AI Agents and Your Next Software Renewal

Gartner has put a number on something a lot of managers have been sensing: up to $234 billion in enterprise application spending is at risk between now and 2030 from what it calls agentic arbitrage, roughly a fifth of enterprise software spending. The logic is simple. Most business software is priced per seat, assuming a human logs in and does the work. When an agent can do part of that work through a connection instead, the seat count, and the bill, come under pressure. You do not need to believe the forecast to act on it. You just need to stop signing long contracts as though nothing is changing.

Why the pressure is real even if the number is not

Treat $234 billion as a direction rather than a prediction, forecasts about 2030 rarely age well. What is durable is the mechanism underneath it. Seat-based pricing was invented for a world where software value scaled with the number of people using it. Agents break that assumption: work gets done through an integration, at odd hours, without a login. Vendors know this. That is precisely why so many are racing to embed their own agents, so the value stays inside the product rather than leaking to something that connects to it. The pricing model is what is genuinely under threat, not the software.

Your bills will change, not necessarily fall

It would be pleasant to conclude that software is about to get cheaper. Nobody watches a fifth of their revenue evaporate quietly. The likelier outcome is a shift in how you are charged.

What vendors will likely do	What it means for you
Bundle their own agents into your plan	Check whether that is included or an upsell
Shift from per-seat to usage pricing	Your cost becomes variable and needs watching
Push longer commitments now	Locks you in right before the ground moves

That middle row is already happening across the AI features you use, and it deserves the attention we gave it in why AI pricing got harder to predict. The third row is the one to guard against most actively.

Four questions before you re-sign

None of these are hostile, and good vendors answer them without flinching. What does the vendor charge for its own AI features, and is that included in your plan or extra? Does pricing stay per seat, or is it moving toward usage? Can you reduce seat counts mid-term if work genuinely shifts to automation? And how easily does your data come out if you leave? Ask them before signing, get the answers in writing, and you have converted a routine renewal into a decision made with information rather than habit.

The mistake in the other direction

A forecast like this can tempt people into ripping out software and replacing it with homemade automations. Resist that. Established business software does far more than the one visible task an agent could imitate: it holds your data, enforces process, keeps records, handles compliance, and integrates with everything else you run. Swapping that for a bespoke automation usually trades a predictable cost for an unpredictable risk. The right posture is restraint, not replacement, do not over-commit, keep your data portable, and let the market prove what works before you rebuild anything that matters.

Where this leaves you

Do three unremarkable things. Note when your major software contracts renew, because timing is the only real leverage most small businesses have. Ask the four questions before any multi-year commitment. And favour shorter terms or seat-count flexibility where the price difference is reasonable, since paying a little more for the ability to adapt is cheap insurance in a market moving this quickly. You do not have to predict how software pricing shakes out by 2030. You only have to avoid being locked in when it starts to move.

Frequently Asked Questions

What did Gartner actually forecast?

Gartner reported that up to $234 billion in enterprise application spending is at risk between now and 2030 from what it calls agentic arbitrage, roughly 20% of enterprise software spending. The idea is straightforward: a lot of business software is priced per seat, on the assumption that a human logs in and does the work. When an AI agent can do part of that work through a connection instead, the number of seats you need, and therefore the amount you pay, comes under pressure. It is a forecast, not a fact, but the mechanism is easy to understand.

Does this mean my software costs are about to drop?

Not automatically, and probably not soon. Vendors are not going to watch a fifth of their revenue evaporate quietly. Expect them to respond by bundling their own agents into your plan, shifting pricing from seats toward usage or outcomes, or adding value that makes the seat worth keeping. The realistic near-term effect is not cheaper bills but changing bills, with pricing models moving underneath you. What matters for you is not predicting the outcome but avoiding getting locked in just before the ground shifts.

How should this change how I approach renewals?

Mainly by shortening your horizon and asking better questions. A three-year commitment made today locks you into assumptions about how work gets done that may not hold in eighteen months. Before signing, ask what the vendor charges for its own AI features and whether that is included or extra, whether pricing stays per seat or moves to usage, whether you can reduce seat counts mid-term if work shifts to automation, and how easily your data comes out. None of these are hostile questions. Good vendors answer them readily.

Should I cancel software and replace it with agents?

Rarely, and not on the strength of a forecast. Established business software does far more than the visible task an agent could imitate: it holds your data, enforces process, keeps records, handles compliance, and integrates with everything else. Replacing that with a homemade automation usually trades a predictable cost for an unpredictable risk. The sensible posture is not replacement but restraint: do not over-commit, keep your data portable, and let the market show you what actually works before you rebuild anything important.

What should a Canadian business do right now?

Three things, none of them dramatic. First, note when your major software contracts renew, because timing is your leverage. Second, before any multi-year commitment, ask the four questions above and get the answers in writing. Third, favour shorter terms or contracts with flexibility on seat counts where the price difference is reasonable, since paying slightly more for the ability to adapt is cheap insurance in a market changing this fast. You do not need to predict how software pricing evolves. You just need to avoid being locked in when it does.