Smartsheet on June 11, 2026, expanded its MCP Server to connect Google Cloud Gemini Enterprise, Microsoft Copilot, and ChatGPT to live Smartsheet work data, while also introducing Smart Assist as an in-product AI companion for users who stay inside the platform. The announcement is less about another chatbot integration than about a more serious contest over where enterprise AI gets its context. Smartsheet is betting that the durable layer in business AI will not be the assistant interface, but the operational system that knows what work is actually happening. That is a useful bet, but it also moves AI from the low-risk world of drafting and summarizing into the messier world of modifying live business records.
The headline names are familiar: Gemini Enterprise, Microsoft Copilot, ChatGPT, and Claude. The more important phrase is live work data. Smartsheet is not merely making it easier to ask an AI assistant about a project plan; it is trying to make Smartsheet the operating memory that external assistants consult before they answer, escalate, summarize, or act.
That matters because the enterprise AI market has spent the past two years learning an awkward lesson. General-purpose assistants are impressive when they rewrite a memo, draft a formula, or summarize a document. They are much less useful when they do not know which dependency is blocked, which regional launch has slipped, which approval is waiting on legal, or which program manager quietly changed the due date yesterday afternoon.
Smartsheet’s argument is that this context already lives in its platform. Projects, workflows, dependencies, owners, comments, attachments, dashboards, and status updates are the working tissue of a company. If AI cannot see that tissue, it behaves like a talented consultant who arrived without reading the packet.
The addition of Gemini Enterprise is significant because it widens the circle beyond Anthropic’s Claude, which was Smartsheet’s earlier flagship MCP integration. Adding Microsoft Copilot and ChatGPT to the rollout signals that Smartsheet is not trying to crown a single AI winner. It is trying to become useful whichever assistant an enterprise has already standardized on.
That is a pragmatic position. Large organizations rarely converge on one AI tool cleanly. Procurement, existing cloud commitments, developer preferences, data residency requirements, and departmental politics all pull in different directions. A finance team may live in Microsoft 365, a product team may prefer ChatGPT, a software group may use Claude Code, and an executive office may be piloting Gemini Enterprise because of a Google Cloud agreement.
Smartsheet’s move says: let the assistant wars continue; the work layer still needs to be governed, current, and actionable.
In Smartsheet’s case, that means the MCP Server can expose work-management objects through a protocol that compatible AI clients understand. The assistant can ask for relevant project information, inspect sheet structures, retrieve status, and, where permitted, create or update work items. The exact user experience varies by assistant, but the architecture is the same: AI reaches into Smartsheet through governed pathways rather than through screenshots, exports, pasted tables, or brittle one-off scripts.
That distinction is not academic. A lot of early enterprise AI usage has been a theater of copy and paste. Workers pull data out of systems, drop it into an assistant, ask for an analysis, then manually transpose the answer back into the original workflow. That can save time, but it also creates obvious problems: stale data, accidental disclosure, missing permissions, and no clean audit trail of what the AI saw or changed.
MCP is an attempt to move from theater to plumbing. It is not magic, and it does not eliminate the need for governance, but it does create a more coherent model. The assistant becomes a client. The enterprise application remains the system of record. Permissions, actions, and auditability can be handled closer to the data rather than improvised in the prompt box.
This is why Smartsheet’s expansion lands differently from a conventional “now with AI” product update. The company is not just placing a chatbot beside the product. It is positioning Smartsheet as a tool-enabled backend for whichever chatbot, agent, or enterprise AI surface the customer already uses.
The more assistants Smartsheet supports, the more credible that posture becomes. Claude gave the company a first mover’s story. Gemini Enterprise, Copilot, and ChatGPT give it a neutrality story.
The word context gets abused in AI marketing, but here it has a concrete meaning. Context is the approved scope of a project. It is the difference between a task that is overdue and a task that was intentionally deferred because a supplier missed a shipment. It is the note buried in a comment thread explaining that a launch risk is no longer material. It is the dependency between two workstreams that only becomes visible when an assistant can inspect the underlying work graph.
This is where Smartsheet has a plausible claim. Its customers often use the platform for cross-functional work that is too structured for email and too fluid for traditional systems of record. Construction programs, marketing launches, IT rollouts, procurement processes, field operations, and transformation programs are exactly the kinds of work where the plan changes frequently and the consequences of outdated information are expensive.
If an assistant can read and reason over that data, it becomes more than a writing aid. It can identify bottlenecks, summarize project health, generate status updates, draft follow-up actions, and help users create or modify Smartsheet assets. That is the shift from individual productivity to operational assistance.
But that shift also raises the stakes. When AI summarizes a document badly, the failure is irritating. When AI updates a row, creates a workflow, or changes a project artifact based on misunderstood intent, the failure can propagate into real work. Smartsheet’s claim that its MCP Server respects user permissions and sits within existing governance structures is therefore central, not decorative.
The enterprise AI winners will be judged less by the sparkle of their demos than by how safely they let machines act around fragile business processes.
