Google's Gemini team made a significant announcement in May 2026. The Gemini app now reaches 900 million users across 230 countries, and two new agentic features (Gemini Spark and Daily Brief) represent the company's most direct push yet into territory where Claude has been building a quiet lead with Australian enterprise teams. If you are already running Claude workflows inside your organisation, here is what you need to know, and why the underlying fundamentals still point Claude's way.
What Google Actually Announced
On 19 May 2026, Google VP Josh Woodward detailed three product moves that matter for anyone evaluating AI strategy.
Gemini Spark is a 24/7 background agent built on Gemini 3.5. It runs in the cloud even after you close your browser, handles tasks across Gmail, Docs, Slides, and newly connected apps including Canva, OpenTable, and Instacart via MCP integrations. Spark is designed to act independently, checking back with users before high-stakes steps like sending emails or spending money. It is rolling out to US Google AI Ultra subscribers in beta, with broader availability planned later this year.
Daily Brief is a morning digest agent that pulls from Gmail and Google Calendar to produce a prioritised summary of what needs your attention. It is available today for US AI Plus, Pro, and Ultra subscribers.
Gemini Omni is a multimodal model that generates video from text and image inputs. It is aimed primarily at consumer and creator use cases rather than enterprise knowledge work.
The central claim from Google: Gemini is shifting from an assistant that answers questions into an active partner that does real work on your behalf. That framing will look familiar to anyone who has used Claude in Cowork mode or Claude Code. The question is not whether agentic AI is the direction of travel; it clearly is moving that way. The question is which platform gives Australian organisations the control, flexibility, and accountability they need when AI agents start touching real workflows.
Where Claude Has a Structural Advantage for Australian Organisations
The Gemini announcement is well-executed. But several structural factors still favour Claude for Australian enterprise deployments:
Trust and transparency documentation. Anthropic's Constitutional AI approach, detailed usage policies, and the published Responsible Scaling Policy give risk and legal teams a paper trail that Google's current agentic marketing language does not yet match. For APRA-regulated entities, that documentation gap is material.
API and deployment flexibility. Gemini Spark is US-only, tied to a consumer subscription tier, and embedded in Google Workspace. Claude via the Anthropic API can be integrated into any workflow regardless of your email provider, ERP, or cloud environment. A Sydney-based professional services firm spending $120,000 a year on knowledge-worker tooling should not have its AI strategy determined by inbox lock-in.
Claude Code for engineering teams. Claude Code's agentic capabilities including multi-file edits, terminal access, and long-horizon task completion are currently ahead for Australian software teams maintaining complex codebases. Gemini has developer tooling, but Code-level depth matters for organisations running custom systems.
Privacy Act and data residency. Australian organisations operating under the Privacy Act 1988 need clarity on where inference happens and whether data is retained for model training. Anthropic's enterprise agreements provide explicit commitments on this. The default Gemini consumer subscription does not.
No Workspace dependency. For Google Workspace organisations, Spark's tight integration is a genuine advantage. For the majority of Australian mid-market businesses running Microsoft 365, Atlassian stacks, or custom ERP environments, that same integration adds friction rather than removing it.
How to Think About Agentic AI Without Losing the Plot
The real question for any Australian leadership team is not 'Claude or Gemini' in the abstract. It is: what does autonomous AI action mean inside your operating environment, and who is accountable when something goes wrong?
Gemini Spark's own documentation notes that it will ask for confirmation before high-stakes actions. Claude-based agentic workflows carry the same category of guard. The difference lies in how each platform lets you configure, audit, and override that behaviour. For organisations subject to AUSTRAC reporting obligations or APRA's CPS 230 operational risk standards, an audit log of what an AI agent did, when, and why is not optional.
Three practical steps for Australian teams working through this:
First, map use cases against your risk profile. Background-running agents that touch finance, legal, or regulated customer data need a governance layer before deployment. Start with lower-stakes automation: meeting prep, internal document drafting, inbox triage. Then build your review cadence before expanding scope.
Second, separate the platform decision from the model decision. Claude can be accessed via the Anthropic API, via Cowork for non-technical teams, or via Claude Code for engineering. Choosing Claude does not mean choosing a single interface or vendor stack. Choosing Gemini Spark, at least today, means choosing Google infrastructure.
Third, treat AUD cost savings as a design constraint, not an afterthought. A Melbourne professional services firm spending $85,000 a year on knowledge-worker productivity tooling should expect AI to reduce rework and research time by 20 to 30 percent in year one when workflows are properly designed. That outcome requires deliberate implementation, not a subscription upgrade.
The Agentic Race Is Accelerating and Australian Teams Need a Position
Google's move is not isolated. OpenAI, Microsoft Copilot, and Anthropic are all shipping agentic capabilities at speed in 2026. For Australian businesses, the risk is not picking the wrong platform. It is waiting so long that competitors in your sector build AI-augmented processes before you do.
Claude's advantage is not that it outperforms Gemini on every benchmark. It is that the combination of strong document understanding, a flexible API, transparent deployment options, and AU-specific implementation support gives organisations a clear route to production that does not depend on a single consumer platform or geographic rollout schedule.
The Gemini Spark announcement is worth watching. For now, it reinforces the broader argument that agentic AI is where enterprise software is headed. The sensible move for Australian teams is to build on Claude now, in controlled ways, with clear governance in place, rather than defer until market pressure forces a rushed decision.
If your team is working out where to start with Claude, or what your current Claude deployment should be doing differently, book a session with Automata AI.
