AI Didn't Eat Your Software Budget. It Opened a New One.

AI Didn't Eat Your Software Budget. It Opened a New One.
Nik Mercado

The story for most of this year was that AI spend would come out of software spend. Agents would replace SaaS seats, the money would move, and the CFO's job was to manage a reallocation. The data says otherwise. Menlo Ventures' survey of enterprise AI buyers found that only about a fifth of this year's AI spend came out of existing software budgets. More than half came from budget that didn't exist a year ago. AI didn't eat the software line. It opened a line of its own, and most companies are still trying to manage it with the tools and reflexes built for the old one. We call that new line the net-new line, and the question it raises is what the company is getting for it, because nobody set a return target for a line they didn't know they'd have.

What is the net-new line?

It's the share of AI spend that came from new budget rather than reallocated budget, and in most enterprises it's the majority.

The reallocation story made intuitive sense. If an agent does work that software used to do, the software budget shrinks and the AI budget grows by the same amount, and total spend holds roughly flat. That's not what happened. Of the AI spend Menlo could trace, about twenty-two percent came from software budgets, most of that from incumbents converting their own seat licenses into AI tiers. Roughly a quarter came from labor and other lines. The rest, around fifty-five percent, was net-new money, approved on top of everything else. The average enterprise ended the year spending about two and a half times what it planned. That's a new category of spend that arrived faster than the planning cycle could absorb it.

Why didn't AI cannibalize software?

Because the work agents do mostly wasn't being done by software before. It was being done by people, or not at all.

The cannibalization thesis assumed AI would substitute for existing tools. In practice the biggest spend categories were coding agents and workflow automation, and the work they absorbed came out of hours, backlogs, and tasks nobody had gotten to, not out of SaaS licenses. A coding agent doesn't replace your issue tracker. It replaces the three weeks the feature would have waited. A contract-processing pipeline doesn't replace your contract system. It replaces the paralegal hours and the deals that stalled in review. That's why the software line barely moved while the AI line exploded, and it's why the labor line is the one that should be watched. We wrote about the efficiency alibi, where a headcount cut gets credited to capability that doesn't exist yet. The net-new line is the honest version of the same question. If the money is new and the work came out of labor, what did the labor get redeployed to?

Isn't net-new spend just a budget overrun?

Only if nobody decided what it was for. Unplanned spend with a return is an investment. Unplanned spend without one is a leak.

This is where the net-new line becomes a leadership problem rather than an accounting footnote. Spend that came out of a software budget inherits the software budget's discipline. Someone owned that line, justified it, and gets asked about it. Spend that arrived as a new line in the middle of the year inherited nothing. It got approved in pieces, team by team, as agents proved useful, and by year end it was the largest new cost category in the company with no return target attached. We've argued that AI should be managed as capital, with a return on intelligence tracked per deployment. The net-new line is where that argument stops being theoretical. The capital is already deployed. The question is whether anyone is measuring what came back.

Does this mean the SaaS disruption story was wrong?

It was early, and it was aimed at the wrong mechanism.

The software line grew more slowly this year than last, so something is happening, but the mechanism is two slower forces rather than agents replacing tools seat for seat. Incumbents are converting their own licenses into AI tiers, which keeps the dollars inside the software line while changing what they buy. And the work that software supported is starting to move to agents that use the software directly rather than through a person clicking, which we've described as headless agents. That second force is the real threat to the software line, and it hasn't shown up in the budget data yet because it's still early. So the honest read is that the reallocation will come, but it will come from agents consuming software on your behalf, not from AI replacing it, and the timing is years rather than quarters.

What should the CFO do with a line that appeared on its own?

Give it what every other major line has: an owner, a return target, and a split between production and discovery.

The net-new line is too large to leave as a reaction. Three moves bring it under management without choking it. First, name an owner. Not the CIO by default, and not finance by default, but whoever can answer what the spend produced, which is usually a business leader whose workflows the agents run. Second, set the target in outcomes. We've made the case for cost per outcome as the unit that survives the shift to agents, and the net-new line is where it belongs, because a line with no outcome attached will get cut the first quarter the market turns. Third, split it. Part of the line is production spend on workflows you trust, and that part gets measured tightly. Part of it is discovery budget, the experiments that found the workflows in the first place, and that part gets protected rather than measured on the same terms. A single undifferentiated line is what gets cut.

What should you do about the net-new line?

The question you're really asking is whether the money your company spent on AI this year was an investment or a leak. The uncomfortable answer is that for most companies, nobody knows, because nobody was asked to.

Pull your AI spend for the year and sort it three ways. How much came from existing software budget, how much from labor and other lines, and how much was new. Then take the new portion and ask what it produced, workflow by workflow. Where you can answer, you have an investment, and it deserves a target and a defender for next year's planning. Where you can't, you have a line that will get cut by whoever is looking for savings, whether or not the work behind it was valuable. The companies that come through the next budget cycle intact will be the ones that gave the net-new line an owner and a return before someone else gave it a haircut.

If you want your AI spend traced to outcomes before the next planning cycle asks, start with an AI Blueprint or reach us at contact@theyor.com

Next
Next

Your AI Plan Didn't Get More Expensive. It Got Smaller.