Cognite makes petabytes of complex industrial data usable. Its platform serves highly technical developers, solution architects, data scientists, and even end users on the shop floor.
Every one of Cognite's audiences needs documentation written for them, across multiple products and over a dozen languages. AI agent readership adds yet another layer of complexity, and the team's current docs setup simply couldn't keep up.
Cognite consolidated its product docs, API references, and internal knowledge onto Mintlify, then built an AI-powered content system on top that lets its lean team serve high-quality knowledge to employees, customers, and agents.
80%+
of contributor drafts publishable, up from 40%
60%+
of docs traffic now comes from agents
$180K
annual savings from support ticket deflection
Challenge
Cognite's docs were spread across disparate systems and publishing pipelines run by the engineering team. As the company grew, having to manage content across this sprawling ecosystem and provide extensive edits before every push was becoming increasingly difficult.
Solution
Cognite consolidated its product docs, API references, and internal knowledge onto Mintlify, then built an AI-powered content system on top. Now the entire company can contribute consistent, high-quality content at scale that's accessible for both human readers and AI agents.
A sprawling estate that couldn't scale
Documentation lived across two self-managed Docusaurus repositories while API references sat in a separate, brittle Redoc.ly instance: three different sites and URLs that created a confusing, disconnected user experience. The publishing pipelines behind it all were operated by internal engineers and required constant maintenance. As the company grew and the number of technical writers stayed the same, none of it scaled.
Our company was growing, our documentation needs were increasing, and it just wasn't working. We needed something that would let us scale, and we knew we could save money by consolidating the systems we'd been managing separately.
– Lise Hauger
Director of Content Operations at Cognite
After evaluating options including GitBook and other competitors, Cognite chose Mintlify because it checked every box, including support for internal content workflows, AI-native artifacts, and AI-assisted authoring. The team ran a test migration before committing, confirmed Mintlify could do what they needed, and the consolidation began.
A content engine, not just a docs site
Because Mintlify treats documentation as code (Markdown in a GitHub repository that AI tools can read and edit), Cognite could layer its own content system on top. They call it CogDocs, a framework powered by custom skills that work with the Mintlify agent. These skills span content planning, writing, scoring, metadata, terminology, and a style guide, encoding everything Cognite knows about good documentation into instructions the agent follows as it drafts.
We built our information architecture and style guides into skills. It means we can help engineers and everyone else produce better content, because the documentation team doesn't scale to write everything ourselves. It's about democratizing who can produce content while maintaining quality.
– Arne Kvadsheim
Information Architect at Cognite
Cognite writers and engineers alike draft in the shared GitHub repo, guided by the skills, while the documentation team acts as code owners, so nothing publishes without review. The scoring skill grades each draft from 0 to 100 on dimensions like SEO, metadata, and style, then hands it back to be revised and re-scored. Cognite gives that skill to contributors directly, so they can self-correct and learn instead of waiting for a writer.
The result is a measurable jump in quality. Drafts that used to arrive at roughly 40% publishable now come in at 80%, and the number keeps climbing as people embrace the system.
We used to do a lot of editing and rewriting. Thanks to Mintlify, we now spend 50% less time on docs cleanup and can allocate those resources to higher-value work, like getting the structure right, making sure content connects, and keeping everything consistent.
– Lise Hauger
Director of Content Operations at Cognite
One platform, two audiences
The Cognite team knows documentation isn't something just people read anymore. As Kvadsheim puts it, "It's kind of like code for agents these days." That's why the team invests in getting structure, components, and cross-links right rather than just polishing prose, since those are the signals agents rely on to find and use the right information.
From a single Markdown source, Mintlify generates the machine-readable artifacts agents expect, including SKILL.md files and an MCP server, so one repository serves a clean, searchable website to humans and structured context to agents. The investment is paying off: AI agents now make up more than 60% of traffic to Cognite's documentation overall. The number jumps to 70% on API references, where agents and MCP clients do most of the reading.
This shift is already changing where Cognite's knowledge gets consumed. Increasingly, its content is read not on its own site but inside ChatGPT, Claude, and other tools.
Before, if you made it to Google, you were good. Now you have to be everywhere, and you don't always know where people are seeing your content, which makes getting the quality right even more important.
– Arne Kvadsheim
Information Architect at Cognite
One source of truth for internal knowledge
In addition to customer-facing docs, Cognite now runs parts of its internal knowledge on Mintlify too. The delivery team's standard operating procedures live on a dedicated Mintlify site, giving employees the same reliable, searchable experience customers get.
Cognite also uses the internal portal as a way to get other teams comfortable working on the Mintlify platform. Once they are, their contributions also help inform external documentation.
Internal knowledge management is just as important as external. Mintlify has enabled us to have one source of truth for humans and agents, both inside the company and out.
– Lise Hauger
Director of Content Operations at Cognite
Analytics that exceeded expectations
Since choosing Mintlify, the Cognite team has come to rely heavily on its analytics. "I expected we'd get good data, that was one of the things we were interested in," reveals Hauger. "But it's been better than I could have imagined."
The team watches where readers come from, where they go, and where they get stuck, then points the Mintlify Agent at those exact gaps to fill them. And the insight doesn't stop at the docs. Product teams are starting to leverage these analytics to better understand how users engage and how features can be improved.
The Mintlify assistant plays a large role in ensuring both internal and external users get what they need from the docs and highlighting when they don't. The early business signal is promising: the team currently saves $180K a year from support ticket deflections.
Powering the full knowledge lifecycle
"We use Mintlify across the full lifecycle of what we do, from writing and publishing, to insights, to improving, and even helping other people contribute," says Hauger. But the team isn't done yet. It recently began piloting Mintlify's new Autopilot feature across a handful of repos.
Autopilot is an always-on agent that watches connected repositories, drafts documentation updates when code changes, and opens a pull request for review. It's a natural fit for how Cognite already works, since it submits every change for review rather than publishing on its own. "It will help us identify the things that fall through the cracks," says Kvadsheim.
In addition, the team's legacy Redoc.ly API reference will be switched off once the remaining links and redirects are mapped over, likely within the next six months. Last but not least, Cognite is in early conversations about bringing its support knowledge base onto Mintlify as well.
A foundation for rapid growth
What began as a migration is now an AI-powered content engine that lets a lean team produce consistent, high-quality knowledge for a growing company. It serves humans and agents alike, in fourteen languages, across product docs and internal knowledge, from a single source of truth.
Migrating to Mintlify, along with our updated style guidelines and workflows, established a foundation for quality and scalability. Because that infrastructure is in place, we're well equipped to support Cognite's rapid growth and capitalize on new automation opportunities.
– Lise Hauger
Director of Content Operations at Cognite
