Somewhere in your company right now, a person who was hired to do difficult, valuable work is instead typing an answer they have typed before. Where is the vacation policy. Who approves the invoice. What is the Wi-Fi password. Multiply that by a whole team and you get the quiet tax every growing company pays: the same questions, answered again and again, mostly in Slack. Tettra was built to collect that tax and hand it back.
The Boston company, founded in 2015 by Andy Cook and Nelson Joyce, sells an internal knowledge base - a company wiki, in older language - wired directly into the chat tools where people already spend their day. Instead of asking teammates to visit yet another app, Tettra lives inside Slack and Microsoft Teams. Ask a question, and an AI bot named Kai looks through the company's documentation and answers on the spot. If it doesn't know, it says so, and quietly nudges the human who does.
01 / The problemThe company that forgets what it knows
Every organization accumulates knowledge and then loses track of it. The onboarding doc lives in someone's Google Drive. The refund process is a Slack message from eight months ago. The one engineer who understands the billing system is on vacation. Cook and Joyce started Tettra because the tools meant to fix this were, in their telling, part of the problem: too complicated to keep up, quick to fill with outdated pages, and stranded outside the daily flow of work. A wiki nobody updates is not a knowledge base. It is a museum.
Their fix rested on a small, stubborn idea. The hard part of knowledge management is not writing things down. It is keeping what you wrote down true. So Tettra built the maintenance into the product: reminders to verify pages, flags for content that has gone stale, suggested edits that route to a subject matter expert for approval. The wiki, in effect, reminds itself to stay honest.
There is a human cost to the alternative, and it hides in plain sight. When knowledge lives only in people's heads, the most capable person on a team becomes the bottleneck for everyone else. Their day fills with interruptions - a tap on the shoulder, a direct message, a meeting that could have been a paragraph. New hires ramp slowly because the answers they need are scattered across chat history and half-finished docs. Tettra's argument is that a company should be able to answer its own routine questions without borrowing an expert's afternoon, and that the way to get there is not discipline but design.
02 / The pivotFollow the traction, not the plan
Tettra did not arrive fully formed. The founders started with a broader idea and, within months, noticed something: people were using it as a Slack tool. Rather than argue with the data, they doubled down on where the pull was. It is the least glamorous lesson in startups and one of the most reliable - build where users already are, not where you wish they would go. Cook has told the story on the Indie Hackers podcast, and it reads like a case study in patience: iterate until you find the version people actually pay for, then stop being clever.
The numbers that resulted are unfashionable in the best way. By 2024 Tettra reported roughly $3.2 million in annual recurring revenue and about 548 customers, run by a team of around a dozen people. It raised only about $1.82 million in total, with Y Combinator among the early backers. There was no mega-round, no blitz-scaling. There was, notably, negative churn - customers on the whole spent more the longer they stayed, expansion outpacing the ones who left.
That last figure is the one investors and operators tend to circle. Negative churn on a knowledge base is unusual. Most software people can quit without much pain - cancel the trial, export the data, move on. A knowledge base is stickier by nature: once a team pours its processes into one, the switching cost is measured in institutional memory, not dollars. Tettra leaned into that stickiness by making the product more useful the more a team fed it. The wiki that started as a place to dump docs became, over years, the place people checked first.
03 / The productMeet Kai, the bot that admits it's stuck
The centerpiece today is Kai, Tettra's AI assistant. Technically it is retrieval-augmented generation - the bot searches the company's own verified pages and composes an answer from them, rather than free-associating from the open internet. The design choice worth noting is what Kai is not allowed to read. Stale pages and private pages are kept out of its index. That sounds like a limitation. It is actually the whole point: a knowledge bot that confidently quotes a document which stopped being true in 2022 is worse than no bot at all.
The rest of the product is deliberately plain. A simple editor for writing pages. Imports from Google Docs, Notion, and PDFs so teams don't start from zero. A question-and-answer system that captures what people ask, assigns it to an expert, and files the answer back into the knowledge base. Integrations with Slack, Microsoft Teams, Google Workspace, GitHub, and Zapier. Nothing here is trying to be a document workspace or a second brain. It is trying to answer questions and keep the answers current.
That handoff between machine and human is more clever than it first appears. Most chatbots fail in one of two ways: they guess, or they dead-end. Kai does neither. An unanswered question is not a failure state - it is a prompt to capture knowledge that did not exist yet. Someone answers it once, the answer is filed, and the next person who asks gets it instantly. Over time the gaps close on their own, which means the product's blind spots become its to-do list. It is a small loop, but it is the loop that separates a knowledge base that grows from one that rots.
04 / The marketThe case for boring software
Tettra plays in a crowded field. Atlassian's Confluence is the enterprise incumbent. Notion is the flexible everything-app. Guru, Slite, Slab, Nuclino, and Document360 all crowd the middle. Tettra's answer to the crowd was not more features. It was fewer. It sold simplicity to teams that found Confluence bloated and Notion a maze - and on G2, users consistently rated it around 4.6 to 4.7 stars, scoring it notably higher than Confluence on ease of use.
05 / The modelTen seats and up
The business is straightforward B2B software as a service, billed per seat with a ten-user minimum. That floor is a quiet strategy decision: Tettra is for teams, not individuals, so it prices for teams from the first invoice. The typical customer runs operations, HR, support, or an agency - the parts of a company where the same procedural questions surface daily and where a single source of truth pays for itself fastest. The math is easy to run: if one senior person spends even a few hours a week re-explaining the same processes, a few dollars a seat is a rounding error against the salary being interrupted.
| Plan | Price | Best for |
|---|---|---|
| Basic | ~$5 | Up to 100 users |
| Scaling | ~$10 | Up to 250 users |
| Professional | ~$7,200/yr | Unlimited users, added controls |
06 / The exitA bootstrapper's ending
In October 2023, Tettra was acquired by GSoft, a Canadian software company that had rebranded to Workleap earlier that year after a CA$125 million capital investment. Workleap is assembling a suite of tools for the modern workplace - employee experience, engagement, productivity - and a lean, focused knowledge base with real revenue and negative churn was a natural piece to slot in. For a company that raised under $2 million, being bought by a well-capitalized acquirer is the kind of ending most founders would take. Andy Cook stepped back from the CEO seat through the integration.
What Tettra leaves behind is less a technology than a discipline. Put the knowledge where the questions get asked. Make the tool responsible for staying accurate, not the humans. And when the AI doesn't know, let it say so and point to someone who does. Companies named Techstars, Fortnox, and SmartBug Media used it to stop being their own help desks. That is the entire promise, and after eight years it still fits in a sentence: stop answering the same question twice.