On this page
Key Takeaways
- The Claude seat is turning into an access pass. On Enterprise, the seat fee buys access to the product and every token is billed on top at API rates. The meter decides the size of the bill.
- Premium seats are a Team-plan question now. Team still sells premium seats with 5x the usage of a standard seat. On Enterprise, Standard and Premium seats are legacy types that convert at renewal.
- Claude Code shows up on the usage line. The seat includes access to Claude Code, but its tokens are metered. Your heaviest developers will drive the bill, whatever you pay per seat.
- Spend caps come before broad access. Anthropic now ships org, group and user spend limits, alerts and default models. Set them first, then widen the rollout.
- Portfolio-wide deals start with portfolio-wide data. Each portfolio company keeps its own Claude organization and bill. What a fund can standardize is the terms, the controls and the reporting.
The most useful sentence about Claude licensing this year is buried in a help article. Anthropic's Enterprise billing guide says the Enterprise seat covers platform access, including Claude Code and Cowork, "but no usage." Every token your team uses is billed separately at standard API rates.
Read that twice if you own an AI budget. The seat used to be the thing you managed. You counted heads, bought seats, and argued about the renewal number once a year. On a usage-based Enterprise plan, the seat is the floor. The meter is everything above it.
And most companies are not set up to manage a meter. Zylo's 2026 SaaS Management Index found that 78% of IT leaders had unexpected charges tied to consumption-based or AI pricing, and the same report says organizations leave 36% of their SaaS licenses unused. So we have two problems at once: seats nobody uses, and usage nobody planned for.
I think 2027 is the year license management splits into three separate jobs. Deciding who gets a heavier seat. Capping what the meter can spend. And, for private equity firms, setting terms once for a whole portfolio instead of eight times. Each of those needs a different owner and a different habit. Here's how I see each one playing out.
The Seat Is Becoming an Access Pass
For most of software history, the license was the product. You paid per user and the user got the tool. Claude licensing is moving away from that, and the direction is clearest on Enterprise.
Anthropic's pricing page lists Enterprise at $20 per seat per month, billed annually, with a 20-seat minimum, plus usage at API rates. The billing guide adds the part that matters: older seat types (Chat, Chat + Claude Code, Standard and Premium) are no longer sold on new contracts and move to the single Enterprise seat at renewal.
That is a quiet change with a loud effect. A seat count tells you almost nothing about what you will spend. Two companies with 200 seats each can have bills that differ by multiples, depending on who uses Claude, for what, and on which model.
I'm in favor of this, by the way. Paying for usage is a more honest model. You pay for the work Claude does, and light users stop costing as much as heavy ones. But honest pricing only helps you if you watch it. A meter with no one reading it is how the 78% happens.
Read the seat line as a floor
When you budget for Enterprise, treat seats as a fixed access cost and model usage separately, by team and by workflow. If your budget has one line called "Claude licenses," it is already out of date.

Premium Seats Belong to the People Who Prove They Need Them
On the Team plan, seats still carry usage, and the standard versus premium mix is the main lever. The pricing page lists standard seats at $20 per month billed annually and premium seats at $100, with 5x the usage of a standard seat. You can mix the two on one plan.
Here's the pattern I expect to keep seeing. A team rolls out Claude. A few people hit their limits in week two. They ask for premium. Someone approves premium for the whole department because it is easier than deciding person by person. Six months later, most of those premium seats are used like standard ones.
Premium is a 5x price step. It should feel like one.
The opposite mistake is real too. A developer who hits the standard limit every afternoon and stops working is a cost you never see on an invoice. Under-licensing your heaviest users is cheap on paper and expensive in practice.
Earn the premium seat with usage data
Make premium a decision backed by numbers. A person gets a premium seat when they have hit standard limits in most of the last few months, and they keep it while that stays true. Review the list before each renewal. The usage analytics that Team owners already have will show you who qualifies.
Claude Code Lands on the Usage Line
This is the trend I think most finance teams have not priced in yet.
On Enterprise, the billing guide says Claude Code access is included in the seat, and all Claude Code token use is metered at API rates. On Team, once someone hits a limit, they can turn on usage credits and keep working at standard API rates. Either way, the people who use Claude the most spend outside the seat. Developers running Claude Code all day are usually at the top of that list.
That is fine. Often it is the best money you spend on AI. But it means the bill now tracks how your engineering team works, which is something nobody in finance can see from a seat count.
Anthropic seems to know this. In July it shipped new analytics and cost controls for Claude Enterprise, including Claude Code tabs in the admin console for active developers, sessions, and cost per commit. When a vendor builds a cost-per-commit view into its own product, it is telling you where the spend is going.
Give developers a budget before you give them access
Before you turn on Claude Code for a new team, agree on a monthly number for that team and who approves going past it. Developers do better work when they know the limit up front than when they find out from a frozen session.
Spend Caps Come Before Broad Access
If 2025 and 2026 were about getting Claude into people's hands, I think 2027 is about rolling it out with the brakes already fitted.
