There is a gap between how AI is being sold and how buyers actually want to buy it. And new research from Infor just made that gap impossible to ignore.

87% of enterprise decision-makers say fixed and predictable AI pricing is important. Not a nice-to-have. Important.

Meanwhile, most AI SaaS companies are still selling on tokens, credits, seats, and consumption tiers — pricing models that guarantee budget uncertainty. That is not a mismatch. That is a sales blocker sitting inside your pricing page.

If you are building or selling an AI product to enterprise buyers right now, this data is a roadmap. Here is what it says.

The Data Nobody Is Talking About

The Infor survey polled 1,000 business decision-makers across the US, UK, Germany, and France — C-suite, VP, director, and head-of-level leaders in industries like manufacturing, logistics, retail, and food and beverage. These are not early adopters. These are the buyers writing the checks.

The headline stat is the 87% figure, but the context behind it tells the real story.

49% of organizations are still in the early stages of AI deployment — running pilots only, paused, or not yet started. That means roughly half of enterprise organizations have not cleared the runway. And these are not small startups. These are businesses with procurement teams, legal review cycles, and board-level scrutiny on every new line of spend.

Three barriers are keeping them stuck. Data security, privacy, and compliance top the list at 36%. That is the number-one blocker. Then comes lack of internal AI talent at 25%, followed by unclear ROI at 23%. These are not technical problems. They are trust problems. Buyers do not trust that the implementation will go smoothly, that their data will be safe, or that the investment will pay off.

But here is the number that reframes everything else.

Enterprises spend 30 to 40% of their total AI budget on integration alone.

Read that again. Before any model runs, before any workflow gets automated, nearly a third to nearly half of the budget is gone — burned on connectors, APIs, data pipelines, legacy system compatibility, and the hidden labor of making AI actually work inside existing infrastructure.

This is the signal most AI builders are missing. The model is not the product. The integration is. And when buyers are already spending 30 to 40% just to get the thing to turn on, unpredictable usage billing on top of that is not a pricing model. It is a deal killer.

Why Usage-Based Pricing Is Becoming a Liability

Token pricing made sense when AI capabilities were experimental. You were paying for API access to a model, not a business outcome. That model is breaking down.

Enterprise procurement teams do not budget in tokens. They budget in dollars, quarters, and annual contracts. When they cannot predict what an AI deployment will cost three months from now, they cannot get the deal approved. The finance team pumps the brakes. The deal goes into "evaluation." And it stays there.

Usage-based pricing creates three specific problems for enterprise buyers.

First, it makes budgeting impossible. If your cost scales with consumption — and consumption scales with adoption — then the more the product works, the more it costs. That is the opposite of what a CFO wants to see in a contract.

Second, it moves financial risk to the buyer. When something unexpected spikes usage — a product launch, a seasonal surge, a process change — the buyer absorbs the cost. That is not a partnership. That is exposure.

Third, it signals that you, the builder, do not fully understand your own cost structure. Fixed pricing requires confidence. It says: I know what this is worth, I know what it costs to deliver, and I am standing behind that. Usage pricing says: let us figure it out as we go.

The market is moving away from this. The signals are already there if you know where to look.

The Pricing Model Shift Happening Right Now

Three major companies made significant pricing moves in a two-week window in April 2026. Each one points in the same direction: away from usage unpredictability and toward buyer confidence.

Model 1: Outcome-based pricing (HubSpot)

On April 14, 2026, HubSpot shifted its Breeze AI agents to outcome-based pricing. The Customer Agent moved from $1 per conversation to $0.50 per resolved conversation. The Prospecting Agent moved from a monthly charge per enrolled contact to $1 per qualified lead.

No resolution. No charge. No qualified lead. No invoice.

This is the most aggressive move in the AI pricing market right now. It completely eliminates the fear of paying for failed attempts. The builder absorbs the cost of underperformance. The buyer pays only when value is delivered.

The builder implication: outcome-based pricing requires a high-confidence product. You need to know your resolution rate, your lead qualification accuracy, and your cost-per-action before you can price this way. But if you have that data, you have a competitive weapon. It removes the buyer's biggest objection — "what if it doesn't work?" — by making the answer obvious.

Model 2: Tiered flat-rate (Microsoft Copilot)

On April 15, 2026, Microsoft restructured Copilot into two tiers — Copilot Chat (Basic) for unlicensed users and M365 Copilot (Premium) at $30 per user per month. For enterprises over 2,000 seats, Microsoft pulled Basic Copilot from Word, Excel, PowerPoint, and OneNote entirely, forcing a decision: upgrade to Premium or lose in-app AI.

