Your Vendors Added AI and Sent You the Bill

The Conventional Wisdom Is That AI Makes Software Cheaper

It does not. Not for you, anyway, and not this year.

Gartner forecasts enterprise software spending rising 14.7 percent in 2026, past $1.4 trillion, with generative AI named as the primary accelerant. Hold that next to your own business for a second. Did your headcount go up 14.7 percent? Did your revenue? For most small operators the answer is no on both counts, which means the increase is not you buying more software. It is the same software costing more.

The pitch attached to every one of these increases is productivity. The AI features will save your team hours. Maybe they will. But a productivity claim is a hypothesis about your future and a price increase is a fact about your March invoice, and those two things get conflated in every renewal conversation I have watched this year.

Here's What's Actually Happening to Your Bill

Three mechanisms, and they stack.

The first is tier restructuring. A vendor introduces an AI-enabled plan above your current one, then moves features you already use into it. Your plan technically did not go up in price. It went down in contents. Over the past year, companies offering three or four packages raised the price of their cheapest offering by roughly 10 percent, and about two-thirds of vendors landed somewhere in a zero to 20 percent increase band. Ten percent on the entry tier sounds survivable until you notice it happened across eleven vendors simultaneously.

The second is the AI add-on, sold per seat, on top of the subscription. Twenty or thirty dollars a user a month for an assistant. On a fifteen-person team that is $360 a month, or $4,320 a year, for a feature nobody asked for and, in my experience, four people will use.

The third is consumption billing, and it is the one that actually hurts. Credits, tokens, tasks, runs. A meter attached to a product that used to be a flat fee.

The Meter Is the Part to Worry About

A subscription is a known quantity. You budget it, you forget it, and the only surprise comes at renewal. A meter is different in kind. It removes the ceiling.

The failure pattern is consistent enough to describe in advance. The vendor seeds a pilot with generous credits. Usage during the pilot is light because three people are testing it. The pilot succeeds, the feature gets wired into a real workflow, and consumption goes up by an order of magnitude the month it becomes load-bearing. Organizations routinely underestimate production costs against pilot costs by several multiples, and some of the gaps reported in the past year have been closer to tenfold.

Then the invoice arrives, and here is the trap. The feature is now inside your process. Your team depends on it. You cannot negotiate from a position where turning it off breaks the workflow, and the vendor's pricing team knows the shape of that conversation better than you do.

Sixty-one percent of organizations cut projects or initiatives in the past twelve months because of unplanned SaaS cost increases. Read that number again, because it is not describing companies that overspent on software. It is describing companies that spent money they had allocated to something else. The software did not get cancelled. The other thing did.

Per-Seat Pricing Is Dying, and Not in Your Favor

Gartner expects 70 percent of businesses to prefer usage-based over per-seat pricing by this year. That gets reported as a win for buyers, on the logic that you pay for what you use instead of for seats sitting idle.

Look at why vendors are moving. If AI tooling genuinely lets a company do the same work with fewer people, then per-seat revenue falls as customers succeed. A vendor whose revenue drops when its product works has a business model problem. Switching the meter from headcount to consumption fixes it, because automated work consumes more, not less. The vendor gets paid more precisely when you need fewer humans.

That is a rational response to a real threat and I do not blame anyone for making it. But it is not a concession to buyers, and it should not be evaluated as one.

Run Your Own Number Before Renewal Season

Export twelve months of software charges from your bank or card statement. Not the vendor list you think you have. The actual charges, because there will be two or three you forgot about entirely.

Divide the annual total by headcount. That is your software cost per employee, and it is the only figure that makes year-over-year comparison meaningful. If it moved more than your revenue did, you have a problem that adding another tool will not solve.

Then sort the list by increase rather than by size. The $400-a-month tool that went up 8 percent is less interesting than the $90-a-month tool that went up 60 percent, because the second one is where the repackaging happened and the second one is where you have leverage. Small line items get renewed on autopilot precisely because they are small, which is why that is where vendors put the aggressive increases.

Anyone running this exercise for the first time should expect to find a seat count that never came down after someone left. That is the most common finding and the easiest money in the entire process.

The Question to Ask Before You Renew

Ask the vendor for a cap. Not a discount, a cap. A contractual ceiling on consumption charges for the term, with overage requiring your written approval rather than arriving as an invoice.

Vendors resist this and then frequently agree to it, because a cap costs them nothing in the median case and they would rather grant it than lose the renewal. Asking is free. Not asking is how a variable line item becomes an unbounded one.

Then ask what happens to your data if you leave, and get the answer in a format you can test. Export formats and retention windows are where the actual switching cost lives, which is the same dynamic that governs choosing a website builder and the same reason most CRM purchases go wrong. A vendor confident in its product will tell you plainly. A vendor that hedges is telling you something too.

The 14.7 percent is an average across a market. It is not a law of nature, and it is not a bill you have to pay in full.