Australia: The release where AI delivers for developers
With the Australia release, developers are armed with AI tools that help you Sense, Decide, Act, and Secure
The Australia release ships the features Platform Developers have been asking for:
- Build Agent in Studio with global scope support
- Flow Execution Analysis that traces through nested subflows in production
- ATF code coverage on your custom scripts
- The ability to build from Cursor, Claude Code, or Windsurf and deploy directly to the platform.
The features in this release map to four connected stages of how AI should work in the enterprise: Sense, Decide, Act, and Secure. With the Australia release, developers get new tools across this spectrum so you can build better, iterate faster, and deploy with confidence.
Sense: See What’s Actually Happening
Before you build anything, you need signal. These features are about giving developers better visibility into what’s breaking, what’s working, and why.
Flow Debugging Got a Serious Upgrade
Two Australia features directly address one of the most time-consuming parts of working with Flow Designer: figuring out what broke and why.
Flow Execution Analysis
When a flow fails, gets stuck, or cancels unexpectedly, you now have an “Analyze” button that sends the full execution context to an LLM and comes back with a structured breakdown: root cause, where the problem occurred, and recommended next steps.
A few things worth knowing about how it works:
- Works in production, not just dev instances
- Can pull from audit and log data even if you’ve turned reporting off, which is a common production configuration
- Traces through nested subflows a few levels deep, so if a parent flow calls a subflow that calls another subflow, the analysis doesn’t just tell you “the subflow failed”
- You can rerun the analysis at any point which is useful when a flow is stuck or paused rather than outright failed and you want to check its status again after something changes
Flow History Version Comparison
Pick any two versions of a flow from flow history and get a visual diff: blue for values that changed within the same action, green for new content, red for removed content.
This was one of the most-requested features from the Flow team. The team heard the feedback, and delivered. Now you can see exactly what changed from the last published version in seconds instead of clicking through every action.
Learn more about improvements to Flows in the Australia release
ATF: Code Coverage and a Test Troubleshooting Agent
Two additions worth knowing about if automated testing is part of your development process.
ATF Code Coverage is now available. Run your test suite and get a view of which of your custom scripts (business rules, client scripts, UI actions) were actually exercised. By default, it only captures scripts you’ve modified or created (identified by update version), which means you get signal on your changes without being flooded with data from ServiceNow’s out-of-box scripts. There’s a system property to capture everything if you want comprehensive coverage across the full stack.
ATF Troubleshooting Agent is an AI skill (Now Assist for Creator) that activates on failed test results. From any failed ATF test, click the analyze button and the agent investigates the failure, surfaces root cause, and suggests fixes. For teams running automated test suites, this cuts down the time spent decoding what went wrong in a failing run.
Learn more about ATF enhancements in the Australia release
Decide: Put Intelligence Inside the Workflow
AI is most useful when it shows up at the moment a decision gets made. These features are about embedding intelligence where work happens.
The Knowledge Layer: Load-Bearing Infrastructure for Every Agent You Build
Any time you build AI-powered features on the platform, one thing determines whether your agents make good decisions or confidently wrong ones: the quality of what’s in Knowledge Center.
The connection is direct. Intelligent Approvals (one of the standout Australia features in the agentic workflow area) handles routine approval requests automatically, drawing from your policy documentation to make decisions. The “why did that decision happen” question has a legible answer: the agent read the policy, the policy was accurate, and the decision followed from it. That chain breaks if the knowledge is stale, duplicated, or missing.
Clean, well-maintained knowledge isn’t documentation hygiene. It’s the input that determines whether your agents are making good decisions or confident wrong ones. For any agent you build, Knowledge Center is load-bearing infrastructure.
AI Agent as a Playbook Step (Already Live)
You can now drop an AI agent directly into a playbook as a step: helper agents (supervised, prompts for human confirmation at decision points) or worker agents (autonomous). The agent’s activity surfaces inline in the activity card, with no separate panel and no Now Assist sidebar. Dynamic UIs that the agent surfaces for human confirmation are rendered automatically, no extra playbook configuration required.
Outputs from the agent step flow into subsequent playbook steps, so you can chain agent work with deterministic process logic.
If you’re on Zurich, this one is already available if you have Now Assist for Platform. It landed in the last Zurich store release, so no need to wait for Australia.
Learn more about bringing AI agents into your workflows in the Australia release
Act: Turn Intelligence Into Execution
This is the heaviest section for platform developers. The Australia release is packed with features that enable you to build and deploy with confidence.
