Google is targeting Australian enterprises with Gemini 2.5 and agentic Workspace integration, promising AI that can read, write and act across Gmail, Docs and Drive.
Tech & Ideas Desk is a contributing writer covering tech and public affairs for The Sydney Times.
Google is targeting Australian enterprises with Gemini 2.5 and agentic Workspace integration, promising AI that can read, write and act across Gmail, Docs and Drive without users having to switch between windows or train new workflows.
Gemini 2.5, released in 2026, is Google's multimodal frontier model with improved reasoning, coding and long-context performance. The model supports text, image, audio, video and code inputs and outputs, with a context window of one million tokens and native multimodality that eliminates the need for separate specialist models. Google says the model is its most capable yet, with a 7B mobile variant and a 32B desktop variant both released as open weights under the Apache 2.0 licence in late 2025.
For Australian enterprises, the practical advantage of Gemini 2.5 is its deep integration with Google Workspace, which many Australian organisations already use as their primary productivity suite. The Gemini app in Workspace can read emails in Gmail, draft replies, summarise long thread histories, extract action items, and create calendar entries without leaving the Gmail interface. In Docs, it can write first drafts, edit existing content, suggest structural improvements and generate presentations from raw notes. In Drive, it can locate documents across the entire organisation and extract relevant information without the user having to search manually.
Gemini 2.5's agentic capabilities allow it to take actions across Workspace on behalf of the user, such as scheduling meetings, sending follow-up emails, updating project timelines and generating reports from spreadsheet data. The model uses Google's Agent Development Kit to chain together actions across multiple services, so a single prompt can trigger a workflow that reads an email, updates a calendar, creates a task in a project management tool and sends a confirmation message.
The agentic layer is optional and opt-in. Users must grant explicit permission for Gemini to act on their behalf, and the actions are logged in an audit trail that administrators can review. The model does not have the ability to send emails without review, delete documents or modify shared files without the user's explicit consent in the same session.
For Australian enterprises in regulated industries, the compliance implications of agentic AI are significant. The Privacy Act 1988, the Australian Privacy Principles and sector-specific regulations such as APRA CPS 234 impose obligations on organisations that process personal information. Gemini 2.5's Workspace integration means that AI can access and process email, calendar and document data that contains personal information. Enterprises must ensure that the use of Gemini 2.5 is consistent with their privacy policies, that staff are trained on appropriate use, and that they can demonstrate compliance if challenged by the OAIC or a sector-specific regulator.
Google charges for Gemini 2.5 at US$1.75 per million input tokens and US$10.50 per million output tokens for the Pro tier, with AI-powered Workspace features billed at US$20 per user per month when added to a Workspace Business or Enterprise subscription. Enterprise customers can negotiate custom pricing based on volume and data residency requirements. The model is also available on Google Cloud's Vertex AI platform, where customers can deploy it in private VPCs and control data residency through region selection.
The cost of deploying Gemini 2.5 locally on Google Cloud is significantly higher than using open-source models on own infrastructure, but the integration with Workspace reduces the engineering effort required to build comparable agentic workflows in-house. For enterprises that are already heavily invested in Google Workspace, the total cost of ownership may be lower than building a comparable system using open-source models and custom orchestration.
Australian enterprises evaluating Gemini 2.5 should consider three factors. First, data residency: Google Cloud offers Australian regions, but enterprises must verify that the specific Gemini 2.5 features they use are available in the Australian region and that data does not leave Australian jurisdiction. Second, integration: the value of Gemini 2.5 is in its Workspace integration, not in its raw model performance. Enterprises that use Microsoft 365 or other productivity suites will need to evaluate whether the integration benefits justify the migration or whether comparable agentic capabilities can be built using open-source models and APIs. Third, governance: enterprises should establish clear policies on AI use, staff training and compliance monitoring before rolling out agentic AI across the organisation.
According to the Google Gemini enterprise blog, Gemini 2.5 is available to Gemini for Google Workspace customers and is already being used by enterprises across finance, healthcare and professional services. For more on Australian enterprise technology, see Tech & Ideas.
Direct inquiries, corrections, or documentation concerning this dispatch to our editorial newsroom desk.
Regulated industries are abandoning commercial AI APIs for self-hosted open-source models running on Australian infrastructure, citing compliance and cost advantages.
Australian enterprises are evaluating DeepSeek open-source models for local deployment as data residency requirements and MIT licensing lower the barrier to sovereign AI.
Nvidia-backed Firmus Technologies has raised A$500 million to advance Project Southgate, targeting 1.6 GW of AI compute capacity by 2028 across Tasmania and Melbourne.
PostgreSQL 18 will ship AI-native queries and improved vector search, strengthening the database's grip on the Australian startup stack.