 2. A plain answer to what the AI does with customer data | One page, in plain words, that a security reviewer can check against your contract |
| Security | 3. Evidence that the AI was tested against misuse | Results of tests that try to trick the AI with hidden instructions, or make it reveal data it should not show |
| Finance | 4. A business case your champion can forward to finance | In the buyer's numbers, with visible assumptions, so finance can check it without you |
| Finance | 5. Payback period in writing | With its assumptions; if longer than six months, the reason written next to it |
| Finance | 6. A usage bill the buyer can predict | An expected monthly range and a limit the customer controls; yes if nothing depends on usage |
| Finance | 7. A contract option under 12 months | Offered in writing, with any extra cost for the shorter term stated up front |
| Adoption | 8. Material for coworkers who resist AI | What changes in their day, what stays with them |
| Adoption | 9. Control over how the AI shows up | Settings to limit or switch off AI features by team |
| Adoption | 10. A plain explanation for the people who will use it | How the AI works, whether customer data trains or tunes it, and what it does not see |
| Agents | 11. Facts an agent can compare | Price and integrations on a public page an agent can read without a login or a sales call |
| Agents | 12. A person who owns the deal on your side | Named, reachable, able to answer what an agent cannot |
Each no is a point where a deal the team approved can still stop. 49% of buyers saw it happen in the 12 months before the survey: approved by the team, reversed by the CFO.
Sources
- G2, 2026 Buyer Behavior Report: The Evaluation Maze, July 22, 2026 (survey fielded in June 2026). https://sell.g2.com/2026-buyer-behavior-report
Analysis and conclusions by yaab, based on the published reports listed above.