Introducing Coda Brain: Your team’s favorite know-it-all

Coda is launching a turnkey AI platform that helps teams spend less time hunting for information—and more time putting data to work.

Blog > About Coda · 6 min read
Earlier this year, we announced a large partnership with Snowflake, the world’s leading AI Data Cloud company, with a focus on turning data into action across the enterprise. We also hinted at something more coming—something big. Today, Snowflake's CEO, Sridhar Ramaswamy, announced Coda Brain during his keynote at Snowflake Summit, saying:
With Coda Brain, any business user can ask questions of their enterprise structured or unstructured data, like documents, and they can get results that an enterprise user can trust.
Sridhar Ramaswamy
CEO, Snowflake

Ready for part 2 of the Coda + Snowflake partnership? Introducing Coda Brain.

Coda Brain, a turnkey AI platform, powered by Snowflake, that helps teams spend less time hunting for information and more time putting it to work. I like to call it “your team’s favorite know-it-all.”

A bit of background: Enterprise AI ≠ Consumer AI.

In the past few years, we’ve all fallen in love with instant access to the world’s knowledge with help from consumer AI tools like ChatGPT, or “the world’s know-it-all.” In 2024, everybody expects to bring the same experience to the enterprise. But, it’s not that simple. As it turns out, consumer AI isn’t the same as enterprise AI. The questions you ask of a consumer AI system are different than what you’d ask of an enterprise AI system. For example, if you ask “what’s our vacation policy,” you expect the know-it-all to have read your company information. Or if you ask about “who are our top customers”, you might expect a list or table as a definitive response. And perhaps the most tricky element is permissions. While data on the open web uniformly accessible to anyone, data in the enterprise tends to have a complex network of permissions — so when you ask “who’s going to be promoted next week?”, you should only see the information you have access to.

A well-matched partnership: The Coda-Snowflake-Coda stack.

In our discussions with the Snowflake team, it became clear that we shared a similar view of the future of enterprise AI. And more importantly, we were both busy building complementary solutions to help address the problem.
On the Coda side, we have invested years of development in Coda Packs — our name for extensions to Coda that are commonly used to sync information from other tools into Coda docs. The most popular Packs get published publicly, and there are now 500+ of them in the Coda gallery, with thousands more built inside companies (often for accessing internal systems). Last month, 30% of active Coda docs used Coda Packs, and last year, Coda Packs synchronized 4B+ rows of data into Coda docs. Most importantly, Coda Packs are architected to directly handle the labyrinth of permission and security constraints in the Enterprise—and to replicate the security rules and permissions of underlying systems. So they are a perfect base for building a permissions-respecting “know-it-all.”
Snowflake Cortex forms all the core building blocks of constructing a scalable, secure AI system. Coda Brain uses almost every component in this stack — the Cortex Search engine that can vectorize and index unstructured and structured data, Cortex Analyst which can magically turn natural language queries into SQL, the Cortex Large Language Models that do everything from interpreting queries to reformatting responses into human-readable responses, and of course the underlying Snowflake data platform which can scale and securely handle the huge volumes of data being pulled into Coda Brain. As we explored this partnership, we quickly determined that between the two platforms, we have all the key ingredients of building a “RAG-based” enterprise AI solution. For those less familiar with the term, RAG stands for Retrieval Augmented Generation and is the name given to an increasingly popular AI technique where AI prompts are split into a Retrieval stage (where the system determines which data the user has access to) and a Generation stage (where the actual response is generated from that data). Between Coda Packs providing an up-to-date permissioned view of hundreds of enterprise data sources and Snowflake Cortex providing all the RAG engine primitives, we were now set to build Coda Brain.

Coda Brain: A turnkey AI platform powered by Snowflake that unites your company’s brain.

On top of these base building blocks, we had the key ingredients of a dream enterprise AI solution, and so we set out to develop Coda Brain. So what’s Coda Brain? Coda Brain is your team’s favorite know-it-all—a turnkey AI platform that understands your company data, can respond in text and tables, and is permission-aware. So your team is never caught in a frustrating loop of searching for context, interrupting the flow of others to find it, and submitting internal requests, just to see the same (and probably stale) data presented in slightly different ways. With Coda Brain, all you need to do is ask.

