What Comes After SaaS?

For the last twenty years, companies have paid for seats.

A sales seat. A finance seat. A support seat. Then another seat for the person hired to move information between the first three systems.

SaaS was a very good business model. The customer rented access to software, the vendor collected predictable recurring revenue, and the employee did the work.

AI agents change that arrangement. They can read the request, use several systems, make a decision within agreed limits and complete the task. The customer no longer needs another dashboard. The customer needs the invoice reconciled, the support case resolved or the compliance review completed.

The wave after SaaS will be Service-as-Software.

From access to completed work

The commercial model follows the product. A company will pay for a result rather than another login.

Bessemer Venture Partners already separates AI products into copilots, agents and AI-enabled services. Copilots still fit the old per-seat model because they help a person work. Agents complete a workflow, which makes pricing per task or outcome more sensible.

Deloitte expects the same pressure on seat-based pricing. One employee with several agents can do work that once required several software users. Charging for the remaining human seat leaves the vendor with less revenue even though its system may create more value.

So pricing will move towards a mix of platform fees, usage and completed outcomes. A support company may charge for each case resolved. A legal AI company may charge for each document prepared. An accounting agent may charge for each invoice processed without human correction.

That sounds neat. The contracts will be less neat. Somebody has to define what "resolved" means, who carries the cost when an agent makes a mistake, and how several agents divide the credit for one result. Lawyers will be fine.

What survives

Salesforce, SAP and Workday will not vanish. They hold company records, permissions, history and audit trails. That makes them useful infrastructure for agents.

Their screens may become less important. Instead of opening five applications, a person will ask one agent to deal with a customer request. The agent will work across those systems and return with the answer, the action taken and anything that still needs approval.

The valuable layer moves from displaying information to executing work.

Where I would build

Generic agents will be difficult to defend. Models improve quickly and the underlying intelligence gets cheaper.

The stronger opportunity sits inside a specific industry and a complete workflow. Know the rules. Connect the systems. Earn the trust to take action. Measure the result. Keep a human close enough to step in when the situation becomes strange, because it will.

The next generation of software companies will compete with software vendors, consultants and outsourced service providers at the same time. That opens a much larger market, but it also brings variable computing costs, operational responsibility and less forgiving customers.

SaaS digitised the tool. Service-as-Software digitises part of the work. Founders should choose one expensive, repetitive and verifiable job, then build a system that can complete it.