What Happens to Software Pricing When AI Replaces the User?

 

Software companies have charged businesses by the number of people using their products for decades. But what happens when the “user” doing the work is an AI agent?

Imagine a company with 100 employees.

It uses business software that costs $30 per employee each month.

The math is simple.

100 people × $30 = $3,000 a month.

If the company grows to 200 employees:

200 people × $30 = $6,000 a month.

This is commonly called per-user or per-seat pricing.

The basic idea is simple:

One person = one software seat.

It has worked extremely well for business software.

But AI agents are starting to create a strange problem.


What If the Software User Isn't Human?

Think about a customer service worker.

A customer asks where an order is.

The worker reads the question.

They open the company's software.

They find the order.

They check the shipping status.

Then they answer the customer.

The company pays for software accounts because humans need those tools to do their jobs.

Now imagine an AI agent doing much of the same work.

The AI reads the question.

It finds the order.

It checks the shipping information.

It responds.

Now we have an unusual question:

Should an AI agent simply count as another software user?

That gets even stranger if one AI agent can handle thousands of tasks.

Calling it "one user" may no longer tell us very much about how much work the software is actually doing.


Salesforce Is Already Charging Differently

This isn't just a future problem.

Salesforce already offers several ways to pay for its Agentforce AI platform.

One model charges $2 per conversation.

Instead of simply counting how many employees have accounts, the customer can pay based on conversations handled by the AI.

Salesforce also offers Flex Credits.

The listed price is:

$500 for 100,000 credits.

When an AI agent performs certain actions, it consumes credits.

The important part isn't the credit itself.

It's what the credit represents.

The unit being sold is moving away from:

How many people use our software?

and toward:

How much work did our AI perform?


What If You Pay Only When the AI Succeeds?

There's another model that goes even further.

It's called outcome-based pricing.

The idea is simple:

Pay for a result.

Imagine an AI customer service agent.

Charging for access means:

"You can use our software, so you pay us."

Charging for an outcome can mean something closer to:

"Our AI solved the customer's problem, so you pay us."

Salesforce has introduced resolution-based pricing for some AI customer-service use cases.

That's a major change in what software companies are actually selling.

For decades, software companies mostly sold access to tools.

AI could allow them to sell completed work.


Microsoft Shows Both Worlds at Once

Microsoft makes the transition particularly easy to see.

Microsoft 365 Copilot still uses a familiar human-centered pricing model for many business customers.

A person gets access to Copilot.

The company pays for that person.

Microsoft lists Microsoft 365 Copilot Business at $30 per user per month under its monthly commitment option.

That's classic per-user software pricing.

But Microsoft's Copilot Studio, which businesses can use to build and operate AI agents, also supports consumption-based pricing through Copilot Credits.

Microsoft's pay-as-you-go rate is listed at $0.01 per Copilot Credit.

So inside the same company, we can already see two different ideas:

Human uses AI → charge for the user.

AI agent performs work → charge for usage.

The two models can exist together.


Why Would Software Companies Want This?

Because AI could break an important relationship that helped make SaaS such a powerful business model.

Traditionally:

A company grows

It hires more employees

It needs more software seats

The software company's revenue grows

Now imagine AI changes the equation.

Consider a fictional company.

Before AI:

100 workers → 10,000 tasks per day

After AI:

50 workers + AI agents → 20,000 tasks per day

The exact numbers are hypothetical, but the problem is real.

The number of employees has fallen.

The amount of work being done has doubled.


If the software company charges only for human seats, its pricing may no longer reflect the amount of value its software is creating.

That's why software companies have a reason to find something else to measure.


Four Ways Software Can Charge

The emerging models can be understood without complicated SaaS terminology.

Per user

Pay for each person using the software.

Per usage

Pay based on how much of the system is used.

Per action

Pay based on tasks performed by an AI agent.

Per outcome

Pay when the AI produces a defined result.

AI could make the last three increasingly important.

But that doesn't mean per-user pricing is about to disappear.


The Software Seat Isn't Dead

Microsoft still sells software by the user.

Salesforce still offers user-based licenses.

And other companies are experimenting with combinations of subscriptions, AI usage and bundled features.

So saying:

"AI will kill per-seat pricing."

would be too simple.

A more realistic possibility is that businesses will pay in several ways at once.

A base subscription.

Plus AI usage.

Plus premium capabilities.

And perhaps, in some cases, a fee for successful outcomes.

AI isn't necessarily replacing the old pricing model.

It is adding new things that software companies can charge for.


There's a Problem for Customers Too

Per-user pricing has one major advantage.

It's easy to understand.

100 employees × $30.

The company knows roughly what the bill will be.

AI usage can be less predictable.

An agent might perform 100,000 actions this month.

Next month it might perform one million.

If the business suddenly grows, it might perform far more.

That's good if the AI is producing valuable work.

But the software bill may grow with it.

This creates another important question for businesses.

Not just:

How much work can our AI do?

But:

How much does it cost every time we ask AI to do that work?

AI could make software more productive while also making software spending more variable.


What Is Software Really Selling?

For a long time, business software largely sold access.

A login.

A CRM seat.

A productivity tool.

A workspace.

AI agents begin to change that because software is no longer only a tool that helps a human do the work.

The software can increasingly do some of the work itself.



And that could change the most important question behind software pricing.

The old question was:

How many people use the software?

The next question may be:

How much work did the software do?

And eventually:

What result did it produce?

Per-seat software isn't going away tomorrow.

But AI is already changing what software companies can measure — and what they can charge for.

The future of software pricing may be less about how many people log in,

and more about how much work gets done.


Sources

Salesforce — Agentforce Pricing

https://www.salesforce.com/agentforce/pricing/


Microsoft — Copilot Studio Pricing

https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio


ServiceNow — AI-Native Platform

https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-moves-beyond-the-sidecar-AI-era-giving-customers-a-complete-AI-native-experience-across-all-products-and-packages/default.aspx