Vendor adoption numbers always deserve careful handling. They are selected to tell a growth story, and “AI actions” is a broad metric whose business value depends on what those actions actually do. A tool call that fetches a sheet is not the same as a tool call that updates a project plan, and neither automatically proves return on investment.
Even so, the shape of the numbers is interesting. Smartsheet says the first 10 days of June generated more than 860,000 AI actions, with record daily activity on two consecutive days as 1,767 and 1,825 organizations were active. It also says nearly 3,000 net-new organizations joined in the last 30 days, with close to 700 new organizations discovering the server each week.
That pattern looks less like a small proof-of-concept group hammering a demo environment and more like broadening organizational usage. The Asia Pacific and Japan figures tell a similar story: in May, Smartsheet reported 561 unique users across 334 customer plans in the region, generating nearly 60,000 MCP AI actions. That works out to meaningful repeated use rather than a one-time curiosity click.
The most important statistic is the one Smartsheet says ties AI to execution: nearly one in three AI-driven actions creates, updates, or modifies live work, rather than simply retrieving information. If accurate, that is the line between AI as a search layer and AI as an operational actor. It also explains why IT leaders will look at the same number and feel both encouraged and nervous.
Retrieval is easier to govern because it mostly concerns what the assistant can see. Write access is harder because it concerns what the assistant can change. The moment AI starts modifying live work, the enterprise conversation shifts from “Is this useful?” to “Who approved this, can we audit it, and how do we reverse it?”
That matters because not every user wants to work through an external assistant. In many companies, the people closest to operational work are not living in a dedicated AI client all day. They are in the project plan, the intake sheet, the dashboard, or the approval workflow. For them, opening a separate AI surface may be just enough friction to make the feature irrelevant.
Smart Assist also gives Smartsheet a way to preserve the in-product experience if the external assistant market fragments further. Today’s enterprise AI stack is unstable. Google, Microsoft, OpenAI, Anthropic, and others are all trying to make their assistants the front door for work. Customers may adopt several at once, switch vendors, or restrict certain tools by department or geography.
A native assistant gives Smartsheet a fallback: even if the customer does not standardize on Claude, Gemini, Copilot, or ChatGPT for Smartsheet workflows, the platform itself can still provide AI-powered interaction. It is a hedge against interface volatility.
The feature also changes the product’s center of gravity. Smartsheet has long been a platform that mixes spreadsheet familiarity with project and workflow management. Smart Assist pushes it toward conversational work management, where users can describe intent rather than manually assemble every workflow, dashboard, or update.
That is attractive, especially for frontline and operations-heavy teams that do not have dedicated administrators for every process. But it also requires discipline. Natural language is flexible; business process is unforgiving. The value of Smart Assist will depend on how well it translates messy requests into precise, reviewable actions.
DPR’s Matthew Feagin described Smartsheet as the backbone for managing dynamic parts of complex projects, from data centers to healthcare facilities. His point was not that AI can write a nicer status update. It was that frontline workers can use natural language to build workflows, test ideas, and get answers faster, with less dependency on specialized configuration.
That is precisely the kind of use case where AI connected to live work data can matter. A worker trying to understand a blocker does not need a generic explanation of project management. They need to know which task is blocked, who owns the next step, what dependency is causing the delay, whether the issue appears elsewhere, and what action can be taken within the system.
If the assistant can help create a tailored Smartsheet solution for a field problem, it reduces the gap between process design and process execution. That has been one of the quiet promises of no-code and low-code platforms for years: people who understand the work should be able to shape the system without waiting for a central IT queue.
AI gives that promise a new interface. Instead of dragging columns, configuring rules, or learning formulas, the user can explain the desired workflow. The system can then propose or create the structure. Done well, that can accelerate operational adaptation.
Done badly, it can create a sprawl of half-understood automations and inconsistent process artifacts. The history of enterprise software is full of tools that empowered business users and then left IT to clean up the governance mess. Smartsheet’s governance claims will matter most in these decentralized environments.
That is powerful because Microsoft’s enterprise advantage is not merely the Copilot brand. It is the Microsoft 365 substrate: identities, files, calendars, Teams chats, meetings, SharePoint sites, compliance controls, and admin tooling. A Copilot user asking about a project may not want an answer that only knows the project sheet. They may want the broader context of the meeting where the decision was made, the document where the requirement changed, and the Smartsheet workflow where the task remains open.
Smartsheet’s presence in that assistant surface could make cross-system answers more useful. It could also make governance more complicated. The assistant experience may feel unified to the user, but the underlying permissions, retention policies, audit logs, and action boundaries span multiple systems. IT teams will need to know where the answer came from and which system authorized the action.
There is also a competitive subtext. Microsoft wants Copilot to be the orchestration layer for enterprise work. Smartsheet wants to ensure that, even if Copilot becomes the interface, Smartsheet remains the authoritative work-management layer. That is a familiar platform tension: the front-end assistant wants to absorb the workflow, while the specialized application wants to retain the workflow’s structure and data.
