Business AI is moving away from predictable per-seat software pricing toward metered consumption — where agent runs, model choices and token usage can directly change the monthly bill.

WHAT’S HAPPENING

Major software companies are changing how businesses pay for AI.

GitHub moved Copilot to usage-based billing on June 1, with AI Credits tied to input, output and cached-token consumption. Customers receive included usage and can pay for additional consumption.

Anthropic’s current Enterprise plan also separates access from consumption. Companies pay for Enterprise seats, while usage across Claude, Claude Code and Cowork is metered separately at API rates. Older Standard and Premium enterprise seats are being transitioned at renewal.

OpenAI now offers eligible Enterprise agreements where usage across ChatGPT, Work and Codex can also be billed by actual token consumption.

WHY IT MATTERS

Traditional business software was easy to budget:

Number of employees × monthly seat price.

AI agents break that model.

One employee may ask a chatbot several questions.

Another may launch an agent that works across files, writes code and runs for an extended period.

Those two users can consume dramatically different amounts of computing power.

That turns AI from a largely fixed software expense into a variable operating cost.

WHO BENEFITS

AI providers can charge more closely according to the computing resources customers actually consume.

Businesses with lighter AI usage may also avoid paying the same amount as organizations running intensive agents continuously.

WHO LOSES

Companies without strong usage controls could face less predictable AI bills.

Finance and IT teams will increasingly need to understand tokens, models, agents, usage limits and budgets instead of simply counting software licenses.

Heavy AI users could also discover that a seemingly inexpensive seat becomes considerably more costly once consumption is added.

WHAT HAPPENS NEXT

More software companies are moving in this direction.

Atlassian announced expanded usage-based pricing for AI and automation, with customers receiving allowances and paying for additional consumption when they exceed them.

The shift is straightforward:

AI software is starting to look less like Microsoft 365.

And more like cloud computing — use more, pay more.