In January 2026, Zylo published its annual SaaS Management Index, based on a survey of 218 IT leaders. One line stands out: "In the last 12 months, 78% of IT leaders reported unexpected charges tied to consumption-based or AI pricing models." And 61% said unplanned SaaS cost increases forced them to cut projects.
That's your customer. And they've had a busy two years.
- January 2025 — Microsoft raised the US price of Microsoft 365 Personal from $69.99 to $99.99 a year, and Family from $99.99 to $129.99 — the first increase since launch, with Copilot added.
- The same month — Google raised Workspace Business Standard from $12 to $14 per user per month on annual plans, with Gemini bundled in.
- July 1, 2026 — Microsoft's commercial prices went up: Microsoft 365 E3 from $36 to $39, Business Standard from $12.50 to $14. CIO Dive's summary: "The price increases are due to expanded availability of AI, security and management capabilities."
Gartner forecasts worldwide software spending will grow 15.2% in 2026. Its analyst John-David Lovelock put it plainly: "The cost of software is going up." SaaStr's Jason Lemkin estimates that about 60% of that growth is price increases on software companies already own.
The vendors aren't wrong. The buyers aren't either.
From the vendor side, these increases make sense. AI features cost real money to serve: ICONIQ puts the average gross margin of AI products at 45% in 2025, and Microsoft's cloud margin fell to 67% on "growing AI product usage." Somebody has to pay for the compute.
From the buyer side, it looks like paying more for the same tools — and the pushback is getting specific. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027, "due to escalating costs, unclear business value or inadequate risk controls." Contract terms are under the same light: a Stanford CodeX article, citing TermScout's review of AI vendor contracts, found that 92% of AI vendors claim broad data usage rights, only 17% commit to full regulatory compliance, and just 33% provide indemnification for third-party IP claims.
The result is a checklist. Procurement, IT, and finance now walk into AI deals with specific questions. If you set the price, you'd rather answer them on your pricing page than on a renewal call.
The six questions coming to your next deal
1. "If our balance hits zero at 2pm on a Tuesday, what stops working?"
This decides whether a CFO can approve a usage-based product at all. A hard stop means automations freeze mid-day. Figma's answer, for example: when you run out, paid AI features are unavailable "until credits reset or an admin adds additional credits."
The stronger design degrades gracefully: warnings well before zero, a switch to a cheaper model instead of a block, and an admin who can approve more without opening a support ticket.
2. "Do we pay for AI on every seat, or only for the people who use it?"
Notion's AI add-on was $10 per member per month, and it was, in one pricing guide's words, an "all-or-nothing workspace feature — you cannot selectively enable it for specific members while excluding others." As another guide put it: "You have to upgrade and pay for every single member in your workspace." Notion has since moved AI into its Business and Enterprise plans.
Buyers check this line first now. If your AI fee lands on people who never touch AI, expect it to be the first thing negotiated away.
3. "What does each action cost — in writing?"
Credits only work if the buyer can translate them into work. On Figma's forum this spring, one user reported dropping from 3,000 credits to 1,220 in about ten prompts — one prompt cost 136. When one action can cost many times more than another, a balance tells the buyer what they've spent, not what they're about to spend. Publish a rate card per action.
4. "Do unused credits roll over?"
If the answer is no, the buyer reads every unused credit as your profit. Figma's included credits "reset monthly and don't roll over" — reasonable for an allowance that comes with a seat. For a large prepaid commitment, rollover is the concession buyers will ask for, and billing platforms like Metronome already support rolling a percentage of the balance into the next contract. Decide your position before the buyer asks.
5. "If you change the model underneath, what protects our rate?"
When Anthropic launched Claude Opus 4.7, it noted that its new tokenizer can map the same input to roughly 1.0–1.35x as many tokens, depending on the content. If your price tracks tokens, a model upgrade can raise your customer's bill with no change in what they do. Price per task, or commit to rate protection.
6. "What do you do with our data?"
When 92% of the AI vendors in that analysis claim broad data rights, this question is now standard. A clear default — no training on customer data without opt-in — is a sales asset, not a legal footnote.
Design so the checklist is boring
None of these questions is hostile. Each one is a buyer trying to forecast a bill and limit a risk. A pricing model that answers all six on the pricing page shortens every sales cycle. One that can't gets found out at renewal — with the buyer's spreadsheet open.
If your pricing can't survive the buyer's checklist, it won't survive the renewal.
Vendors arrive at renewal with usage reports. Smart buyers arrive with a value file: what the tool actually delivered, measured in their own KPIs. Build that file with them before they build it against you.
Which of these six questions would your pricing page answer today — without a sales call?
Free companion
Find Your Value Metric — Pricing Worksheet
A fill-in-the-blanks pricing worksheet to pick the metric that scales with customer value — the four-question test, a scorecard for candidate metrics, and a one-page template you can run through with your team before the next buyer does.
Get the worksheet →This is one piece of a longer framework I teach in Chapter 7 of Product Strategy in the AI Era — including how to match pricing complexity to deal size, and how to price for the risks large organizations pay to remove: security, compliance, and governance.
Sources
- Zylo, 2026 SaaS Management Index (January 29, 2026): zylo.com
- GeekWire, Microsoft 365 consumer price increase (January 16, 2025): geekwire.com
- 9to5Google, Google Workspace price increase (January 15, 2025): 9to5google.com
- Microsoft, 2026 M365 packaging and pricing updates: microsoft.com
- CIO Dive, Microsoft 365 AI tools, higher price (December 5, 2025): ciodive.com
- Gartner, IT spending forecast for 2026 (October 22, 2025): gartner.com
- SaaStr, Gartner enterprise software spend (November 27, 2025): saastr.com
- Gartner, agentic AI project cancellations (June 25, 2025): gartner.com
- Stanford Law CodeX, "Navigating AI vendor contracts" (March 21, 2025): law.stanford.edu
- Figma Forum, AI credits enforcement (March 11, 2026): forum.figma.com
- Monetizely, "How much does Notion AI cost?": getmonetizely.com
- Metronome docs, applying credits and commits: metronome.com
- Anthropic, "Introducing Claude Opus 4.7" (April 16, 2026): anthropic.com