Customers may benefit from that tension if it produces better integrations and less lock-in. They may suffer if the result is a confusing matrix of licensing, capabilities, regional availability, and partial feature parity. Smartsheet says Microsoft Copilot and ChatGPT connections are available first to US customers, while Smart Assist, the MCP Server, and connections to Claude and Google Cloud Gemini Enterprise are available to all customers. That staggered availability is a reminder that AI rollouts are still constrained by geography, vendor readiness, and compliance posture.
For administrators, the practical question is not whether Copilot can “talk to Smartsheet.” It is whether the resulting workflow is predictable enough to support at scale.
They are not the whole security story. Enterprise AI introduces risks that are not identical to traditional API integrations. A normal integration usually performs known operations in known patterns. An AI assistant may choose tools dynamically based on a user’s natural-language prompt, intermediate reasoning, and the data it retrieves along the way.
That difference changes how organizations should think about control. It is not enough to ask whether a user has permission to edit a sheet. IT also needs to understand when an assistant is allowed to make edits on the user’s behalf, whether the user must explicitly approve write actions, how prompts and responses are logged, how tool calls are audited, and how sensitive data is protected when responses cross assistant boundaries.
There is also the problem of intent. A user may ask, “Clean up the overdue tasks and notify owners,” expecting a draft plan. An assistant might interpret that as permission to change statuses, create comments, or trigger alerts. The safest implementations keep humans in the loop for consequential writes and make proposed changes visible before committing them.
Smartsheet’s claim that nearly a third of AI actions involve creating, updating, or modifying live work makes these controls urgent. Read-only AI is already a governance challenge. Read-write AI is a change-management challenge.
The other concern is data minimization. Work-management platforms often contain more than task names. They can include customer details, budget hints, personnel issues, vendor disputes, legal constraints, and operational risks. Connecting that context to multiple assistants increases the need for clear policies about which AI tools are approved for which data classes.
The promise of MCP is standardized connectivity. The risk is standardized overexposure if organizations connect first and design controls later.
The optimism is understandable. Enterprises are tired of building custom integrations for every system pair. Developers are tired of brittle glue code. Business users are tired of waiting for dashboards and reports that are out of date by the time they are published. A standard protocol for AI-to-application interaction sounds like the missing connective tissue.
But open plumbing does not guarantee open outcomes. The major AI vendors still have incentives to pull users into their own interfaces, clouds, identity systems, and billing models. Support for MCP can coexist with platform lock-in elsewhere. A company may technically be able to connect multiple assistants to Smartsheet while practically finding that one vendor’s admin controls, model quality, or licensing terms make it the path of least resistance.
Smartsheet is navigating that reality by connecting to all the major assistants rather than betting solely on one. That is customer-friendly in theory. In practice, customers will still need to test each assistant’s behavior, latency, action handling, permission mapping, and administrative experience.
The same Smartsheet data may produce different user experiences depending on whether it is accessed through Claude, Gemini Enterprise, Copilot, ChatGPT, or Smart Assist. Models differ. Tool-use behavior differs. Enterprise admin surfaces differ. The protocol may standardize the doorway, but the room on the other side still changes.
That means organizations should resist treating MCP support as a checkbox. The real evaluation is operational: does the assistant retrieve the right data, preserve permissions, ask before changing important records, produce auditable actions, and behave consistently under messy human prompts?
The standard is useful because it lowers integration friction. It does not remove the need for testing, governance, and user education.
A dashboard is a designed answer to a known question. An assistant is an improvised answer to a question that may not have existed yesterday. In volatile projects, that matters. Leaders still need standard reporting, but workers often need situational answers: what changed since the last review, which risk has no owner, which dependency is now blocking the launch, which tasks can be safely deferred, and what should be escalated before Friday.
That is the appeal of connecting AI directly to live work data. It lets users ask questions that were not anticipated when the dashboard was built. It also lets them move from analysis to action in the same flow.
This is why Smartsheet’s line about “where the work happens” is more than marketing. The company is trying to make the work-management layer conversational without surrendering the underlying structure that makes it useful. A spreadsheet-like grid remains valuable because it imposes order. AI becomes valuable when it helps users navigate and manipulate that order without becoming spreadsheet mechanics.
There is a danger, however, in overstating the post-dashboard future. Executives still need consistent metrics. Compliance teams still need fixed reports. Program offices still need baselines. AI-generated answers can be fluid, but fluidity is not always a virtue.
The likely future is hybrid. Dashboards will remain the canonical view for recurring governance. Assistants will become the exploratory and operational layer around them. Smartsheet is trying to serve both, which is the right posture for enterprise software that cannot afford to treat every workflow as a chat.
But it increases the burden of judgment. Administrators must decide which assistants can connect, which users can use them, what actions require approval, what data classes are off-limits, and how to monitor usage without smothering adoption. The technical connection is only the beginning.
This is particularly true in organizations where Smartsheet is used by business teams without heavy central oversight. Work-management platforms often spread because they are useful and flexible. That flexibility can create a long tail of departmental processes that IT only partially understands. Adding AI actions into that environment raises the possibility of faster work and faster mistakes.
