Smartsheet has opened the floodgates for multi‑model AI inside its work management platform. On June 11, 2026, the Bellevue‑based company announced that enterprise customers can now connect Microsoft 365 Copilot, OpenAI ChatGPT, and Google Cloud Gemini Enterprise directly to its Model Context Protocol (MCP) server. The integration, branded Smart Assist, lets teams summon any of those three AI assistants from within sheets, dashboards, and automated workflows — all governed by the same permission model that protects the rest of their Smartsheet data.

It’s a sharp departure from the single‑assistant AI strategies most SaaS vendors have pursued. Instead of locking customers into one ecosystem, Smartsheet is betting that large organizations need flexibility to match the right model to the right job, whether that’s drafting a project summary in Word, classifying invoice line items, or generating Gantt charts from natural language prompts.

What the MCP server brings to the table

Model Context Protocol, originally open‑sourced by Anthropic in 2024, describes a standard way for AI models to request information from business applications without copying sensitive data. An MCP server sits inside the application, exposes controlled “resources” and “tools,” and lets approved models query those resources on behalf of a user. Smartsheet launched its MCP server in early 2026, initially supporting a handful of Anthropic Claude models. With today’s update, the protocol layer now speaks to three additional platforms: Microsoft’s Copilot infrastructure (including the Copilot extensibility stack that underpins Copilot in Excel, Word, and Teams), OpenAI’s ChatGPT Enterprise API, and Google Cloud’s Gemini Enterprise endpoints.

That means a project manager can, for example, highlight five rows of a sheet and ask Copilot to “create a status report in Word using the Project Status template,” or have ChatGPT surface risks based on historical schedule variance stored in Smartsheet. Meanwhile, a data analyst might call Gemini Enterprise inside a dashboard to generate a Python script that visualizes resource allocation, all without leaving the Smartsheet browser tab.

How Smart Assist works in practice

Smart Assist is exposed through a uniform prompt bar that appears in sheets, reports, and the Smartsheet desktop app. Users with the appropriate license — Smartsheet Enterprise plan plus a corresponding license for the AI service they want to use — pick their assistant from a drop‑down list next to the prompt bar. The system then authenticates against the customer’s existing Microsoft Entra ID, OpenAI account, or Google Cloud identity, no separate API key management required.

Once connected, the MCP server presents a curated set of resources: sheet metadata, cell values, comments, attachment summaries, and automation rules. The AI model never sees the raw file store. Instead, the MCP server acts as a broker, returning only the data necessary to fulfill a prompt. For instance, if someone asks “Which tasks are overdue and who owns them?,” the server delivers a structured JSON payload with task names, due dates, and assignee emails, but not the entire sheet.

Early enterprise adopters such as a Fortune 500 construction firm and a global healthcare network have been quietly testing Smart Assist since March. According to a Smartsheet spokesperson, the most common use case so far is the “cross‑app memo” — a Copilot‑fuelled workflow where a sheet summary is drafted in Smartsheet, polished in Word, and then published to a Teams channel with a single voice click. Another frequently cited scenario is risk scoring: ChatGPT examines project‑schedule data in Smartsheet against a company’s historical benchmarks stored elsewhere, returning a 1–10 risk score and a short explanation.

Governance that IT departments will appreciate

The specter of shadow AI has made many enterprises nervous about giving rank‑and‑file employees free‑form access to large language models. Smartsheet’s approach is to wrap every Smart Assist interaction in the same administrative controls that already govern sheets and dashboards.

Admins define which AI models are available to which groups, restrict which sheets can be queried by an external model, and enforce data‑loss‑prevention rules. For Copilot connections, the integration leans on Microsoft Purview sensitivity labels, so a sheet tagged “Highly Confidential” won’t be accessible to Copilot even if a user tries to prompt it. For ChatGPT Enterprise, Smartsheet uses API‑level policies that OpenAI introduced in early 2026, which prevent prompts and output from being used for model training. Google Cloud Gemini Enterprise connections inherit the organization’s VPC Service Controls and Assured Workloads settings, ensuring data remains within the prescribed compliance boundary.

“We’re not just handing people a chat box and hoping for the best,” said CTO Praerit Garg during a press briefing. “Every prompt flows through the same governance plane that audits who viewed what cell. If a user doesn’t have permission to see a column in Smartsheet, an AI model won’t see it either.”

