For most of last year, a Claude Enterprise bill was simple arithmetic: seats multiplied by a per-seat price. Agentic work broke that model. When Claude is running skills, editing files, and working through connectors, spend follows activity rather than headcount, and the finance team finds out at quarter-end. On 2 July 2026, Anthropic shipped a set of analytics and cost controls for Claude Enterprise built for exactly this problem. Here is what is in the release, and what Australian finance and IT teams should do with it.
What shipped on 2 July
Usage and cost by group and by user in the admin analytics dashboard, with output such as artifacts created, files edited, and skills and connectors used shown directly next to what it cost.
Model defaults and entitlements, so admins choose which model new conversations start with across chat, Cowork, and Claude Code, and which models are available to specific roles or the whole organisation.
Spend alerts that notify admins at 75% and 90% of an org-level spend limit, with in-app notifications for users at 75% and 95% of their own limits.
New usage and value tabs for Claude Code in the admin console, including active developers, top commands, cost per commit, and an adjustable productivity model.
Analytics chat that answers plain-language questions such as "which teams doubled their Claude usage this month?" and returns charts you can export and share.
An Analytics API that feeds usage and cost data into Datadog Cloud Cost Management, CloudZero, and anything else that can consume an API.
Cost reporting that follows your org chart
The dashboard now breaks usage and cost down by the SCIM groups your IT team already manages. That is a small detail with a large payoff: finance sees the same team structure it budgets against, rather than reconciling a separate mapping spreadsheet every month. Cost also sits next to output. If the operations group spent $4,200 last month, you can see the artifacts, edited files, and connector activity that money produced, in the same view.
Admins can extend that visibility to individual users: cost, product and model breakdowns, and progress against spend limits. Users see their own trends too, which changes the conversation from "why is your team expensive?" to "here is what your usage looks like, and here is what it returned."
Model defaults keep routine work off the most expensive option
Model-level entitlements are the quiet headline of this release. Admins now set which Claude model new conversations start with across chat, Cowork, and Claude Code, and control which models each role can access at all. A coordinator summarising meeting notes does not need the same model as an engineer refactoring a payments service, but until now both defaulted to whatever the product suggested. Setting sensible defaults by role is the single largest cost lever in the release, and it costs nothing to pull.
Alerts arrive before the cutoff does
Spend-threshold alerts notify admins at 75% and 90% of an org-level limit, which leaves time to raise the cap before anyone is blocked mid-task. Users get their own in-app notifications at 75% and 95%, and can request a limit increase from their admin without leaving Claude. For IT teams that have fielded a Friday-afternoon "my access just stopped" ticket, this is the fix: nobody hits a surprise cutoff, and nobody quietly burns through budget either.
Claude Code gets a value column, not just a cost column
Two new tabs in the admin console cover Claude Code specifically. The usage tab shows active developers, session counts, and top commands across the organisation, updated daily. The value tab estimates productivity lift, cost per commit, and annual value. Importantly, every formula is visible and the inputs are adjustable, so the number is defensible in front of a CFO rather than a black box you have to take on faith. For Australian engineering leads writing next financial year's tooling budget, this is the evidence page.
Pull the numbers into the stack you already run
The Analytics API makes usage and cost data available programmatically, filterable by date range, team, product, or model. Finance and IT can pull Claude spend into Datadog Cloud Cost Management or CloudZero and see it beside the rest of their cloud and AI spend. Skills report their own usage and cost, and new endpoints track plugin adoption and artifact creation. If your platform team already treats cloud cost as an engineering metric, Claude now fits the same pipeline.
What this means for Australian finance and IT teams
For an Australian business scaling past a pilot, the numbers get real quickly. A 60-seat Enterprise rollout can pass $70,000 a year in all-in Claude spend once agentic usage ramps, and boards in Sydney and Melbourne are already asking what that spend returns. These Claude Enterprise cost controls are how you keep the answer legible: spend broken down by the teams finance already budgets for, defaults that stop routine work landing on premium models, and alerts that fire before anything is blocked.
The features do not configure themselves, though. Getting value from this release in the first month looks like:
Map your SCIM groups to the cost centres finance actually reports against, before the first dashboard review.
Set model defaults by role, and reserve the top-tier models for the work that justifies them.
Set org and per-user spend limits with the alert thresholds on, so the first budget conversation happens at 75%, not at 100%.
Wire the Analytics API into whatever already holds your cloud spend, so Claude is reviewed like any other line item.
Set up well, this release is the difference between a predictable bill and a quarter-end surprise. Automata AI helps Australian teams stand up exactly this: groups, entitlements, limits, and the reporting pipeline behind them. If you are moving Claude from trial to rollout, book a brainstorm with us and we will map it to your org chart in a session.
