Atlassian Is Putting AI and Automation on a Meter: Here's What's Changing and What To Do About It

If an email from Atlassian landed in your inbox this week talking about “usage-based pricing” and “included allowances”, you’re not alone. Every Atlassian cloud customer is getting one. If you skimmed it and moved on, it’s worth a proper look, so we’re breaking it all down here.

The short version: from 3 December 2026, Atlassian starts billing for usage beyond the allowances included in your plan. The meters are already running. The good news is you have three months of free visibility before a single dollar is charged, and what you do with that window will decide whether December is a non-event or an unpleasant surprise.

What’s Actually Changing

Atlassian is expanding usage-based pricing to five capabilities:

• Rovo credits: the AI features, including Rovo Chat, agents, Confluence AI Slides, Jira Coding Agent and Teamwork Graph queries. In-line summaries, rewrites and Rovo Search remain free and do not consume credits.
• Automation steps: every automation rule you run in Jira, Confluence and JSM
• Assets objects: the things you track in Assets
• AI agent resolutions: autonomous request resolution in Customer Service Management
• Bitbucket meters: build minutes, storage and network, now pooled at the organisation level

Each meter comes with a monthly allowance pooled across your organisation. Allowances reset every month and unused amounts don’t carry over.

Two default settings deserve your attention right now. First, extra usage is switched on by default, billed in arrears on your next invoice. Second, if you switch it off instead, the capability pauses for new requests when you hit 100% of your allowance until the monthly reset. Any work already in flight (a running Chat conversation, an automation mid-flow) will complete; only new requests are held. One default costs you money, the other can pause your automations mid-month. Neither is wrong, but it should be your decision, not something you discover in December.

The Numbers

What’s included depends on your products and plan. Allowances are per user, per month, pooled across the organisation.

Rovo credits

Rovo credits are consumed when a user or agent invokes AI to take an action. Basic AI features (Quick Answer mode in Rovo Chat, basic Rovo Agents) consume a fixed 10 credits per interaction. Premium AI features (Think Deeper mode in Rovo Chat, Confluence slides, Jira Coding Agent, advanced agents) consume a variable number of credits depending on the compute effort and output generated.

Allowances for standalone Jira or Confluence:

• Standard: 25 credits per user per month
• Premium: 70 credits per user per month
• Enterprise: 150 credits per user per month

Allowances for Service Collection and Teamwork Collection (10x standalone):

• Standard: 250 credits per user per month
• Premium: 700 credits per user per month
• Enterprise: 1,500 credits per user per month

Allowances stack across products. If your organisation has 100 Jira Premium users and 100 Confluence Premium users, your pooled allowance is (100 × 70) + (100 × 70) = 14,000 credits per month.

One more thing: enriched Teamwork Graph API calls, CLI calls and MCP server calls also consume Rovo credits. Organisations building integrations or using third-party tools against the Teamwork Graph should factor this into their estimates.

Automation steps

Jira figures shown below. JSM allowances are much higher, Confluence much lower. All automation step allowances pool into a single shared pool across all Atlassian apps.

• Standard: 400 Jira steps per user per month
• Premium: 750 Jira steps per user per month
• Enterprise: 1,000 Jira steps per user per month

When an automation rule includes a Rovo-powered step (e.g. “Use Agent”), that step consumes Rovo credits, not automation steps. The rest of the rule (triggers, conditions, standard actions) still consumes automation steps. A single rule can draw from both meters in one run, each tracked and billed independently.

Assets objects

Assets work differently to the other meters. This is a standing count of the objects you hold, not a monthly consumption figure that resets. Allowances vary significantly by product, not just by plan.

Allowances pool across your organisation and across products, so a customer with both Jira Premium and JSM Premium gets 1,000 + 50,000 = 51,000 objects included. But note: a Jira Premium-only customer gets 1,000, not 50,000. If you’re running a serious CMDB, the product you’re on matters as much as the plan tier.

System schemas (e.g. Services) don’t count toward your allowance. Sandbox objects don’t count either. Deleted objects are excluded, and Assets has no archiving, so everything that exists counts.

Bitbucket

Bitbucket allowances are now pooled at the organisation level across all your workspaces, so a quiet workspace’s capacity is automatically available to a busy one. Unlike Rovo and automation, Bitbucket allowances are flat per organisation, not per user.

When you go over

Prepaid usage packs are available for Rovo credits, automation steps and CSM AI resolutions, with volume discounts on annual commitments. Rovo packs range from 2,000 to 10,000,000 credits per month; automation packs from 40,000 to 200,000,000 steps.

The Bit That Actually Matters

The automation meter doesn’t count what the old limits counted. Previously, you were measured on successful rule runs. From December, every trigger, condition, action, branch and loop iteration counts as a step, and steps are consumed by unsuccessful executions too, including rules that fire and match nothing.

