Enterprise AI Assistants: A Field Guide to Gartner's First Magic Quadrant
Gartner published its first Magic Quadrant for Enterprise AI Assistants, a category that didn't formally exist three years ago.
On September 2, Gartner published its first Magic Quadrant for Enterprise AI Assistants, a category that didn't formally exist three years ago. It evaluates ten vendors and arrives at a moment when most organizations already have at least one of these tools in production, often without having compared alternatives.
Figure 1: Magic Quadrant for Enterprise AI Assistants, Gartner (August 2026).
It's worth reading carefully, because the quadrant sorts the market but doesn't answer the question that matters: which of these platforms fits the organization you actually have today.
Gartner defines an enterprise AI assistant as an application built on one or more generative models, designed to be the daily entry point into an employee's work. To qualify for the evaluation, a product must have its own orchestrator, out-of-the-box connectors to corporate applications, RAG-based search over internal and web data, an ecosystem of both proprietary and third-party agents, and enterprise-grade security and governance controls.
An exclusion criterion explains several notable absences. If the assistant is bundled at no cost within another suite, or billed solely on consumption, it falls outside the analysis.
The figures Gartner attaches to the market show the scale of the shift: from $17 billion in 2025 to more than $70 billion by 2030. And a more concrete projection for anyone planning the digital workplace: by 2029, knowledge workers would spend more time in these tools than in any other application.
The absence of visionaries says something about where the market stands. Vision-stage innovation is concentrated in the same players that already have scale.
Before looking at positions, it's worth answering these questions internally. They organize the decision better than the chart does.
Microsoft Copilot. The most widely deployed assistant on the market, with roughly 30 million paid licenses. Its advantage is being embedded in the applications employees already use, and leaning on Work IQ as a context layer. It fits when a well-governed Microsoft 365 environment is already in place. Worth checking: it requires a qualifying M365 plan, its licensing scheme is the most complex of the group, and Microsoft is the only Leader without its own frontier model behind its assistant.
Google Gemini Enterprise. Integrates chips, infrastructure, models, and agents into a single stack. Its pricing scheme is among the simplest in the market, and agent governance controls come at no extra cost, which matters when the goal is governing agents at scale. Worth checking: the experience is more complete with Google Workspace, and its computer-use capabilities are still maturing.
OpenAI ChatGPT Enterprise. Combines ChatGPT's mass familiarity with product depth in research, analysis, and agents, without forcing a commitment to a productivity suite. Worth checking: its connector strategy is focused on productivity tools and has less coverage over corporate applications like SAP, Workday, or ServiceNow. There are also governance gaps, such as DLP and runtime controls, usually closed by integrating third-party tools.
Anthropic Claude Enterprise. The vendor Gartner singles out as the category's innovation leader. It created MCP, now the de facto standard for connecting agents to tools and data, and its long-running agentic capabilities are among the most advanced in the group. Worth checking, and this weighs heavily in the region: its hosting and data residency are limited to the United States. It also bills per-interaction consumption on top of the per-user license, which complicates budget predictability.
Amazon Quick. Unusually broad functional scope, combining search, research, automation, BI, and no-code app building on a single platform, plus access to models from multiple providers via Bedrock. Fits well where there's already investment in AWS. Worth checking: it's a recent entrant, and its collaboration and co-authoring capabilities are less developed.
Glean. Its personal and corporate knowledge graph works as a context layer, even underneath another assistant that's already been adopted. It offers the most flexible deployment options in the group, including managed private cloud on the customer's own infrastructure and on-premise. Worth checking: it comes from the enterprise search world and trails the leaders in agentic execution. Several of its corporate connectors are read-only.
The four Niche Players. ClickUp Brain² makes sense with an already-consolidated ClickUp deployment. ServiceNow EmployeeWorks works as an entry point for IT and HR workflows, with narrower reach for general knowledge work. Superhuman Go integrates inline over the browser and existing applications, reducing context switching. ZoomMate leverages meeting, call, and chat context, though today it operates only in English and out of U.S. data centers.
Reading Gartner's vendor-by-vendor descriptions surfaces a pattern the quadrant itself doesn't show. Presence in Latin America is limited across much of the group. Anthropic, Glean, OpenAI, and Zoom have customer bases concentrated in North America and Europe, with narrower reach in the region. Microsoft shows up with the broadest global and sovereign coverage in the market, and Google with significant global reach, including in regulated industries.
For a financial institution, an insurer, or a company with data residency requirements in LATAM, this isn't a nuance. Gartner states it plainly: an assistant without the required deployment footprint can't be implemented, no matter how well it scores on product.
The most practical data point in the report is that most organizations are already deploying two or more of these tools. Gartner's recommendation follows the same logic: pilot at least two, with the user groups that stand to benefit most, before standardizing. The risk to manage is functional overlap and a governance perimeter that becomes harder to administer.
And a warning about the order of operations. Gartner suggests settling the governance, security, and data-access framework first, and only then rolling out broadly. Who can create agents, which platforms they can operate on, and what information they can reach are decisions best agreed upon across the workplace, security, compliance, and identity teams beforehand.
Choosing the platform is the visible part of the decision. What takes time comes after: sorting out permissions across corporate repositories, defining the agent governance framework, connecting the assistant to the applications where the process actually happens, and sustaining adoption with metrics.
At EDSA we work on that layer, with Microsoft, Google, and Anthropic platforms among others, and with the criterion of choosing the tool based on the business objective rather than the other way around.
Source: Gartner, Magic Quadrant for Enterprise AI Assistants, September 2, 2026.
Access the full report:Gartner® Magic Quadrant™ for Enterprise AI Assistants
Write to us at talk@edsa.com