The right governance model should be proportional. A team using AI to summarize project status may not need the same controls as a team using AI to update customer-facing implementation plans. A sandboxed assistant creating draft sheet structures is different from an assistant modifying production workflows. Not all AI actions carry the same risk.
Smartsheet’s adoption metrics suggest users are moving quickly. IT rarely moves at the same speed, especially in regulated or risk-sensitive organizations. That mismatch is where shadow AI tends to grow.
The best version of this rollout gives administrators enough visibility to embrace the tools rather than block them. The worst version encourages users to find unofficial routes because official governance is too slow, too opaque, or too restrictive. Smartsheet’s success will depend partly on whether its controls feel like guardrails rather than gates.
The meaningful questions are harder. Did projects finish faster? Were errors reduced? Did teams spend less time in status meetings? Did risk surface earlier? Did frontline users create better workflows without admin intervention? Did the AI actions replace busywork or merely create a new stream of outputs to review?
Smartsheet’s DPR example points toward plausible value in complex, operational environments. The company’s claim that nearly one in three actions modifies live work suggests users are not merely asking novelty questions. Still, the next phase of credibility will require more than usage counts.
This is not unique to Smartsheet. The whole enterprise AI industry is moving from adoption theater to ROI scrutiny. Executives who approved early AI pilots are now asking what changed in the business. Vendors that can connect AI activity to cycle time, error reduction, risk mitigation, or revenue-impacting workflows will have stronger stories than vendors counting prompts.
Smartsheet may be better positioned than many because work management produces measurable artifacts. Tasks have owners, dates, dependencies, statuses, and completion histories. If AI genuinely improves execution, the platform should be able to show it.
That evidence will matter. The first wave of enterprise AI was sold on possibility. The next wave will be renewed, expanded, or cut based on operational proof.
The challenge is equally clear. The more useful these AI connections become, the closer they get to the systems where mistakes matter. Smartsheet’s expansion gives customers a glimpse of a more practical enterprise AI future: not a chatbot hovering beside work, but an assistant wired into the work itself. Whether that future feels empowering or reckless will depend on how well Smartsheet, Google, Microsoft, OpenAI, Anthropic, and enterprise IT teams turn AI action into something users can trust.
Smartsheet Wants to Be the Work Graph Beneath Every Assistant
The headline names are familiar: Gemini Enterprise, Microsoft Copilot, ChatGPT, and Claude. The more important phrase is live work data. Smartsheet is not merely making it easier to ask an AI assistant about a project plan; it is trying to make Smartsheet the operating memory that external assistants consult before they answer, escalate, summarize, or act.That matters because the enterprise AI market has spent the past two years learning an awkward lesson. General-purpose assistants are impressive when they rewrite a memo, draft a formula, or summarize a document. They are much less useful when they do not know which dependency is blocked, which regional launch has slipped, which approval is waiting on legal, or which program manager quietly changed the due date yesterday afternoon.
Smartsheet’s argument is that this context already lives in its platform. Projects, workflows, dependencies, owners, comments, attachments, dashboards, and status updates are the working tissue of a company. If AI cannot see that tissue, it behaves like a talented consultant who arrived without reading the packet.
The addition of Gemini Enterprise is significant because it widens the circle beyond Anthropic’s Claude, which was Smartsheet’s earlier flagship MCP integration. Adding Microsoft Copilot and ChatGPT to the rollout signals that Smartsheet is not trying to crown a single AI winner. It is trying to become useful whichever assistant an enterprise has already standardized on.
That is a pragmatic position. Large organizations rarely converge on one AI tool cleanly. Procurement, existing cloud commitments, developer preferences, data residency requirements, and departmental politics all pull in different directions. A finance team may live in Microsoft 365, a product team may prefer ChatGPT, a software group may use Claude Code, and an executive office may be piloting Gemini Enterprise because of a Google Cloud agreement.
Smartsheet’s move says: let the assistant wars continue; the work layer still needs to be governed, current, and actionable.
MCP Turns Integration From Demo Theater Into Plumbing
The reason this expansion is possible is the Model Context Protocol, usually shortened to MCP. The acronym may sound like one more piece of AI vendor alphabet soup, but the idea behind it is straightforward: give AI tools a standardized way to connect to external systems and call tools against them. Instead of every assistant needing a bespoke integration with every enterprise app, MCP offers a common pattern for exposing data and actions.In Smartsheet’s case, that means the MCP Server can expose work-management objects through a protocol that compatible AI clients understand. The assistant can ask for relevant project information, inspect sheet structures, retrieve status, and, where permitted, create or update work items. The exact user experience varies by assistant, but the architecture is the same: AI reaches into Smartsheet through governed pathways rather than through screenshots, exports, pasted tables, or brittle one-off scripts.
That distinction is not academic. A lot of early enterprise AI usage has been a theater of copy and paste. Workers pull data out of systems, drop it into an assistant, ask for an analysis, then manually transpose the answer back into the original workflow. That can save time, but it also creates obvious problems: stale data, accidental disclosure, missing permissions, and no clean audit trail of what the AI saw or changed.
