For twenty years a software bill was simple arithmetic: people multiplied by price. Then the software started doing the work itself, and the question "how many people use this?" stopped describing what you were paying for. If your latest renewal quote has a line called credits, usage or outcomes, this is why.

The short answer: per-seat pricing is not dead, but it is being hollowed out. Most vendors are moving to a hybrid: a predictable seat fee plus a variable charge for what the software's automation actually does. The variable part is where your budget surprises will come from, so learn to read it before you sign.

The three models you will see on one invoice

What each pricing unit measures and what it rewards

Model	You pay for	Predictable?	Where it bites
Per seat	Named users	Yes	Automation lets fewer people do the work, so the vendor's revenue falls; expect pressure to change
Usage or credits	Actions, tasks, tokens, compute time or a per-agent fee	Partly; depends on caps	An enthusiastic team or a runaway workflow can burn a month's allowance in a week
Outcome	A business result, such as a resolved support ticket	No	Defining and auditing the outcome; who decides a ticket was "resolved"?

Deloitte's 2026 technology predictions, published in November 2025, expect hybrid models to become the standard and describe usage-based pricing as attractive because it is quantifiable and therefore auditable. The same report treats outcome pricing as early and hard to measure. Read Deloitte's prediction.

What the market data says

• Deloitte cites Gartner's projection that by 2030 at least 40% of enterprise software-as-a-service spend will shift toward usage-, agent- or outcome-based pricing.
• Gartner's own release of 1 July 2026 estimates that up to $234 billion of enterprise application software spending is exposed to what it calls "agentic arbitrage" by 2030, roughly a fifth of enterprise application SaaS spending, as automated agents complete work across systems without a person opening each application. Gartner's announcement is the primary source for that figure.
• Orb's 2026 study of 80 companies selling automated agents found hybrid pricing at 95%, usage-based pricing at 91.3% (up from 83.3% in 2025) and outcome-based pricing at just 3.8%, down from 4.5%. See Orb's figures. Orb sells billing software, so treat this as an industry survey rather than neutral research.

Put together: the industry talks about outcomes and bills for usage. Plan for usage.

One real price list, decoded

monday.com's work-management plans, as listed on its pricing page on 21 October 2026, show the hybrid shape clearly. Basic is $9 per seat per month billed annually with 1,000 "AI credits" a month; Standard is $12 with 2,000; Pro is $19 with 3,000. The page says credits "reflect the complexity of each task", invites you to "add credits anytime", and separately caps automation and integration actions per month. Prices exclude tax and vary by country.

Read that as three dials, not one price: a seat dial you control at renewal, a credit dial that moves every time someone clicks an automated feature, and an action cap that can stop workflows mid-month. None of this is a criticism of monday.com; it is simply a well-documented example of where the category is going.

A hypothetical ten-seat team, month by month

This is an illustration with invented numbers, not a quote from any vendor.

1. Month one. Ten seats at $12 is $120. The team barely touches the automated features. Credits unused. The bill looks like 2019.
2. Month four. Someone builds a workflow that summarises every incoming request. It is brilliant. It also consumes credits per request, and requests are not something you control. The included allowance runs out on the 19th.
3. Month five. You buy an extra credit pack. The seat fee is unchanged; the effective price per person has risen by a third. Nobody added a user, so nobody noticed in the headcount report.

The lesson is not "avoid automation". It is that the variable line needs an owner, a cap and an alert, exactly like a cloud computing bill. Our 2018 piece on SaaS pricing structures described tiers and seats; the structures have grown a meter.

Eight questions to ask before you sign

1. What exactly consumes a credit, and is the rate per action published or "based on complexity"?
2. What happens at zero: does the feature stop, slow down or bill overage automatically?
3. Do unused credits roll over, and do purchased packs expire?
4. Can an automated workflow spend credits without a person present? Who approves that?
5. Is there a hard monthly cap and an alert threshold you set, not the vendor?
6. If the vendor's model gets more efficient, does your credit price fall, or does the saving stay with the vendor?
7. For outcome pricing: who measures the outcome, with what data, and can you audit it?
8. What is the all-in cost at last year's actual usage, not at the vendor's "typical" team?

Question six is the quiet one. Usage pricing shifts forecasting risk to the buyer, and efficiency gains can land on either side of the invoice depending on how the credit is defined.

Frequently asked questions

Is per-seat pricing going away?

Not soon. Deloitte expects hybrids, not a clean replacement, and Orb's survey shows subscriptions still present at 71.3% of agent companies alongside usage fees. Seats are becoming the floor of the bill rather than the whole of it.

Are credits the same as tokens?

Usually a credit is a vendor-defined unit that bundles tokens, compute and margin into one number. That is convenient, but it also hides the rate. Ask for the conversion.

Should a small business prefer usage pricing?

If your use is light and bursty, usage pricing can be cheaper than paying for idle seats. If a workflow runs on every customer interaction, a flat seat price with generous limits may be safer. Model both on your real numbers.

What does "agentic arbitrage" mean for a buyer?

Gartner's term describes software agents doing tasks across several systems so that people stop opening each one. If your team stops logging into a tool but its automation still runs, you may be paying seats for an empty room and usage for a busy one. Review both.

Research note — 21 October 2026: Based on Deloitte's 2026 TMT Predictions, Gartner's 1 July 2026 press release, Orb's 2026 agent pricing study and monday.com's public pricing page, all opened on 21 October 2026. Vendor prices change; the ten-seat example uses invented figures and is labelled hypothetical. No vendor was interviewed and no product was tested for this article. For how automated agents actually work, start with our guide to AI agents.

AI assistance — 21 October 2026: AI tools assisted with research discovery, drafting and editing. Factual claims and citations were cross-checked against the linked source material on 21 October 2026; prices and product details can change.