Published in 2026. This article examines developments from 2024 to 2026.
Introduction
Enterprise knowledge management — helping employees find and use information scattered across documents, tickets, chat history, and internal wikis — has become one of the clearest practical use cases for Generative AI inside large organizations, precisely because the problem it solves (too much unsearchable internal information) predates generative AI by decades and does not depend on any single model breakthrough to matter. This article looks at how a handful of companies are approaching the problem differently, and where each fits in an enterprise’s broader AI stack. Unlike categories built entirely around brand-new generative AI capability, knowledge management is a case where AI has mostly improved a problem enterprises already knew they had — findability, consistency, and speed of internal information retrieval — rather than introducing an entirely novel workflow employees had never encountered before in some form, which is part of why adoption in this category has tended to move faster than in categories requiring employees to learn a genuinely new way of working.
Permissions-aware Enterprise Search
Glean, founded by former Google search engineer Arvind Jain in 2019, is built specifically around the idea that enterprise search has to respect the same access permissions that already govern each underlying system — a document a user cannot open in Google Drive should not appear as a generated answer in a company-wide AI assistant either. Glean connects to an organization’s existing SaaS tools and generates answers grounded in that permissions-aware retrieval layer, which the company positions as its core differentiator against generic Retrieval-Augmented Generation implementations that enterprises might otherwise build in-house without the same access-control rigor. That distinction matters more than it might sound: a poorly permissioned internal search tool can accidentally surface confidential information to employees who should not see it, which is precisely the failure mode Glean’s architecture is designed to prevent by design rather than by after-the-fact filtering.
Governed generative content at brand scale
Writer, founded by May Habib in 2020, approaches the knowledge management problem from a different angle: rather than focusing primarily on search and retrieval, Writer builds its own Palmyra language models and wraps them in governance tooling, style guides, and deployment controls aimed at making sure AI-generated marketing and operational content stays on-brand and compliant across a large organization. For companies whose knowledge management challenge is less “can employees find the right document” and more “can multiple teams generate consistent, compliant content without a bottlenecked review process,” Writer’s full-stack approach — owning both the model and the governance layer — is a distinct alternative to retrieval-focused tools. Writer’s decision to build and control its own underlying models, rather than relying entirely on third-party model providers, also gives it more direct control over how those governance and style constraints are enforced at the model level rather than only through a wrapper layer on top of someone else’s model.
Knowledge management inside a broader data platform
Dataiku approaches knowledge management as one output of a broader governed data and machine-learning platform rather than as a standalone product category. Because Dataiku already sits on top of an enterprise’s data infrastructure for MLOps and analytics, its generative AI workflows can connect Large Language Models to internal data with the same governance and audit controls the company has built for traditional machine-learning models over more than a decade — an approach that appeals most to enterprises that already use Dataiku for data science and want generative AI to inherit the same guardrails rather than introducing a separate, ungoverned tool. Florian Douetteau has described this continuity as central to Dataiku’s pitch: governance built for one generation of machine-learning models should not need to be rebuilt from scratch simply because the newest models happen to be generative rather than predictive.
Conversational assistants and the Moveworks-ServiceNow combination
Moveworks took a more conversational, assistant-first approach to knowledge management, letting employees resolve IT, HR, and other workplace requests through natural-language search and automated actions across enterprise apps via its Reasoning Engine. ServiceNow’s acquisition of Moveworks, completed in December 2025, combined that conversational search and multi-step Automation directly with ServiceNow’s own workflow platform, and by February 2026 the companies had launched ServiceNow EmployeeWorks specifically to bring Moveworks’ knowledge and search capabilities into a unified portal serving what ServiceNow described as close to 200 million employees across its installed base. Notably, ServiceNow chose to keep Moveworks available as a standalone product rather than retiring it into the broader platform, which suggests knowledge-management-specific buying decisions remain distinct enough from full ITSM platform decisions that customers still want the option to adopt one without the other.
How to think about fit, not just features
The four approaches above are not simply competing on feature checklists; they solve somewhat different problems. Glean is the strongest fit for organizations whose core challenge is fragmented search across many disconnected SaaS tools. Writer fits organizations whose primary pain point is generating consistent, on-brand content at scale rather than searching existing content. Dataiku fits organizations that already run governed machine-learning pipelines and want generative AI to plug into that existing discipline. Moveworks, now inside ServiceNow, fits organizations that want employee-facing conversational support tied directly to IT service management workflows. Enterprises evaluating this category should therefore start from their dominant knowledge-management pain point rather than from a generic “which AI assistant is best” comparison, since the four platforms described here were built to answer meaningfully different questions even though they are often discussed together under a single “enterprise AI” umbrella.
The procurement pattern behind knowledge management deals
One pattern worth noting from this period is how these deals actually got signed inside large organizations. Knowledge-management tools tend not to be bought by a single centralized “AI team” the way many outside observers assume; instead, they are frequently championed by an IT service management leader (in Moveworks’ and now ServiceNow’s case), a chief data officer or analytics leader (in Dataiku’s case), a search or knowledge-operations owner (in Glean’s case), or a marketing operations leader (in Writer’s case). That buyer diversity is itself a reason the category has not consolidated around a single dominant platform the way some other enterprise software categories have: each tool answers to a different budget holder with a different success metric, whether that is ticket deflection, model governance coverage, search satisfaction, or content production velocity.
Why the ServiceNow-Moveworks deal is a useful bellwether
The ServiceNow-Moveworks acquisition deserves a second look because of what ServiceNow chose to do after closing it. Rather than immediately retiring the Moveworks brand or forcing existing Moveworks customers onto a full ServiceNow contract, the companies kept Moveworks available as a standalone product while also launching ServiceNow EmployeeWorks as an integrated offering. That dual-track approach suggests ServiceNow’s own diligence concluded that a meaningful share of Moveworks’ value came from customers who wanted a best-of-breed conversational AI layer without committing to the broader ITSM platform, and that forcing a bundled purchase risked losing exactly the customers the acquisition was meant to capture. Other large platform vendors evaluating similar acquisitions in adjacent categories are likely watching how this dual-track integration performs before committing to a similar structure themselves.
Conclusion
Enterprise AI knowledge management matured from 2024 to 2026 into a genuinely differentiated set of approaches rather than a single commoditized category, and the ServiceNow-Moveworks acquisition stands out as the period’s clearest signal that large workflow platforms consider this capability important enough to buy outright, while still preserving it as an independently purchasable product. Readers assessing options in this space should look closely at Glean, Writer, Dataiku, and Moveworks against their own dominant use case and internal budget owner rather than assuming one platform serves every knowledge-management need equally well.