MCP is an attempt to move from theater to plumbing. It is not magic, and it does not eliminate the need for governance, but it does create a more coherent model. The assistant becomes a client. The enterprise application remains the system of record. Permissions, actions, and auditability can be handled closer to the data rather than improvised in the prompt box.
This is why Smartsheet’s expansion lands differently from a conventional “now with AI” product update. The company is not just placing a chatbot beside the product. It is positioning Smartsheet as a tool-enabled backend for whichever chatbot, agent, or enterprise AI surface the customer already uses.
The more assistants Smartsheet supports, the more credible that posture becomes. Claude gave the company a first mover’s story. Gemini Enterprise, Copilot, and ChatGPT give it a neutrality story.
The Real Product Is Context, Not Conversation
Pratima Arora, Smartsheet’s chief product and technology officer, framed the announcement around a common enterprise frustration: the problem is not access to AI, but the fact that AI does not know how the organization works. That line is vendor positioning, but it identifies a real failure mode. A model may understand language, code, and business jargon, yet still know nothing about the actual commitments inside a company.The word context gets abused in AI marketing, but here it has a concrete meaning. Context is the approved scope of a project. It is the difference between a task that is overdue and a task that was intentionally deferred because a supplier missed a shipment. It is the note buried in a comment thread explaining that a launch risk is no longer material. It is the dependency between two workstreams that only becomes visible when an assistant can inspect the underlying work graph.
This is where Smartsheet has a plausible claim. Its customers often use the platform for cross-functional work that is too structured for email and too fluid for traditional systems of record. Construction programs, marketing launches, IT rollouts, procurement processes, field operations, and transformation programs are exactly the kinds of work where the plan changes frequently and the consequences of outdated information are expensive.
If an assistant can read and reason over that data, it becomes more than a writing aid. It can identify bottlenecks, summarize project health, generate status updates, draft follow-up actions, and help users create or modify Smartsheet assets. That is the shift from individual productivity to operational assistance.
But that shift also raises the stakes. When AI summarizes a document badly, the failure is irritating. When AI updates a row, creates a workflow, or changes a project artifact based on misunderstood intent, the failure can propagate into real work. Smartsheet’s claim that its MCP Server respects user permissions and sits within existing governance structures is therefore central, not decorative.
The enterprise AI winners will be judged less by the sparkle of their demos than by how safely they let machines act around fragile business processes.
The Adoption Numbers Suggest Curiosity Has Become Habit
Smartsheet says MCP Server usage has climbed rapidly since the Claude integration launched in March. The company reported more than 22,000 unique users globally and more than 3 million AI actions over the period. Weekly active users reportedly rose from fewer than 1,000 at launch to more than 9,000, while weekly tool calls climbed from 42,000 to more than 700,000.Vendor adoption numbers always deserve careful handling. They are selected to tell a growth story, and “AI actions” is a broad metric whose business value depends on what those actions actually do. A tool call that fetches a sheet is not the same as a tool call that updates a project plan, and neither automatically proves return on investment.
Even so, the shape of the numbers is interesting. Smartsheet says the first 10 days of June generated more than 860,000 AI actions, with record daily activity on two consecutive days as 1,767 and 1,825 organizations were active. It also says nearly 3,000 net-new organizations joined in the last 30 days, with close to 700 new organizations discovering the server each week.
That pattern looks less like a small proof-of-concept group hammering a demo environment and more like broadening organizational usage. The Asia Pacific and Japan figures tell a similar story: in May, Smartsheet reported 561 unique users across 334 customer plans in the region, generating nearly 60,000 MCP AI actions. That works out to meaningful repeated use rather than a one-time curiosity click.
The most important statistic is the one Smartsheet says ties AI to execution: nearly one in three AI-driven actions creates, updates, or modifies live work, rather than simply retrieving information. If accurate, that is the line between AI as a search layer and AI as an operational actor. It also explains why IT leaders will look at the same number and feel both encouraged and nervous.
Retrieval is easier to govern because it mostly concerns what the assistant can see. Write access is harder because it concerns what the assistant can change. The moment AI starts modifying live work, the enterprise conversation shifts from “Is this useful?” to “Who approved this, can we audit it, and how do we reverse it?”
Smart Assist Is the Hedge Against Assistant Fragmentation
The external connections will draw the most attention because they attach Smartsheet to the biggest names in enterprise AI. Smart Assist may be the more strategically defensive product. It gives Smartsheet users a built-in AI companion inside the platform, allowing them to ask questions or describe tasks without leaving Smartsheet.That matters because not every user wants to work through an external assistant. In many companies, the people closest to operational work are not living in a dedicated AI client all day. They are in the project plan, the intake sheet, the dashboard, or the approval workflow. For them, opening a separate AI surface may be just enough friction to make the feature irrelevant.