The builder implication: tiering is not just a packaging strategy. It is a forcing function. By creating a clear Basic vs. Premium split, Microsoft is telling large enterprises: "you have outgrown the free tier, and you now have to decide what AI is worth to you." Flat-rate tiers give enterprise buyers the predictability they need to get a contract approved. The question is not "how much will this cost us" — it is "which tier are we on."

Model 3: Credit-based with a free trial window (OpenAI Workspace Agents)

OpenAI's Workspace Agents launched as a free research preview, ending May 6, 2026, then shifting to credit-based pricing with no minimum commitments. No per-credit price published yet. Available only on Business, Enterprise, Edu, and Teachers plans.

The builder implication: the free trial window is a competitive forcing function. It accelerates adoption, builds workflow dependency, and then introduces pricing before the buyer can evaluate alternatives. If you are a vertical AI SaaS competing in a space where OpenAI now offers a free trial, you need to move fast on your own pricing clarity. Credit-based models can still be unpredictable — the difference is the cap. Buyers can work with credit budgets if they know what a credit buys and what the ceiling is.

The thread connecting all three: buyers want to know what they are signing up for. Outcome, tier, or credit-cap — all three approaches reduce financial uncertainty more than pure usage billing.

What This Means If You're Building an AI Product

The data is pointing in one direction. Here are three moves that translate it into an enterprise AI pricing strategy.

Move 1: Anchor your pricing to an outcome your buyer already measures.

Not tokens. Not seats. Outcomes.

Every enterprise buyer has KPIs they are already measured on. Support resolution rate. Lead qualification volume. Document processing speed. Time-to-close. If your AI product moves one of those numbers, price against that number.

HubSpot did not invent outcome-based pricing. They had the data to justify it — 65% resolution rate, 39% reduction in resolution time — and they built the pricing around what their buyers already measured. You need to do the same work. What metric does your buyer report to their boss? Price your product as a percentage of that value.

Move 2: Bury the integration cost inside your price, not as a line item.

The Infor data is explicit: enterprises are spending 30 to 40% of their AI budget on integration. They hate it. It creates surprise bills, scope creep, and implementation fatigue that kills deals before they close and poisons renewals.

If you quote integration as a separate line item, you are asking the buyer to absorb a cost they already resent. Bundle it. Build it into your standard package. Or create a fixed-fee onboarding SKU and cap it.

Buyers who are already burned on integration costs will pay a premium for a vendor who says: "implementation is included, and the price will not change." That is not a concession. That is a positioning decision.

Move 3: Give enterprise buyers a predictable ceiling.

Even if your underlying infrastructure cost is variable — and it will be — your price to the customer does not have to be. You can absorb the variability. You can set usage caps. You can offer flat monthly rates with overage options at clearly defined price points.

The goal is simple: a CFO should be able to look at your contract and write a number in a budget spreadsheet without putting a question mark next to it. If they cannot do that, someone else's contract will get approved and yours will not.

Predictable ceiling pricing is not about leaving money on the table. It is about getting the deal done. You can always expand the account once you are in.

The Gartner Warning

Here is the number that should stop every AI SaaS founder cold.

42% of companies plan to deploy AI agents in the next 12 months. The wave is real. The budgets are moving.

But Gartner predicts that over 40% of those agentic AI projects will be canceled by the end of 2027 — due to escalating costs, unclear business value, and inadequate risk controls.

That is not a technology failure rate. That is an expectation failure rate.

Read the three reasons Gartner cites again: escalating costs, unclear business value, inadequate risk controls. Two of those three are pricing problems. "Escalating costs" means the buyer did not know what they were getting into. "Unclear business value" means nobody priced against an outcome the buyer could measure. The third — risk controls — is downstream of the same problem. When buyers cannot predict cost, they cannot justify controls, and the whole project eventually gets cut.

The failure is not happening in the model. It is happening in the contract. Buyers expected predictable value and got unpredictable bills. The CFO kills the renewal. The project gets labeled a failed experiment. The vendor loses the account and a reference.

This is the pricing problem in disguise. And it is going to take out a lot of AI companies between now and 2027.

The Close

The founders who win the next 18 months are not the ones with the best models.

They are the ones who make it easiest for a CFO to say yes.

Fixed pricing is not a concession to an unsophisticated buyer. It is a competitive weapon. When your competitor is still explaining token consumption and your prospect can put your contract number directly into a budget line, you have already won half the deal before the demo ends.

The data is clear. The market is moving. The only question is whether your pricing page is keeping up.

To go deeper, read the two companion guides in this enterprise AI product strategy cluster: AI Product Metrics That Drive Renewals, which explains the numbers that protect retention and expansion, and AI Product Packaging for Enterprise Approval, which shows how to structure AI offers so buyers, finance teams, and procurement can say yes faster.