The tools exist now. The same July release added spend limits by group on top of org and user limits, alerts to admins at 75% and 90% of an org limit, in-app warnings to users at 75% and 95% with a way to request more, and default models by role across chat, Cowork and Claude Code. The billing guide adds that when a limit is hit on a sales-assisted plan, usage stops until the next period or until someone raises it.
So the order of operations matters. Most rollouts I see go access first, controls later, usually after the first surprising invoice. Flip it.

Set group limits first
An org-wide cap is a blunt tool. If one team's agent runs away, a single org cap can stop everyone. Group limits put the ceiling where the risk is. Give your Claude Code and agent-heavy groups their own limit, set the rest of the org lower, and let the alerts tell you when the numbers were wrong.
One caveat, said plainly. Caps protect the budget, and they can also throttle your best people. A cap that nobody reviews turns into a ceiling on adoption. Every limit needs an owner who looks at spend next to what that team produced.
A Simple Way to Sort the Three Trends
When an operating team asks me where to start, I sort the work like this:
| Trend | What it changes | Who should own it | First move |
|---|---|---|---|
| Seat mix (Team) | Who gets premium usage and who stays on standard | Team owner with each department lead | Write a premium rule based on hitting limits, and review it before renewal |
| Usage on the meter (Enterprise) | The bill follows usage, so seat counts stop predicting spend | Finance with the Claude admin | Split the budget into a seat line and a usage line by team |
| Spend caps | How much any org, group or user can spend before someone decides | Claude admin with team leads | Set org and group limits and alerts before adding the next group |
| Portfolio terms | Whether each portco negotiates alone | Fund operating team | Build one rollup of plan, seats, spend and renewal date for every portco |
The last row is where private equity firms have an edge, if they use it.
Portfolio-Wide Contracts Start With Portfolio-Wide Data
Let me be precise about what "portfolio-wide" can mean, because I hear it used loosely.
Anthropic's course on deploying Claude Enterprise says to run one organization per separate contract, and that several organizations can link under a parent to share identity, domain verification, single sign-on and provisioning. Billing and usage tracking stay with each organization. So for most funds, each portfolio company keeps its own Claude organization and its own bill. Seats do not pool across companies.
What can be portfolio-wide is everything around the bill: the plan rules, the controls, the reporting, and the conversation with the vendor. And the market is moving toward portfolio-level AI work. In May, Anthropic, Blackstone, Hellman & Friedman and Goldman Sachs announced a new enterprise AI services firm whose initial customers are the portfolio companies of its investors. When that much capital organizes around portfolios, I expect licensing conversations to follow the same shape.
I don't know how far vendors will go on portfolio pricing. Nobody has shown me a clean public model for it yet. But I do know who gets the better conversation: the fund that can show its total seats, total usage, renewal dates and growth plan across every company on one page. A fund that negotiates eight times, with eight partial views, has none of that.
Negotiate from the rollup
Before any portfolio-level talk with a vendor or a partner, build the rollup. One row per portco: plan, seat mix, active users, monthly usage spend, spend limits, and renewal date. Then standardize the parts you control (seat policy, cap settings, default models, review cadence) so every new portco starts from the same playbook.
What I'm Watching in 2027
I'm confident about the direction. Seats will keep getting cheaper relative to usage, usage will keep growing as agents take on more work, and controls will keep getting finer. The companies that treat licensing as a monthly habit will spend less and get more out of the same tools.
What I'm less sure about is the pricing model on the other side. Will heavy agent users end up on something closer to per-workflow pricing? Will portfolio-level terms become a standard offer or stay a negotiation? I don't have clean answers. I don't think anyone does yet.
If you want a place to start, pick one portfolio company and answer three questions this month. Who on your premium seats actually needs them? What is your usage line by team? And which spend limits would stop a runaway agent before finance finds it? We help teams set this up through Claude Deployment & Setup and Token & Usage Analytics, and when the bill is the problem, Token Optimization. If you want to compare notes on what you are seeing across your portfolio, Book a Discovery Call.
FAQ
On the Claude Team plan, a standard seat costs $20 per month billed annually and a premium seat costs $100, with 5x the usage of a standard seat. You can mix both seat types on one Team plan. On Claude Enterprise, Standard and Premium are legacy seat types that move to a single Enterprise seat at renewal.
No. On usage-based Claude Enterprise plans, the seat fee covers access to Claude, Claude Code and Cowork, and every token is billed separately at standard API rates. Budget for the seat line and the usage line separately.
Claude Code access is included in the Enterprise seat, but all Claude Code token use is metered at API rates. That makes developers who use Claude Code heavily one of the biggest drivers of the usage bill, so give those teams their own spend limits.
Claude Enterprise admins can set spend limits at the organization, group and user level, get alerts at 75% and 90% of an org limit, and set default models by role. Users see their own progress against limits and can request an increase in Claude.
Usually each portfolio company needs its own Claude organization and contract, and billing and usage stay with each organization even when organizations link under a parent. A fund can still standardize plan rules, spend controls and reporting, and bring one portfolio-wide view of seats and usage to vendor conversations.
Treat Claude license management as three jobs: set a premium seat rule on Team, split the Enterprise budget into seats and metered usage, and set spend limits before widening access. Then review seats, usage and limits every month, and start renewal work at least 90 days out.


