A data and AI control plane disguised as a catalog. The company every Fortune 500 CDO seems to have on speed-dial - and most of their analysts have never heard of.
Walk into a data team meeting at any large company today and there is a 50/50 chance Atlan is the tab someone has open. The other tab is usually Slack, where someone is asking what a column means.
Atlan is a metadata workspace. That sentence will make your eyes glaze over, which is roughly the experience the company has been fighting against since 2019. What Atlan actually is, in the way the people who use it describe it, is the place you go to figure out whether the number on the dashboard is real, who owns it, and whether the AI agent your boss just deployed is allowed to touch it.
The company is headquartered in San Francisco, with most of its engineering muscle in New Delhi. About 460 people work there. Customers include Cisco, Nasdaq, HubSpot, Unilever, Ralph Lauren, Plaid, FOX, News Corp, and Autodesk. In May 2024, the company raised $105 million in Series C funding at a $750 million valuation, led by GIC and Meritech Capital, with participation from Salesforce Ventures and Peak XV (formerly Sequoia Capital India).
Pull quote · Untitled, from a customer onboarding call · 2024
Pick a mid-sized enterprise. They likely have Snowflake. They likely have Databricks. They have dbt, Fivetran, Looker or Tableau, and a quietly growing number of LLM-powered tools that need to be pointed at data the company half-trusts. The "modern data stack" is, charitably, a stack. Less charitably, it is a pile.
The tools are good. The connections between them are not. Lineage breaks. Definitions drift. A column called revenue means one thing to finance and something almost-but-not-quite-the-same to sales. When the CEO asks why two dashboards disagree, the answer is usually a Slack thread that goes for nine hours.
Atlan's founders had a sharper version of this complaint. They had built a data team at SocialCops, their previous company, and watched their analysts spend something like half their time hunting for context that should have been a click away. They tried buying the existing catalogs. They were, in the founders' telling, joyless, expensive, and built for a 2010 idea of how data teams work.
Paraphrased from Prukalpa Sankar · Peak XV podcast · 2023
Prukalpa Sankar and Varun Banka founded SocialCops in 2013, fresh out of college, with a stubborn idea that data could be used to do socially useful things. The company helped build India's national data platform and partnered with the World Bank. It also forced them to live, daily, with the gap between how data teams work and how the rest of software has moved on.
The bet they made in 2019, when they spun Atlan out as a separate company: data work is now a team sport, and the team sport needs its own workspace. Engineers got GitHub. Designers got Figma. Product managers got Notion. Data people got, somehow, a series of disconnected query editors and a Confluence page nobody updated.
Atlan would be opinionated about collaboration. Documentation would live where the work happened, not in a separate wiki. Metadata - the boring, structural information about data - would be active, meaning it would flow back into Slack, Jira, BI tools, and increasingly, the prompts feeding LLMs.
Internal language · used in pitch decks, recruiter calls, and most blog headers
Atlan looks like a search bar with opinions. Type a term - say, customer_ltv - and the platform returns every table, dashboard, model, and column that uses it. It also returns the people who own those assets, the lineage between them, the tests that have passed or failed, and the policies that govern who can see what.
Underneath, it is a metadata graph. Connectors pull metadata from Snowflake, Databricks, BigQuery, dbt, Looker, Tableau, Fivetran, Power BI, and a long tail of others. The graph is two-way: tags applied in Atlan flow back into the source systems. Definitions written here show up as tooltips in BI dashboards. Policies enforced here propagate to the warehouse.
The cynical answer: data catalogs are not new. The honest answer: large-language-model deployments inside enterprises have made metadata urgent. An agent that queries the warehouse on behalf of an executive needs to know which columns are sensitive, which definitions are authoritative, and which datasets are stale. That is exactly what Atlan stores. The category is suddenly load-bearing.
A paraphrase. Several different analysts have said something approximately this in 2024.
Approximate amounts, USD millions. The Series C is the one everyone wrote about.
SaaS companies love to list logos. The logos are usually meaningless. Atlan's are a little less meaningless because the buyers are not marketing departments - they are Chief Data Officers at organisations that fail audits when their data is wrong.
Cisco runs Atlan. So do Nasdaq, HubSpot, Unilever, Ralph Lauren, Plaid, News Corp, FOX, NextGen Healthcare, and Autodesk. The buyers tend to be data leaders who have already tried at least one of the legacy catalogs and decided the second purchase needed to be something the analysts would actually open.
The analyst houses noticed. In late 2024 Atlan was named a Leader in Enterprise Data Catalogs by a major global research firm, which is the unsexy version of being placed on a magazine cover. In this category, it counts.
A quote that gets repeated, in various phrasings, in roughly half the customer case studies.
The mission, in the founders' words, is to be a home for the humans of data. The phrase is a small piece of stubbornness. It refuses to call the user "the customer," refuses to describe data work as a pipeline problem, and quietly insists that the bottleneck in most data teams is not the technology but the social fabric around it.
That bias shows up in the product. Atlan invests heavily in features that look like Notion or Linear - inline commenting, mentions, embedded glossaries, shareable links that resolve in Slack. The technical bits, lineage and policy and connectors, are the table stakes. The collaboration is the moat.
It also shows up in how the company hires. Atlan is remote-first, distributed across the US, India, and a long tail of countries. The internal writing culture is unusually heavy, a habit borrowed from years of running a globally distributed team at SocialCops. Memos beat meetings. The company runs its own internal version of Atlan to manage Atlan, which is the kind of thing only people who actually believe in their product do.
A data catalog used to be a compliance line item. Today it is, depending on who is selling, an AI readiness platform, a data product hub, or a context layer for agents. The naming is marketing. The underlying point is real: as more decisions are delegated to software, the metadata describing what the software is allowed to do, with which data, on whose authority, becomes the thing that determines whether any of it works.
Collibra, Alation, Informatica and Microsoft Purview are all running at the same problem from different angles. So are a generation of leaner challengers - Secoda, Select Star, data.world. The category is crowded and getting more so. Atlan's argument is that it spent the years between 2019 and 2024 quietly building a product that data teams actually liked using, and that liking-it turns out to be a defensible technical position when budgets tighten.
The closing argument. Make of it what you will.
Walk back into that data team meeting from the top of this story. The dashboard still has a number on it. Someone still asks what it means. The difference, if Atlan has done its job, is that the answer is one click away, the owner is tagged, the lineage is visible, and the AI agent watching the conversation already knows whether it is allowed to act on it.
That is the bet. It is not a flashy bet. It is the kind of bet you make when you think the next decade of software is going to be quietly decided by who controls the metadata. Atlan has $200 million in the bank and a 460-person team that believes that very specific thing. Worth keeping the tab open.