Gemini Enterprise updates and Salesforce integration expansion

Agentic workflows and in-window editing

Gemini Enterprise differs from Google Workspace with Gemini because it functions as an ecosystem-independent platform rather than a simple in-app assistant. While the Workspace version helps with writing and summarizing, Gemini Enterprise uses AI agents to automate processes across different business systems. Gemini Enterprise shifts from simple chat to agentic workflows through long-running agents and secure sandboxes. These sandboxes allow agents to iterate in loops and build their own tools to solve complex problems without constant human prompts. Canvas provides a way for users to open and edit documents or slides directly within the chat window. This removes the frustration of looping back and forth between multiple prompts to refine language or layouts without having to export content to external applications. A2UI creates bespoke graphical interfaces, such as editable Salesforce tables, directly within a session. Users also use "bring your own MCP" to connect organization-specific Model Context Protocol servers to Gemini Enterprise. This setup centralizes authentication and authorization so users only access data they have permission to see. The new Inbox feature allows users to delegate tasks and operate asynchronously instead of staying in the chat window. Projects allow teams of users to collaborate with agents and share context in an organized way. A new Data Science Agent, currently in preview, automates data wrangling and ingestion by generating multi-step plans for training and inferencing. This agent helps users find patterns and accelerates complex model development. Gemini Enterprise unifies information from disparate systems like Jira, ServiceNow, OneDrive, SharePoint, and Confluence.

Salesforce and Google Cloud connectivity

The Salesforce and Google Cloud partnership connects Agentforce and Gemini to reduce data movement. This integration allows agents to act across Slack, Google Workspace, and CRM systems, providing agents with the ability to move work between tools. Agentforce Sales agents work inside Gemini Enterprise to engage leads, create meeting briefs, and update CRM records. Gemini models power the Agentforce Reasoning Engine through the Atlas Reasoning Engine, which includes multimodal understanding of text, image, and video. Because the integration utilizes zero-copy access via Google Lakehouse, Agentforce can read data from Google Lakehouse without moving information, which reduces data fragmentation and simplifies the security posture for many large enterprise organizations. A2UI and MCP servers work together to improve how agents interact with external systems. You might find these specific interface tools useful for maintaining workflow continuity. Users can request outputs from Slack, where Slackbot pulls relevant context from Slack threads and Google Workspace files. Gemini becomes accessible in Slack as an assistant that can summarize a Google Meet transcript alongside a Slack thread. Salesforce workloads run natively on Google Cloud’s AI-optimized infrastructure via Hyperforce. For commerce leaders, Commerce Cloud powers product discovery inside Google Search and the Gemini app. The partnership also includes Tableau MCP, which adds governed analytics with row-level security on every agent query. To support communication, the companies developed the Agent2Agent Protocol and the Agent Payments Protocol to enable agents to transact. Will these deep integrations eventually eliminate the need for third-party AI add-ons in CRM systems?

Model availability and pricing tiers

Google provides several pricing tiers and model options for enterprise and developer needs. On October 20, 2026, Google removed Gemini 2.5 Pro from in-country deployments in Canada and Japan. Users in those regions now use Gemini 3.5 Flash. The Gemini 2.5 Flash model faces deprecation on October 16, 2026, according to recent schedule updates. Developers using Gemini 4 Argon for Workflow Builder can access specialized reasoning capabilities. Thinking tokens in models like Gemini 2.5 Pro increase expenses because Google bills these as output tokens.

The Gemini API offers different tiers for developers. The free tier in Google AI Studio allows for prototyping with certain limits. The paid tier includes access to context caching and higher rate limits. For enterprise deployments, the Gemini Enterprise Agent Platform provides dedicated support channels and advanced security. Gemini 3.8 Flash arrived on September 2, 2026, with an introductory price of $0.75 per million input tokens. Gemini 3.8 Flash-Lite serves as a replacement for earlier versions to support high-throughput production. Developers can also use Gemini 3.8 Live for real-time voice applications. Vertex AI pricing runs 10-20% higher than AI Studio for the same model. The premium buys enterprise data handling, SLA guarantees, and integration with Google Cloud’s IAM and VPC controls. Gemini 4 Argon is priced at an introductory $2.00 per million tokens until November 30, 2026.