Build on ServiceNow With the Tools You Already Use
developers can now build ServiceNow applications using Cursor, Claude Code (CLI), Windsurf (IDE), and Codex in their local environment and deploy directly to the platform. Starting with @Servicenow/sdk v4.6, the now-sdk exposes AI-consumable skills designed explicitly for helping external coding agents build, edit, and deploy Fluent applications.
This is the answer to the “why build here” question that full-stack developers have been asking. Practically, this means you can use VS Code with the Claude plugin, jump straight into Cursor or Windsurf, or use the OpenAI Codex API from your terminal environment. Whatever dev environment makes you most productive, you can write your Fluent application there and deploy directly to the ServiceNow platform when you’re ready.
For developers who’ve been evaluating whether ServiceNow fits their existing workflow, this removes one of the most common blockers. You don’t have to give up your local development setup to build on the platform.
Learn more about Build Anywhere with your preferred tools in the Australia release
Get a step-by-step guide to building ServiceNow apps via Claude Code and the ServiceNow SDK
Build Agent Arrives in ServiceNow Studio
This is the headline for platform developer tooling in Australia. Build Agent (previously only available in the ServiceNow IDE) is now available in ServiceNow Studio.
That matters more than it might sound. A lot of developers and admins who build on ServiceNow never left Studio. The IDE was a great option, but it required a different workflow, and not every team adopted it. But the goal was always to meet developers where they are — and ServiceNow Studio is where most of us already work. Going forward, ServiceNow Studio and IDE stay in sync and new capabilities land in both.
What can Build Agent do today?
- Tables, columns, business rules, script includes, catalog items, flows, security configurations, around 40+ types of metadata
- Now works at global scope, not just scoped apps which is a significant expansion for developers maintaining legacy global customizations
- The default model Build Agent uses is Claude Opus 4.6, with the option to select Sonnet 4.5 for when you want faster iteration on lighter tasks
- Every action gets captured as a checkpoint backed by an update set so you can see your full history of what was built, and roll back if needed
On PDIs: Build Agent in Studio is now available on PDIs. The existing prompt limit has been increased from 10 free calls to 25 free calls per month. For the full Australia feature set without hitting those constraints, a regular instance is the better path. The trial version in Studio is the fastest way to get hands-on without an existing Now Assist for Creator entitlement.
Learn more about Build Agent in ServiceNow Studio in the Australia release
Build Agent Can Now Build (and Edit) Flows
This deserves its own callout because it’s been one of the most-requested Build Agent capabilities: flow generation.
You can now describe an automation in natural language and Build Agent will produce a fully wired flow with trigger conditions set, actions populated, data pills connected. You can also ask it to edit an existing flow with a follow-up prompt, because it understands context from previous steps and applies changes surgically.
What makes this more than “it writes a flow” is that the Flow team contributed out-of-box knowledge sources that tell Build Agent how to build flows correctly, like which trigger type to use for different scenarios, how to structure conditions, which actions to reach for. You’re not prompt-engineering your way through this from zero. Ask for a flow, and Build Agent already knows what a well-built flow looks like.
There’s also a diagram preview in Build Agent before you commit to opening the canvas, so you can sanity-check the logic structure without diving all the way in.
Learn more about using Build Agent to edit flows in the Australia release
Playbooks: More Composable, AI-Ready
Nested Playbooks
If you’ve been building playbooks for a while, you know the pain: every time you need an approval escalation process, or a standard compliance checklist, or a review cycle, you rebuild those same steps inside each new playbook. Maintaining them across twelve playbooks when requirements change is painful.
Nested Playbooks solve this. Build your reusable process once as a standalone playbook, then nest it wherever it’s needed. From the end user’s perspective, the steps look identical to any other step in the parent playbook. The maintenance win is entirely on the development side.
The prerequisite is Standalone Playbooks (also new in Australia): non-record-driven playbooks that can kick off without a parent record anchoring them. They’re the enabler for nesting, but they also open up questionnaire-style workflows and guided decision trees independently.
The CSM guided decision tree capability is also migrating to the platform as playbooks in Australia. If your team has been building decision trees, you can now build them as platform playbooks, with nested children and all the composability that comes with it.
Learn more about Nested Playbooks in the Australia release
UI Interactions: A Better Way to Customize Workspaces
Anyone who has built workspace customizations knows the friction: you want to add a button that opens a custom modal, you end up doing UXF client action configuration, then go into UI Builder to wire up add-on event mapping, then back to declarative actions, and hope it all connects.
UI Interactions are a new framework in Australia that changes this. A UI Interaction is a self-contained unit (UI, logic, and script) that can be triggered from anywhere without modifying the page it lives on. Whether from a declarative action, component event, or another UI Builder page: same UI Interaction, same behavior, zero page modifications.
