Disclosure: This post is sponsored by Sigma.
We’ve all been to a user conference. At some of them, you sit through a lot of slides and nod politely. You eat a mediocre lunch and fly home with a tote bag and zero new ideas. Workflow 2026 was not that conference.
On March 5th, Sigma took over Convene in San Francisco. They spent a full day showing the community what it looks like when a platform is genuinely building toward something. Not roadmap theater. Not buzzword bingo. Actual, working product, big strategic bets, and a few moments that made the room go quiet in the best way.
Here’s what stood out.
Workflow 2026 Keynote
Mallory Busch, Mike Palmer, Zalak Trivedi, Katrina Menne, and Marwan Mattar kicked off the day, and the energy was right. Roadmap previews, product positioning, and a clear message: Sigma is an AI Application and BI platform, and they want you to know the difference between those two things.
Two announcements in particular got the crowd going.
Sigma Tables.
If you’ve been building in Sigma for a while, you know Input Tables. You love them, you use them, and you’ve also bumped into their biggest limitation: they live inside a workbook. They’re tied to it. You can’t reuse them elsewhere without rebuilding from scratch.
Sigma Tables changes that. The new model lets tables live independently of workbooks, data models, and reports. They can be shared across content. That is a genuinely big architectural shift. Think about what it means to build an AI Data App where the data structure can persist and be referenced across multiple workbooks or use cases. The friction that’s been quietly frustrating advanced builders for a while? This is how it gets removed.
Sigma Public.
If you caught the full breakdown on this one, you already know how excited this makes me. The short version: Sigma is building a public platform for AI Data Apps, giving practitioners a place to build, share, and get visibility outside their org. It’s the right call. The community is ready.
Sessions Worth Talking About
301: Integrating AI Into Your Workflows with AI Query
Fran Britschgi walked through three concrete use cases for AI Query: row-level analysis, column-level analysis, and multi-column analysis where you’re dealing with multiple value sets on the same record. That last one is where it gets interesting, and it’s where prompt engineering really starts to matter.
The session spent a meaningful chunk of time on exactly that. They discussed how to write prompts that actually work at scale inside a warehouse-native context. One specific technique to highlight is using functions like LISTAGG() and LISTAGGDISTINCT(). These functions aggregate multiple values before feeding them into an AI call. If you’re working with complex records where a single row has multiple related values you want analyzed together, that pattern is a game changer. It keeps your prompts cleaner and your results more coherent.
The broader point Fran made landed well: you don’t need to leave Sigma to build intelligent, AI-powered workflows. The capability is right there inside the workbook, sitting on top of your warehouse.
(Side note: these sessions were not recorded, so if you weren’t in the room, that’s genuinely a bummer. Another good reason to make the trip next year.)
201: Best Practices for Building AI Applications
Katrina Menne and Greg Bonnette introduced a framework for building AI Applications that actually scale, and they gave it a name worth remembering: BUILD.
- B – Business Objective
- U – User Path
- I – Input Modeling
- L – Look and Feel
- D – Development Cycle
It’s clean, it’s memorable, and it maps to real decisions you face every time you start a new build. A lot of people are building in Sigma right now without necessarily thinking about what happens when requirements triple or fifty more users show up. This framework is the “build it right the first time” conversation the community needed.
The Art of the Possible: Sigma Eating Its Own Cooking
This was the session of the day. No contest.
Orla Clifford, Sigma’s VP of Operations, was the main storyteller, and she’s genuinely good at it. The setup: Sigma, a company of about 800 employees, looked hard at their SaaS spend and asked themselves an uncomfortable question. If we’re an application-building platform, why are we paying millions of dollars for tools we could build ourselves?
So they challenged themselves to find out.
Greg Bonnette went first, walking through the internal AI App built to handle expense submission and approval. Real writeback to the warehouse, real workflow automation, running in production.
Jake Hannan, Sigma’s Head of Data, covered how they’ve rebuilt sales department workflows inside Sigma. The kind of stuff that usually lives in a CRM.
And then Luke Stanke closed it out with the one that made the room lean forward: a full content marketing pipeline, powered by a stack of Sigma Agents. One agent handles web search and SEO optimization. One works from a Sigma knowledge base to ensure factual accuracy. One manages voice and writing style. The whole pipeline runs inside Sigma, in production, generating and optimizing marketing content.
That’s not a demo. That’s production.
And then came the headline: Sigma believes they are on a clear path to fully replacing Salesforce within the next year.
Let that land for a second. Salesforce. The platform everyone assumes is untouchable infrastructure. Sigma is betting they can build everything they need natively, with Sigma Agents doing the heavy lifting. And the way Orla framed it was something close to: if we can do it, so can you. With Sigma.
Sigma Agents: The Thread Running Through Everything
Sigma Agents are not fully GA yet, but the direction is crystal clear, and the Art of the Possible session made the case better than any product slide ever could.
Agents are autonomous, action-taking AI components that live inside your Sigma Data Apps. They don’t just answer questions about your data. They do things. Execute tasks. Trigger workflows. Write back to the warehouse. Operate without a human in the loop at every step.
The 302 session on Chat Agents made the technical case. The Art of the Possible made the business case. Together, they painted a picture of what Sigma is actually building toward: not a better dashboard tool, but the operational layer your business runs on.
That’s a big swing. And based on what Sigma showed with their own internal stack? It’s a swing they’re already landing.
Sigma Builder Challenge
Capping off the day was the Builder Challenge, where four expert app builders went head-to-head for the coveted Golden Goat Award. Equal parts competition and community energy, it was a genuinely fun way to close a dense day of learning. Big shoutout to Katrina Menne and Kaela Dickens for running it!
The Bigger Picture
Workflow 2026 told a coherent story. Sigma Tables removes friction for builders. Sigma Public gives the community a stage. Sigma Agents give the platform a reason to exist in places far beyond the dashboard. The Art of the Possible session made a convincing argument. It showed that Sigma themselves are betting their own operations on it.
That’s not marketing. That’s commitment.
The community has been building cool things in Sigma for a while. It feels like the platform is finally building the infrastructure to match the ambition of the people using it.
What session would you have wanted to be in the room for? And if Sigma Agents go GA tomorrow, what’s the first thing you’re automating? Head over to LinkedIn and let’s talk about it.
Go Forth and Viz.
This post is sponsored by Sigma. All opinions are based on firsthand attendance at Workflow 2026.