Read that last part again. A global rule listening on every issue update consumes a step on every event, even when it does nothing at all. A rule that counted as one run under the old model might be 5, 10 or 50 steps under the new one. And if you’re on Enterprise, note that unlimited automation is gone. This is the first time the top tier has had a cap.

Sandbox usage counts too. Both Rovo credits and automation steps consumed in sandbox environments count against your production pool. Assets sandbox usage does not. If your teams do heavy testing in sandbox, this is a real gotcha.

Assets has its own quiet catch: because it’s a standing count rather than a monthly meter, going over the allowance isn’t a one-off charge, it recurs every month until you reduce the count. Stale discovery imports and abandoned schemas become a permanent line on your bill. And the allowance cliff between products is dramatic: 1,000 objects on Jira Premium, 50,000 on JSM Premium, 500,000 on Enterprise. A serious CMDB can make the product and plan conversation for you.

Bitbucket is mostly good news. Allowances now pool across all your workspaces at the organisation level, so a quiet workspace’s build minutes cover a busy one automatically. The change to watch is that annual customers get extra usage billing switched on from 3 December, something that previously wasn’t available to them at all.

The other thing that matters is the gap on the Rovo side, and it’s best explained with Atlassian’s own numbers.

We Ran Atlassian’s Calculator So You Don’t Have To

Atlassian has published a usage calculator to help you estimate what you’ll need. We put it through its paces, dozens of scenarios across every meter, and the model underneath is simple: a per-user consumption band multiplied by your headcount.

Here’s what it tells you. Atlassian estimates an organisation in “routine adoption” of Rovo, meaning AI used weekly for drafting, summaries and recurring agent tasks, consumes 500 to 2,000 credits per user per month. Your Premium plan includes 70. By Atlassian’s own numbers, a team genuinely adopting Rovo blows past its included allowance by 7 to 28 times. For a 100-user Jira Premium organisation, that’s roughly US$430 to US$1,930 a month in extra usage, as an indicative illustration.

Automation is the opposite story. That same 100-user organisation has 75,000 steps included, which covers Atlassian’s middle usage profile almost entirely. Even the heaviest profile tops out around US$122 a month in overage.

Assets is in the calculator too, and the numbers are reassuring for most organisations. Even the heaviest “scaling” profile estimates 11 to 50 objects per user, so a 100-user organisation lands at 1,100 to 5,000 objects against a 50,000 JSM Premium allowance (or just 1,000 if you’re on Jira Premium only). Unless you’re running a serious CMDB with automated discovery feeding it, you’re unlikely to trouble the JSM limit. If you are, the Enterprise allowance of 500,000 changes the equation.

One gap worth knowing: Bitbucket isn’t in the calculator at all. For build minutes and storage, your only estimator is your own history, which you’ll find in the Platform usage dashboard.

So the practical read is this: automation overage is cheap, and rarely worth re-engineering to avoid, but it does reward hygiene. Assets rarely bites unless you’re importing at scale, and when it does, it bites monthly. Rovo is where the real budget exposure sits, and where your plan-tier and governance decisions actually matter.

What To Do Before 3 December

The window between now and December is free telemetry. Use it.

1. Open the Platform usage dashboard now. It’s in Atlassian Administration under Insights → Platform usage. Data can take up to a week to populate, so start early and baseline a full month of real consumption.
2. Find your noisiest automation rules. Use the Automation steps view in the Platform usage dashboard, and export the CSV for granular data. Look for global rules with broad triggers, schedules running more often than needed, and rules that constantly fire without matching. Scoping a rule to the projects it actually serves stops the meter running on everything else.
3. Set your limits deliberately. Decide per meter whether you want a hard cap or extra usage with a ceiling. Remember extra usage is on by default.
4. Model your Rovo exposure. Run your numbers through Atlassian’s usage calculator, then measure real credit consumption against your pooled allowance and cost the gap before Atlassian does it for you. Consider whether Teamwork Collection or Service Collection (10x the Rovo credits of standalone apps) is a better fit.
5. Count your Assets objects. Unlike the other meters, this one doesn’t reset monthly. Stale imports, duplicate schemas and objects nobody uses sit on the count permanently, so an Assets clean-up pays for itself every month from December onwards. Check your allowance carefully: it depends on your product, not just your plan tier.
6. Check your sandbox. Rovo credits and automation steps consumed in sandbox count against your production allowance. If your teams test heavily in sandbox, factor that into your estimates.
7. Ask the annual-contract question. Atlassian hasn’t published how mid-term annual customers are billed from 3 December versus at renewal. If you’re on an annual agreement, get that answer in writing.

You’ll also get notifications at 80% and 100% of each allowance, but by the time those arrive the month is largely spent. The dashboard beats the alert.

Why We’re Offering a Usage Readiness Check

We’re running usage readiness reviews ahead of the December change: baselining your consumption and modelling what usage-based pricing means for you. If you’re interested, please get in touch with our team.