Smart Assist also gives Smartsheet a way to preserve the in-product experience if the external assistant market fragments further. Today’s enterprise AI stack is unstable. Google, Microsoft, OpenAI, Anthropic, and others are all trying to make their assistants the front door for work. Customers may adopt several at once, switch vendors, or restrict certain tools by department or geography.
A native assistant gives Smartsheet a fallback: even if the customer does not standardize on Claude, Gemini, Copilot, or ChatGPT for Smartsheet workflows, the platform itself can still provide AI-powered interaction. It is a hedge against interface volatility.
The feature also changes the product’s center of gravity. Smartsheet has long been a platform that mixes spreadsheet familiarity with project and workflow management. Smart Assist pushes it toward conversational work management, where users can describe intent rather than manually assemble every workflow, dashboard, or update.
That is attractive, especially for frontline and operations-heavy teams that do not have dedicated administrators for every process. But it also requires discipline. Natural language is flexible; business process is unforgiving. The value of Smart Assist will depend on how well it translates messy requests into precise, reviewable actions.
The Construction Example Shows Why This Is More Than Office Automation
Smartsheet cited DPR Construction as a customer using the MCP Server in project-heavy environments. That example is useful because construction is not a tidy knowledge-work demo. Large-scale technical construction involves many participants, changing conditions, regulatory constraints, suppliers, subcontractors, documentation, and expensive consequences when coordination fails.DPR’s Matthew Feagin described Smartsheet as the backbone for managing dynamic parts of complex projects, from data centers to healthcare facilities. His point was not that AI can write a nicer status update. It was that frontline workers can use natural language to build workflows, test ideas, and get answers faster, with less dependency on specialized configuration.
That is precisely the kind of use case where AI connected to live work data can matter. A worker trying to understand a blocker does not need a generic explanation of project management. They need to know which task is blocked, who owns the next step, what dependency is causing the delay, whether the issue appears elsewhere, and what action can be taken within the system.
If the assistant can help create a tailored Smartsheet solution for a field problem, it reduces the gap between process design and process execution. That has been one of the quiet promises of no-code and low-code platforms for years: people who understand the work should be able to shape the system without waiting for a central IT queue.
AI gives that promise a new interface. Instead of dragging columns, configuring rules, or learning formulas, the user can explain the desired workflow. The system can then propose or create the structure. Done well, that can accelerate operational adaptation.
Done badly, it can create a sprawl of half-understood automations and inconsistent process artifacts. The history of enterprise software is full of tools that empowered business users and then left IT to clean up the governance mess. Smartsheet’s governance claims will matter most in these decentralized environments.
Microsoft Copilot Makes This a Windows and Microsoft 365 Story
For WindowsForum readers, the Microsoft Copilot connection is the one to watch. Many organizations are already evaluating Copilot through Microsoft 365 licensing, security reviews, and executive mandates. If Smartsheet can make its work data available through Copilot, it potentially plugs project execution into the same assistant surface employees use for documents, email, meetings, and Teams conversations.That is powerful because Microsoft’s enterprise advantage is not merely the Copilot brand. It is the Microsoft 365 substrate: identities, files, calendars, Teams chats, meetings, SharePoint sites, compliance controls, and admin tooling. A Copilot user asking about a project may not want an answer that only knows the project sheet. They may want the broader context of the meeting where the decision was made, the document where the requirement changed, and the Smartsheet workflow where the task remains open.
Smartsheet’s presence in that assistant surface could make cross-system answers more useful. It could also make governance more complicated. The assistant experience may feel unified to the user, but the underlying permissions, retention policies, audit logs, and action boundaries span multiple systems. IT teams will need to know where the answer came from and which system authorized the action.
There is also a competitive subtext. Microsoft wants Copilot to be the orchestration layer for enterprise work. Smartsheet wants to ensure that, even if Copilot becomes the interface, Smartsheet remains the authoritative work-management layer. That is a familiar platform tension: the front-end assistant wants to absorb the workflow, while the specialized application wants to retain the workflow’s structure and data.
Customers may benefit from that tension if it produces better integrations and less lock-in. They may suffer if the result is a confusing matrix of licensing, capabilities, regional availability, and partial feature parity. Smartsheet says Microsoft Copilot and ChatGPT connections are available first to US customers, while Smart Assist, the MCP Server, and connections to Claude and Google Cloud Gemini Enterprise are available to all customers. That staggered availability is a reminder that AI rollouts are still constrained by geography, vendor readiness, and compliance posture.
For administrators, the practical question is not whether Copilot can “talk to Smartsheet.” It is whether the resulting workflow is predictable enough to support at scale.
The Security Story Begins With Permissions, But It Does Not End There
Smartsheet emphasizes that these products sit on the same governance structure, giving IT teams oversight as AI systems connect to work data across the organization. Its MCP materials also describe requests flowing through Smartsheet’s existing API infrastructure and respecting authenticated user permissions. Those are necessary foundations.They are not the whole security story. Enterprise AI introduces risks that are not identical to traditional API integrations. A normal integration usually performs known operations in known patterns. An AI assistant may choose tools dynamically based on a user’s natural-language prompt, intermediate reasoning, and the data it retrieves along the way.
