Nobody Reads the Company Wiki. Now AI Reads It For You

Ask anyone who has worked at a company with more than fifty people to describe the internal wiki, and you get the same tired look. It exists. It is out of date. The one page you need was last edited by someone who left in 2021. So you do what everyone does - you ping a colleague on Slack and ask. The knowledge was written down. Nobody read it.

That gap, between knowledge stored and knowledge used, is the entire business these eight companies are fighting over. Atlassian's Confluence, Notion, Nuclino, Zendesk, eGain, Paligo, Glean and Knowmax do not look like direct rivals. One is a wiki, one is a help center, one is a technical-docs engine, one is an AI search box. For years they sold into separate budgets and rarely met in a bake-off. In 2025 that stopped being true.

The old mapFour businesses that never used to compete

Start with the honest version of the market, before the AI marketing swallowed the labels. There are really four jobs here, and each of these products was built to do one of them well.

A support leader buying Zendesk in 2019 was not evaluating Paligo. An engineering team standardizing on Confluence was not comparing it to eGain. These were four aisles in four different stores. What connected them was only the underlying promise: put knowledge somewhere central, and people will find it. The dirty secret was that people mostly did not.

What changedRetrieval flipped the product from filing cabinet to answer

The technical unlock has an ugly name and a simple effect. Retrieval-augmented generation, or RAG, means an AI model first pulls the relevant documents from your company, then writes a grounded, cited answer from them. The user never opens the document. They ask a question and get a sentence back, with a link to the source if they want to check.

Once that works, the shape of every product on this list changes. The wiki stops being a destination and becomes raw material. The help center stops being a page a customer skims and becomes the source a bot quotes. The value moves from the place you store knowledge to the layer that hands it back. The industry even has a slogan for it now.

Notion's chief executive, Ivan Zhao, has publicly framed RAG as the future of knowledge management, which is a notable thing to say when your company was built on selling people a better document. Atlassian bolted an AI layer called Rovo onto Confluence and made it generally available in October 2024, wiring in more than 50 third-party apps so the assistant can answer across tools, not just inside the wiki. Twenty years after launching Confluence as the original wiki, Atlassian effectively admitted people would rather ask than browse.

The striking part is how synchronized the shift was. Generative AI hit the whole category in a single wave around 2023 - Notion AI, eGain's AssistGPT, Nuclino's Sidekick, and RAG-style question-answering all shipped in the same window. Nobody wanted to be the vendor still selling a static repository while a competitor demoed a chat box that answered in plain English. By 2025 the demos had turned into agents. Notion 3.0 launched AI agents in September 2025 and added custom agents in a 3.3 release the following February. Glean shipped its own agents in early 2025. The unit of the product quietly moved again, from a search box that answers to a worker that acts.

Follow the moneyA boring category, suddenly priced like AI

The clearest evidence that something shifted is the funding. Glean, founded in 2019 by former Google engineer Arvind Jain to build search over a company's own data, raised a $150M round in June 2025 at a $7.2B valuation. Its revenue went from roughly $100M to $200M in annual recurring revenue in about nine months. That is not the growth curve of a filing cabinet.

Notion, founded in 2013, crossed 100 million users and opened an employee share sale near an $11B valuation, with the company crediting AI for the acceleration. eGain, the veteran of the group at nearly thirty years old, is a public company doing roughly $88M in annual revenue - a reminder that the newest AI category often runs on the oldest plumbing. Here is how the headline valuations and revenue markers stack up.

The tellWhen pricing changes, the market has already moved

If you want to know where a software category is heading, watch the pricing model, not the press release. Zendesk, taken private in 2022 for about $10.2B, spent 2024 and 2025 buying its way into AI - Ultimate, Klaus, and in December 2025 the RAG search company Unleash. Then at its Relate 2026 event it did the thing that actually matters: it started charging per resolved answer instead of per seat.

That is a bet with teeth. Per-seat pricing rewards you for having more human agents logged in. Per-resolution pricing rewards the software for closing the ticket without a human at all. A knowledge base vendor that charges by the answer is telling you, in the only language a finance team respects, that the document was never the point. The resolution was.

The independentsSmall, structured, and stubbornly useful

Not everyone is chasing an $11B valuation, and the exceptions are instructive. Nuclino, run out of Munich, is a lightweight team wiki that is profitable, founder-owned, and has taken no venture money. It competes against companies worth thousands of times more by being fast and simple - the thing that loads instantly while the enterprise suite spins. In a category obsessed with adding AI, "it opens quickly and gets out of the way" remains a real strategy.

Paligo is the specialist most people have never heard of. It is a component content management system, which is a mouthful for structured, single-source technical documentation - the manuals, the compliance docs, the content that gets reused across a dozen outputs and cannot afford to be wrong. It raised $29M in 2023, led by GRO Capital. In an AI moment defined by plausible-sounding answers, there is quiet demand for documentation that is allowed to be exactly one thing. A regulator does not want a creative summary of the safety procedure. That distinction - structured facts versus generated prose - is likely to be one of the fault lines that survives the AI wave, because some knowledge has to be retrieved exactly, not paraphrased.

Knowmax and eGain sit on the customer-experience side, building knowledge for contact centers - decision trees, agent-assist, self-service. eGain has done this since 1997 and now wraps it in generative AI branded Max AI and AssistGPT. Knowmax, run from Gurugram in India, does much the same for support teams with visual how-to guides and cognitive decision trees. The pattern repeats: an established knowledge product, a language model welded on top, a new category name on the invoice. It is worth remembering that this is not the first time knowledge management was declared the next big thing. The phrase has been cycling through enterprise software since the late 1990s. What is different now is that the technology finally does the one thing the category always promised and never delivered - it reads the pile for you.

Why it mattersWhoever owns the answer owns the customer

In November 2025 Gartner made the reshuffle official, defining two fresh markets - Generative AI Knowledge Management Apps and Enterprise AI Search - and naming Glean an emerging leader in the first. When the analysts rename the shelf, the store is being rebuilt. Gartner also projects AI search and assistants will be embedded in about 60% of enterprise apps by 2028, up from roughly 20%.

The strategic prize is worth spelling out, because it explains the money. If AI reads the docs and hands users the answer, the open question is who owns that answer layer. Own it for a support team and you own the customer relationship. Own it for an entire company and you become the default front door to everything anyone knows. That is why a wiki company, a help-desk company, and a search company all suddenly look like they are building the same product. They are.

For anyone actually buying this software, the practical read is calmer than the valuations suggest. Small teams still reach for Notion or Nuclino. Large engineering orgs live in Confluence. Support teams pick Zendesk, eGain or Knowmax. Regulated technical writing needs Paligo. Company-wide AI search points at Glean. Most organizations end up running several at once and quietly wishing they did not. The convergence is real, but it is happening at the answer layer on top, not in a single winner underneath. The wiki is not dead. It has just been demoted to something the AI reads so you do not have to.