AI agents and seat-based pricing: what SaaS vendors say

01

Fewer seats for the same work

A vendor that charges per user or per employee can lose revenue when customers need fewer of either. Workday says its subscription model is largely based on customers' employee headcount and that changes in headcount, including layoffs, can affect subscription revenue. If AI allows a customer to operate with fewer employees, that is a possible route to lower revenue for a vendor charging on that basis.

A smaller renewal is therefore a reason to ask what changed. It is not enough to infer an AI effect from the seat count. The customer may need fewer users, have fewer employees, or be changing what work it does in the product. Establish the reason before deciding whether the response belongs in product, pricing or the renewal conversation.

02

Customers who build their own

AI can let customers build their own alternatives to software they now buy. Docusign writes that AI has made some foundational capabilities for agreements, such as analysing, summarising and comparing text, "cheaper and easier to replicate". It adds that AI may let companies "quickly and cheaply create agents or other homegrown alternatives" to Docusign.

To win a sale, Docusign, Dynatrace, HubSpot and Workday must show they beat the customer's other options. Those options include "generic LLMs, software created using natural language prompts and generative AI (referred to as vibe coding)". The competitor in your next deal may be a general model, or software someone made with prompts.

One response is to give customers a way to build on the platform. Manhattan Associates describes Agent Foundry, part of its Active Agents add-on, as a way for customers to create agents running on the Manhattan Active Platform. The commercial question is what customers will still need from the platform as they build.

03

Moving from seat-based pricing to credits and usage

Your customers may ask you to move off seats. In February 2026, Forrester advised enterprise buyers to "Renegotiate contracts with enterprise SaaS vendors". It expected contracts based on seats to "shift to consumption (and even outcome-based) pricing as AI agents are deployed". ZoomInfo warns that competitors may offer pricing that follows "the actual usage and realized value obtained by customers", which may appeal more to small and mid-market customers on tight budgets.

Moving off seats changes the exposure. Workday offers AI agents through its subscription-based Flex Credits model and says it has limited experience determining the best pricing for those contracts. Palo Alto Networks warns that customers who rapidly increase consumption may later reduce or reconfigure it to lower costs. The contract needs to show how much spend is committed for the term and how much can fall when usage falls. The words "credits" or "usage" do not answer that question.

04

What customers still pay for

For enterprise software, integrations, security and the consequences of failure can slow a switch and give customers reasons to pay for dependability. If years of software development can be reconstructed in a few days, I don't think we can build the same kinds of software businesses on that work anymore. A company can deliver useful software and still find that the market no longer supports its pricing or cost structure. I explain why in my article on the software industry's economics.

05

Four questions before your next renewals

First establish what your price counts. A customer may be able to reduce licensed users while keeping the same employee headcount. A model based on employee headcount has a different exposure. Check when the contract lets either quantity change, and what that does to the amount the customer pays.

Which renewals came back with fewer seats? Ask those customers why, including what their people now do with AI and what work they still need the product to do. They may be asking what I ask myself about the companies I buy from: "As I become able to do more myself, what will I still need their help with?"

Which parts of the product could the customer replace, and what would it take to use that replacement in practice? Look at the work it must do, the systems it must connect to and the consequences of failure. The sales conversation needs to establish why paying you remains worthwhile against that particular alternative. A list of features a general model cannot reproduce from text is not enough.

If part of the price is credits or usage, what is the minimum commitment and what can vary during the term? Look at the customers whose usage grew fastest, and whether they are now trying to reduce the bill. Palo Alto Networks identifies that optimisation behaviour as a revenue risk.

06

Before you invest in a software company

Ask management the same questions and distinguish scenarios from effects already visible in contracts and renewals. AI due diligence on a software company sets out what else to ask for. That includes the AI cost lines, the reasons behind smaller renewals and lost deals, and questions for customer calls. Before I accepted a claim about years of future software earnings, I would want to understand what keeps the functionality worth paying for once reproducing it gets easier. These questions belong in my AI Executive Advisory and in the examination of competitive position in an independent commercial assessment.