That difference changes how organizations should think about control. It is not enough to ask whether a user has permission to edit a sheet. IT also needs to understand when an assistant is allowed to make edits on the user’s behalf, whether the user must explicitly approve write actions, how prompts and responses are logged, how tool calls are audited, and how sensitive data is protected when responses cross assistant boundaries.
There is also the problem of intent. A user may ask, “Clean up the overdue tasks and notify owners,” expecting a draft plan. An assistant might interpret that as permission to change statuses, create comments, or trigger alerts. The safest implementations keep humans in the loop for consequential writes and make proposed changes visible before committing them.
Smartsheet’s claim that nearly a third of AI actions involve creating, updating, or modifying live work makes these controls urgent. Read-only AI is already a governance challenge. Read-write AI is a change-management challenge.
The other concern is data minimization. Work-management platforms often contain more than task names. They can include customer details, budget hints, personnel issues, vendor disputes, legal constraints, and operational risks. Connecting that context to multiple assistants increases the need for clear policies about which AI tools are approved for which data classes.
The promise of MCP is standardized connectivity. The risk is standardized overexposure if organizations connect first and design controls later.
Open AI Plumbing Does Not Mean Open Outcomes
Smartsheet’s expansion sits inside a broader industry push toward agentic workflows. Vendors increasingly want AI systems not only to answer questions but to call tools, update records, create artifacts, and coordinate work across applications. MCP has become a favored standard because it offers a common way to expose tools to models.The optimism is understandable. Enterprises are tired of building custom integrations for every system pair. Developers are tired of brittle glue code. Business users are tired of waiting for dashboards and reports that are out of date by the time they are published. A standard protocol for AI-to-application interaction sounds like the missing connective tissue.
But open plumbing does not guarantee open outcomes. The major AI vendors still have incentives to pull users into their own interfaces, clouds, identity systems, and billing models. Support for MCP can coexist with platform lock-in elsewhere. A company may technically be able to connect multiple assistants to Smartsheet while practically finding that one vendor’s admin controls, model quality, or licensing terms make it the path of least resistance.
Smartsheet is navigating that reality by connecting to all the major assistants rather than betting solely on one. That is customer-friendly in theory. In practice, customers will still need to test each assistant’s behavior, latency, action handling, permission mapping, and administrative experience.
The same Smartsheet data may produce different user experiences depending on whether it is accessed through Claude, Gemini Enterprise, Copilot, ChatGPT, or Smart Assist. Models differ. Tool-use behavior differs. Enterprise admin surfaces differ. The protocol may standardize the doorway, but the room on the other side still changes.
That means organizations should resist treating MCP support as a checkbox. The real evaluation is operational: does the assistant retrieve the right data, preserve permissions, ask before changing important records, produce auditable actions, and behave consistently under messy human prompts?
The standard is useful because it lowers integration friction. It does not remove the need for testing, governance, and user education.
Smartsheet Is Selling the Post-Dashboard Future
One of the more interesting implications of this announcement is what it says about dashboards. Smartsheet, like many work platforms, has invested heavily in dashboards, reports, views, and structured summaries. AI does not make those obsolete, but it changes their role.A dashboard is a designed answer to a known question. An assistant is an improvised answer to a question that may not have existed yesterday. In volatile projects, that matters. Leaders still need standard reporting, but workers often need situational answers: what changed since the last review, which risk has no owner, which dependency is now blocking the launch, which tasks can be safely deferred, and what should be escalated before Friday.
That is the appeal of connecting AI directly to live work data. It lets users ask questions that were not anticipated when the dashboard was built. It also lets them move from analysis to action in the same flow.
This is why Smartsheet’s line about “where the work happens” is more than marketing. The company is trying to make the work-management layer conversational without surrendering the underlying structure that makes it useful. A spreadsheet-like grid remains valuable because it imposes order. AI becomes valuable when it helps users navigate and manipulate that order without becoming spreadsheet mechanics.
There is a danger, however, in overstating the post-dashboard future. Executives still need consistent metrics. Compliance teams still need fixed reports. Program offices still need baselines. AI-generated answers can be fluid, but fluidity is not always a virtue.
The likely future is hybrid. Dashboards will remain the canonical view for recurring governance. Assistants will become the exploratory and operational layer around them. Smartsheet is trying to serve both, which is the right posture for enterprise software that cannot afford to treat every workflow as a chat.
The Admin Burden Moves From Integration to Judgment
For IT and operations leaders, Smartsheet’s expansion reduces one kind of burden and increases another. It reduces the burden of bespoke integration. A standardized MCP Server connecting to multiple assistants is cleaner than separate custom connectors, scripts, exports, and unofficial automations.But it increases the burden of judgment. Administrators must decide which assistants can connect, which users can use them, what actions require approval, what data classes are off-limits, and how to monitor usage without smothering adoption. The technical connection is only the beginning.
