You’ve finally figured out AI at work—Now comes the bill
Agentic AI has destroyed the SaaS seat-based subscription model
By Alan Jacobson, AI Economics Strategist
This week, stories in The New York Times and Wall Street Journal highlighted something that’s been quietly building inside companies: employees are deploying AI agents that generate massive volumes of tokens—and massive, unexpected costs.
AI is working.
And now the bill is arriving.
The moment the SaaS model broke
For years, enterprise software has been priced per seat.
One user. One license. Predictable cost.
That model worked because usage was naturally bounded by human time and attention.
Agentic AI breaks that completely.
One “seat” is no longer one user doing one task.
One seat can now deploy:
- hundreds of agents
- running continuously
- generating thousands—or millions—of interactions
24/7.
The result:
usage is no longer tied to people
And when usage detaches from people, seat-based pricing loses its foundation.
This isn’t a pricing problem. It’s a cost problem.
Everyone is focused on what AI can do.
Almost no one is focused on what AI costs to run.
That’s the problem.
SaaS worked because the economics were simple:
- build once
- sell many times
- cost doesn’t scale with usage
- margins expand
AI breaks that.
Every interaction has a cost.
Every workflow has a cost.
Every “improvement” increases compute load.
So the more value you deliver, the more cost you incur.
That’s not a product issue.
That’s a business model problem.
Why current pricing models fail
Most companies are trying to force AI into old models:
Seat-based pricing → hides cost until margins collapse
Token / usage pricing → exposes cost but creates anxiety and unpredictability
Neither works.
One breaks the provider.
The other breaks the customer.
And agents make both failures happen faster.
The real requirement: cost predictability
Customers don’t need perfect pricing.
They need:
- predictability
- control
- no surprises
If a CFO can’t forecast it, they won’t approve it.
If a user can accidentally trigger runaway costs, they won’t trust it.
Agents introduce a new failure mode:
cost can be generated autonomously, at scale, without intent
That’s not a billing issue.
That’s a control failure.
Introducing Bounded AI Pricing (BAP)
Bounded AI Pricing is a simple idea:
AI usage must be bounded, visible, and controllable
It has three parts.
1. Fixed price per enterprise with included compute (not tokens)
Each service agreement includes a defined allocation of compute.
Not tokens. Not abstractions.
Actual compute capacity.
This aligns pricing with the real cost driver while keeping it predictable.
Customers know:
- what they are buying
- what is included
- what it can do
2. Fixed overage with early detection and warning
When usage exceeds the allocation:
- overage pricing is predefined and transparent
- customers receive real-time alerts
- soft caps and throttles prevent runaway usage
No surprise invoices.
No “we didn’t know” conversations.
3. Customer-controlled governance
The customer controls usage inside their organization.
- set limits by user, team, or workflow
- restrict high-cost operations
- enforce policies before cost is incurred
This is critical.
Because with agents:
cost can scale without human awareness
Governance closes that gap.
Why this works
For customers:
- predictable spend
- no surprise bills
- control over internal usage
For providers:
- cost recovery aligned to compute
- protection against unbounded usage
- ability to scale without margin collapse
The shift
This is not:
- seat pricing
- or usage pricing
It’s:
bounded consumption with governance
The bottom line
AI doesn’t kill SaaS.
Unbounded cost does.
Who wins AI?
Not the company with the best model.
Not the company with the most features.
The company that makes AI cost predictable and controlable.
– Published on Sunday, March 22, 2026