The new workflow:
- Open UI Builder (there’s a new UI Interactions tab alongside presets and controllers)
- Create a UI Interaction and pick a type: form, list, or generic
- Define your steps from the toolbox (modals, modeless dialogs, user input modals, navigation, server scripts, if/else conditions)
- Set up inputs, wire up events between steps
- Reference it from a declarative action or event handler
One place to define the interaction and trigger it from multiple places. Update it once and the change propagates everywhere it’s used.
Form and list types give you access to the full client scripting environment out of the box. Server script steps have role enforcement applied automatically. And the team shipped demo data: working example UI Interactions you can inspect, duplicate, and learn from.
Learn more about UI Interactions in the Australia release
All Menu Configuration: Bringing Users Into the Modern Era
For teams that still have users working in classic navigation (and haven’t made the full jump to workspaces), Australia adds a way to give them a better experience without a full workspace buildout.
The All Menu Configuration experience lets you create and edit application navigation using a drag-and-drop editor, AI-assisted module creation, and a mix of next-experience and classic modules side by side. Inline analytics dashboard configuration, form layout editing without UI Builder, contextual sidebar presets, all designed for admins who need to quickly improve what their users see without becoming full UI Builder experts first.
Next-experience module pages created this way also support UI Interactions out of the box, without needing a full workspace implementation.
Learn more about UI updates in the Australia release
Release Ops Gets Runbooks
Release Ops launched in Zurich as the automation foundation for deployments: pipeline orchestration that moves update sets, app repo artifacts, and plugins across your instance chain, with ATF and instance scan checks built in. Australia adds the part that was always the gap: the manual steps that can’t be automated.
Every release process has them. Training the user group. Uploading docs to SharePoint. Running a fix script after deployment. Getting UAT sign-off. These have lived in spreadsheets, email threads, and people’s memories, all completely invisible to the Release Ops pipeline.
Runbook Tasks bring all of that into the system. You define manual tasks tied to specific pipeline stages (before move to test, after test deployment, before ready-for-deploy), and the pipeline pause points won’t advance until those tasks are marked complete. Release managers get a single view of both automated pipeline status and manual task status. No more cross-referencing your deployment spreadsheet against your ATF results.
One infrastructure note: Release Ops still doesn’t support PDIs, GCC instances, or on-prem. It depends on MIF/MIM (ServiceNow’s inter-instance communication tech) and those environments aren’t compatible yet. Cloud instances are fully supported.
Learn more about Release Ops in the Australia release
Secure: Run AI as a Governed System
Shipping agents is one thing. Having confidence that they’re doing the right thing in production, and that you can verify it, is another. These features are about keeping developers and platform engineers in control as AI adoption scales.
Governance Starts With What Your Agents Know
The governance story for AI agents runs through knowledge quality. When Intelligent Approvals makes a decision, the audit trail starts with the policy document it read. When an agent gives a user guidance, the reliability of that guidance traces back to whether the knowledge it drew from was accurate and current. Build Agent’s flow generation is grounded in curated out-of-box knowledge sources from the Flow team for the same reason. Getting your knowledge infrastructure right isn’t a post-launch cleanup task, it’s part of the build.
Build Agent’s Checkpoint System: Every Action Is Reversible
Every action Build Agent takes gets captured as a checkpoint backed by an update set. You can see the full history of what was built and roll back if needed. That’s a real confidence boost for anyone considering using it beyond quick prototypes, and it’s part of why the trial is worth actually exploring rather than just testing once.
Release Ops Runbooks Close the Manual Audit Gap
The Runbook Tasks feature in Release Ops isn’t just a convenience feature for release managers, it’s closing an audit gap. When manual tasks (training, UAT sign-off, fix scripts) lived in spreadsheets and email, they were invisible to any governance or compliance review. Bringing them into the release pipeline means every step, automated or manual, is tracked in one place with a clear completion record.
What to Do Right Now
The clearest first step: upgrade your PDI to Australia (or request a fresh one) at developer.servicenow.com. Most of what’s in this post is available to try immediately on an Australia instance.
If you’re figuring out where to spend your first few hours with Australia, Build Agent in ServiceNow Studio and the Flow version comparison are probably the highest-leverage places to start. Build Agent in ServiceNow Studio is worth exploring even if you’ve been using the IDE: the parity is real, and bringing it to Studio opens it up for team members who never made the switch. The Flow version comparison requires no setup — it’s just there.
Full deep-dive videos for each of these feature areas are on the Developer Program community hub: a full week of Developer Passport live streams with the PMs who built these features, including live Q&A.
Have questions about any of these features or want to share what you’re building? Drop into the ServiceNow Developer community or find me in the unofficial ServiceNow Developer Slack.