This is particularly true in organizations where Smartsheet is used by business teams without heavy central oversight. Work-management platforms often spread because they are useful and flexible. That flexibility can create a long tail of departmental processes that IT only partially understands. Adding AI actions into that environment raises the possibility of faster work and faster mistakes.
The right governance model should be proportional. A team using AI to summarize project status may not need the same controls as a team using AI to update customer-facing implementation plans. A sandboxed assistant creating draft sheet structures is different from an assistant modifying production workflows. Not all AI actions carry the same risk.
Smartsheet’s adoption metrics suggest users are moving quickly. IT rarely moves at the same speed, especially in regulated or risk-sensitive organizations. That mismatch is where shadow AI tends to grow.
The best version of this rollout gives administrators enough visibility to embrace the tools rather than block them. The worst version encourages users to find unofficial routes because official governance is too slow, too opaque, or too restrictive. Smartsheet’s success will depend partly on whether its controls feel like guardrails rather than gates.
The Numbers Are Impressive, but the ROI Is Still Unproven
Three million AI actions sounds impressive. So does growth from fewer than 1,000 weekly active users to more than 9,000. But action volume is not the same as productivity, and productivity is not the same as business value.The meaningful questions are harder. Did projects finish faster? Were errors reduced? Did teams spend less time in status meetings? Did risk surface earlier? Did frontline users create better workflows without admin intervention? Did the AI actions replace busywork or merely create a new stream of outputs to review?
Smartsheet’s DPR example points toward plausible value in complex, operational environments. The company’s claim that nearly one in three actions modifies live work suggests users are not merely asking novelty questions. Still, the next phase of credibility will require more than usage counts.
This is not unique to Smartsheet. The whole enterprise AI industry is moving from adoption theater to ROI scrutiny. Executives who approved early AI pilots are now asking what changed in the business. Vendors that can connect AI activity to cycle time, error reduction, risk mitigation, or revenue-impacting workflows will have stronger stories than vendors counting prompts.
Smartsheet may be better positioned than many because work management produces measurable artifacts. Tasks have owners, dates, dependencies, statuses, and completion histories. If AI genuinely improves execution, the platform should be able to show it.
That evidence will matter. The first wave of enterprise AI was sold on possibility. The next wave will be renewed, expanded, or cut based on operational proof.
The Smartsheet AI Bet Now Has Four Front Doors
Smartsheet’s latest announcement is best read as a platform strategy rather than a feature bundle. Gemini Enterprise, Copilot, ChatGPT, Claude, and Smart Assist are different doors into the same proposition: AI becomes useful when it can safely see and act on the current state of work.- Smartsheet added Google Cloud Gemini Enterprise connectivity to its MCP Server on June 11, 2026, while also expanding toward Microsoft Copilot and ChatGPT.
- The move extends Smartsheet’s earlier Claude integration and positions the MCP Server as a neutral bridge between live work data and multiple enterprise AI assistants.
- Smart Assist gives users an in-product option for asking questions and initiating tasks without leaving the Smartsheet platform.
- Smartsheet says MCP usage has reached more than 22,000 unique users and more than 3 million AI actions since the Claude launch period.
- The most consequential adoption figure is that nearly one in three AI-driven actions reportedly creates, updates, or modifies live work rather than merely retrieving information.
- IT teams should treat the rollout as a governance project as much as an AI productivity project, because write-capable assistants require auditability, approval flows, and clear data boundaries.
The challenge is equally clear. The more useful these AI connections become, the closer they get to the systems where mistakes matter. Smartsheet’s expansion gives customers a glimpse of a more practical enterprise AI future: not a chatbot hovering beside work, but an assistant wired into the work itself. Whether that future feels empowering or reckless will depend on how well Smartsheet, Google, Microsoft, OpenAI, Anthropic, and enterprise IT teams turn AI action into something users can trust.
References
- Primary source: IT Brief Australia
Published: 2026-06-11T22:50:07.781918
Loading…
itbrief.com.au - Related coverage: smartsheet.com
Smartsheet MCP Server Achieves Exceptional Customer Adoption in First Week; 4,000 Users with 1.74 Million Total Actions Since Launch
Smartsheet brings AI to where complex work actually lives, connecting the most critical systems and data sources across the enterprisewww.smartsheet.com - Related coverage: financialcontent.com
Loading…
www.financialcontent.com - Related coverage: stackone.com
The Best MCP Gateways for Gemini Enterprise in 2026 | StackOne
The best MCP gateways for Gemini Enterprise (formerly Agentspace), compared: StackOne, Composio, Zapier, Arcade vs Google's native connectors.
www.stackone.com
- Related coverage: workmanagementhub.com
Smartsheet MCP Claude Integration: Complete Setup Guide (2026) - Work Management Hub
To set up Smartsheet MCP Claude integration, open Claude Desktop Settings → Connections, find Smartsheet, authorize via OAuth, and select which workspaces toworkmanagementhub.com - Related coverage: developers.smartsheet.com
Loading…
developers.smartsheet.com
- Related coverage: ad-hoc-news.de
Loading…
www.ad-hoc-news.de