What this means for Windows shops and the Microsoft ecosystem

For organizations that have standardized on Windows 11 and the Microsoft 365 suite, the Copilot integration is the headline item. Because Smartsheet already offers a best‑in‑class Microsoft Teams app and a set of Power Automate connectors, adding Copilot to its MCP server effectively transforms Smartsheet into a first‑class data source for Microsoft’s AI ecosystem.

A Windows user working in Excel can now ask Copilot to fetch live project data from Smartsheet and drop it into a spreadsheet without switching context. Similarly, a project lead can dictate a status update into the Copilot pane in Outlook and have it insert a Smart‑Assist‑generated summary that pulls directly from a Smartsheet report. The integration uses the same Copilot extensibility model that third‑party ISVs like Adobe and ServiceNow have adopted, meaning IT admins can manage it through the Microsoft 365 Admin Center just like any other Copilot extension.

It also deepens Smartsheet’s relationship with Windows enterprise customers. Two years after Microsoft itself revamped Planner and Project, Smartsheet has carved out a niche for cross‑functional work that spans marketing campaigns, product launches, and construction RFPs — areas where a native Microsoft tool may not be the best fit. By plugging into Copilot, Smartsheet makes itself more visible to the millions of knowledge workers who live inside Teams and Outlook all day.

The competitive landscape

Smartsheet isn’t the first collaboration vendor to embrace multi‑model AI. Asana introduced a similar “AI Studio” in late 2025 that allows customers to choose between Anthropic, OpenAI, and Google models. Monday.com recently released an “AI Blocks” framework that supports custom models deployed on AWS Bedrock. What sets Smartsheet apart is the depth of its governance layer and its decision to support Copilot, which Monday.com and Asana have yet to do.

That Copilot support could prove sticky. Microsoft CEO Satya Nadella has repeatedly stated that Copilot extensibility is a cornerstone of the company’s AI platform play. Smartsheet, by becoming an early anchor tenant in that ecosystem, may gain a distribution advantage as Copilot matures from a writing assistant into a full‑fledged orchestration layer.

Two big unknowns: cost and compatibility

Smartsheet has not yet published a pricing page for Smart Assist. In the press briefing, executives said the feature is included in the existing Enterprise license “at launch” but warned that consumption‑based pricing may follow once usage patterns stabilize. Customers who use Copilot will also need Microsoft 365 Copilot licenses, which currently cost $30 per user per month. Similarly, ChatGPT Enterprise and Gemini Enterprise carry their own per‑seat or per‑token charges.

There is also an open question about how Smart Assist will handle older Windows environments. The Smartsheet desktop app, a Progressive Web Application that runs on Edge WebView2, works on Windows 11 and fully patched Windows 10 22H2. While the AI prompts are processed server‑side, the prompt bar depends on WebView2 version 120 or later. Organizations still running unpatched Windows 10 builds — support for which officially ended in October 2025 — may encounter UI glitches. A Smartsheet support document recommends upgrading to Windows 11 or enabling Extended Security Updates before deploying Smart Assist.

What early testers are saying

Several participants in the Smartsheet Community have shared first impressions. One IT manager at a Midwest manufacturing firm noted that “the Copilot connection practically pays for itself by automating our weekly status meetings.” Another user, a program coordinator at a university, praised the Gemini Enterprise option: “We’ve been building custom Google Cloud Functions to process student enrollment data; now a simple prompt in Smartsheet does the same thing in seconds.”

Not all feedback has been glowing, however. Some testers reported latency when the MCP server routes prompts through a company’s on‑premises data gateway, a common requirement for highly regulated industries. A Smartsheet engineer active on the forum acknowledged the issue and said a performance optimization patch is scheduled for July 2026.

A template for the AI‑governed workplace

Smart Assist reflects a broader industry shift toward what Gartner calls “agentic workflows” — sequences of tasks where AI agents act semi‑autonomously under human supervision. By making its MCP server a neutral hub that can talk to three competing model families, Smartsheet is positioning itself as the governance layer for those workflows. That layer authenticates the user, scopes the data, logs the prompt, and returns the answer — all without the model ever touching the corporate file server.

If the strategy succeeds, other SaaS vendors may follow suit. Already, several enterprise software makers are exploring MCP servers of their own, and the MCP specification continues to gain traction through the Linux Foundation’s AI and Data group. For Windows users, the immediate payoff is simpler: less time copying data between apps, fewer context switches, and an AI assistant that actually understands what’s inside their most complex Smartsheet project.