Microsoft Copilot & Google Gemini Enterprise

AI Where Your Employees Already Work

The previous unit examined Amazon Bedrock as a developer-centric platform for building custom AI applications. Microsoft and Google take a fundamentally different approach: they embed AI directly into the productivity tools your employees already use every day.

This distinction matters because the fastest path to AI value in many banking organizations is not building a custom application -- it is making existing workflows faster. When a relationship manager can summarize a 40-page credit memo inside Word, or when an analyst can query loan data using natural language in Excel, the ROI is immediate and visible.

Microsoft Copilot Ecosystem

Microsoft's AI strategy spans two distinct but complementary products that banks should evaluate separately.

Microsoft 365 Copilot

Microsoft 365 Copilot (now branded Microsoft Copilot) embeds Large Language ModelsLarge Language Model (LLM)A neural network trained on vast amounts of text data that can understand and generate human language. LLMs power chatbots, document analysis, code generation, and many enterprise AI applications.See glossary directly into Word, Excel, PowerPoint, Outlook, and Teams. It operates on your organization's Microsoft Graph data -- emails, documents, calendars, chats -- to provide contextual AI assistance.

Banking applications:

• Word: Draft credit memos, summarize regulatory guidance
• Excel: Analyze loan portfolio data with natural language queries
• PowerPoint: Convert reports into client or executive presentations
• Outlook: Summarize email threads, draft responses, identify action items
• Teams: Summarize meetings and search transcripts for decisions

Governance considerations for banks: Copilot accesses your organization's data through Microsoft Graph, which means it inherits your existing Microsoft 365 permissions model. If an employee does not have access to a document through SharePoint permissions, Copilot cannot access it either. However, banks should audit their Graph permissions carefully -- many institutions discover overly permissive sharing settings when Copilot makes that data more accessible.

Copilot is now multi-model. Anthropic's Claude models run inside Copilot alongside OpenAI's. Anthropic acts as a Microsoft subprocessor for this, and its processing sits outside Microsoft's EU Data Boundary. The models are switched on by default for most commercial tenants and off by default in the EU, EFTA and UK -- though a separate in-app setting is on by default for EU, EFTA and UK tenants created after March 2026. Your third-party risk team should review these settings in the admin center, as it would any new subprocessor in a vendor's supply chain.

Agents and their control plane. Copilot Studio lets teams build agents. Microsoft Agent 365 (generally available May 2026, $15/user/month) is Microsoft's registry and "control plane for AI agents" -- its answer to the question "who approved this bot, and what can it touch?"

Government clouds: Microsoft lists Copilot as available in its GCC, GCC High and DoD government clouds. Anthropic models are not available in GCC High or DoD.

Microsoft Foundry

For banks building custom AI applications, Microsoft Foundry (called Azure AI Foundry until its rename in November 2025; Azure OpenAI now sits inside it) provides foundation modelFoundation ModelA large AI model trained on broad data that can be adapted to many tasks. Examples include OpenAI's GPT, Anthropic's Claude, Google's Gemini and Meta's Llama families. Banks evaluate these for capabilities, safety, and regulatory fit.See glossary access through Azure's enterprise cloud infrastructure. It offers OpenAI's GPT models and Anthropic's Claude, alongside models from other providers, within Azure's compliance boundary.

Key banking advantages:

• Data stays within your Azure tenant and selected region
• Integrates with your existing Microsoft identity management
• Azure Private Link support keeps traffic off the public internet
• Available in Azure Government regions for federal banking requirements
• Content filtering is enabled by default with configurable severity levels

Foundry is the natural choice for Azure-centric banks building custom APIAPI (Application Programming Interface)A standardized interface that allows software systems to communicate. In AI, APIs let your applications send prompts to a model and receive generated responses programmatically.See glossary-driven AI applications.

Google Gemini

Google's enterprise AI offering centers on Gemini -- its multimodal foundation model family -- delivered to employees and to developers.

Gemini in Google Workspace and Gemini Enterprise

Gemini is now included in Google Workspace business plans, integrating into Gmail, Docs, Sheets, Slides, and Meet with capabilities similar to Copilot. While Workspace adoption is lower in traditional banking than Microsoft 365, digital-first banks and fintech partners often use it as their primary productivity suite.

Gemini Enterprise (launched October 2025, replacing Agentspace, about $30 per user per month) is Google's separate agent product for employees. It can also work over Microsoft 365 data, which matters for Microsoft-centric banks.

Gemini Enterprise Agent Platform

Google's developer platform, formerly Vertex AI, became the Gemini Enterprise Agent Platform in April 2026. It includes Agent Studio for building agents and a Model Garden that offers Gemini alongside other providers' models, including Claude. It differentiates on:

• Multimodal and document AI: Strong processing of text, images, audio, and video -- relevant for banks handling check images, ID documents, and recorded customer interactions
• Grounding with Google Search: The ability to ground model responses in real-time web search results, useful for market research and competitive intelligence workflows

Large context windows and multimodality are no longer unique to Google -- leading Claude and GPT models offer them too.

Banking adoption note: Google Cloud's financial services footprint is smaller than AWS or Azure, but growing. Google's Workspace Gemini and its generative AI cloud services received FedRAMP High authorization in March 2025, under their earlier product names; confirm the status of the specific current service you plan to use.

Choosing Your Approach

The fundamental decision is not "Microsoft versus Google" -- it is "embedded productivity AI versus custom-built applications." Most banks will use both:

Cost Considerations

Microsoft 365 Copilot is licensed per-user per-month ($30/user/month for the enterprise add-on). For a bank with 10,000 employees, this represents $3.6M annually before any custom development. Microsoft also sells Microsoft 365 E7 ($99/user/month), which bundles E5, Copilot and Agent 365; agent usage can add consumption charges on top. Most banks start with targeted rollouts to high-value roles -- relationship managers, analysts, compliance officers -- rather than organization-wide deployment.

Microsoft Foundry and Google's Agent Platform use consumption-based pricing (per token). Frontier models cost significantly more per token than smaller models, so architecture decisions about which model to use for which task directly impact operating costs.