Here’s our vision:

Coda Brain understands your company data.

To bridge the gap between data and action across the enterprise, we knew that we’d need to make data accessible to the whole team. So we’re leveraging our ecosystem of 500+ Packs. Instead of syncing pieces of data from other apps into individual docs, your team can now sync all your relevant data into Coda Brain—a single, shared workspace that updates in real-time. And, as always, if you have an internal system or more niche integration that doesn’t have an existing Pack, you can always build your own.

Coda Brain is secure, inside and out.

When companies first started deploying AI tools, they focused on how to keep data from leaking outside of the company. But, these days, most companies are worried about security inside the company. To combat those concerns, we’ve built a RAG-based system that allows for personal authentication for each integration, which means that Coda Brain understands data permissions at the most granular level and ensures that everyone only gets answers they have access to. So, when you ask Coda Brain, “Who’s getting promoted next week?” you’ll only see the data you have specific access to.

Coda Brain is fluent in words and data.

The next challenge is getting people the right answers to the right questions. With ChatGPT and other enterprise assistant tools, you ask an unstructured (natural language) prompt and get back an unstructured (text) response and Coda Brain certainly does this well. But in the enterprise, we’ve found that many questions deserve a data-centered answer as well, something that both Coda and Snowflake are very good at. Building on Snowflake’s Cortex Analyst text-to-SQL engine, Coda Brain will not only return unstructured responses, but structured ones as well. So, when you ask Coda Brain “What are our Salesforce opportunities over $10,000?” you’ll see results in the form of a table full of data, which you can filter down further, turn into a chart, and use directly in your docs.

Coda Brain combats hallucinations with verifiable answers.

With consumer AI, we’ve all become used to hallucinations where the AI system seems to “make up” a response. In low-risk use cases, this can be tolerable and sometimes they’re even a bit fun. But in the world of enterprise AI, it’s important to make sure that companies can trust the quality of the answers they’re receiving. Like many RAG-based systems, Coda Brain will produce citations for answers—the same way that Coda Brain determines what you have access to, it will also include reference links in the answers. But Coda Brain goes a step further with structured data. When you ask a question like “what are our Sales opportunities over $10k,” you’ll not only get a structured response, but a live filter that you can change on the fly. Our early users have not only found this to be a delightful way to refine and expand on their question, but also a great way to verify the result is complete — just move the filter to see the rest of the underlying data.

Together, Coda Brain and Coda docs become brain-powered docs.

Getting thoughtful answers from Coda Brain is a useful (and rewarding) experience. But in many cases, the results are the basis for an ongoing workflow, not just a one-time question. We’ve made it incredibly simple to add any result given to you by Coda Brain into a Coda doc—so you can turn one-time queries into continuously refreshing results. And the resulting data is just normal Coda building blocks. For example, you can take a query for “customers who care about [feature X]”, and turn it into a live table where you add columns to track their feedback, or buttons and automations to automatically update them as their requests are addressed. In this way, Coda Brain can become a fantastic starting point for powerful Brain-driven workflows in Coda docs.

One more thing: Coda Brain has an accessible API.

Our early testers loved what Coda Brain can do for knowledge workers, but given the architecture of the product, we thought it would be useful to go beyond end users and give direct access to the Coda Brain backend as well. So your developer and data teams can use API and SQL interfaces to directly connect to Coda Brain: giving them a pre-filled, pre-configured, and pre-encoded data warehouse. In that way, Coda Brain is a platform truly available to your team, not just another application in your stack. Rather than spending months rebuilding your data warehouse to be permissions-aware and AI-ready, Coda Brain can do it for you.

Join the private preview for Coda Brain today.

Coda Brain is currently in a limited beta for enterprise teams. Sign up to learn more and get a spot on the waitlist.

Related posts

Explore more stories about Coda.
Coda + Snowflake: Democratizing data for everyone

Coda and Snowflake are partnering to turn data into action.

Coda acquires Plato to jointly spread engineering excellence

Coda has acquired Plato to better serve high-performing engineering teams.

How to simplify your SaaS stack and keep your data secure

Coda offers enterprise admins more control over their workspace.