Stack AI's co-founders, Antoni Rosinol and Bernardo Aceituno, were PhD students at MIT who are finishing their degrees in 2022 as large-scale language models become mainstream. ChatGPT was scheduled to be released to the world at the end of the year, but even before then, we recognized the problem within the company of combining data with models without a lot of expertise and knowledge, and we wanted to change that.
After graduation, they moved to San Francisco and joined Y Combinator's Winter 23 cohort to launch Stack and refine their ideas. Today, the company has built low-code workflow automation tools designed to help businesses build AI-based workflows, including chatbots and AI assistants. The company has raised $3 million to date.
“Our platform allows people to connect disparate tools to build workflows that require them to work together. We focus on connecting data sources with LLM. Because it allows you to build powerful workflow automation. We also offer many other tools and features to automate complex business processes,” Aceituno told TechCrunch. They've only been launching the product for six months, but they already have over 200 customers using it, they report.
Basically, this involves dragging components onto the workflow canvas. This typically includes data sources such as Google Drive and LLM along with other workflow components such as trigger components or action components to build the workflow, allowing customers to create generative AI programs without much coding. Although the coding itself is not AI-based, the tasks in a workflow often require manual coding to ensure that the workflow operates smoothly.
Some of its early customers are in the healthcare industry and Aceituno acknowledges that it needs to be careful with applications related to doctors and patients, especially when internal data sources may not always be reliable, contain contradictory or useless information.
In such cases, he says, it's important to rely on human experts – doctors – to determine the quality of the answers. As another level of protection, all responses include source citations, allowing medical professionals to verify sources before accepting responses.
“So it’s true that if you put in garbage, the citations will also be garbage. That’s why these assistants shouldn’t completely take over the process,” he said.
Rosinol, who started his startup right out of MIT, says that going to YC really helped him understand the business side of things and how to work with customers to bring startup ideas to life.
“We started with an early version of this API that was much more developer-centric. And we started with a few of our customers with this idea that they wanted to use AI to automate RFP responses or automate sales. And as we worked with our customers, it became very clear that the real challenge was not training models, but effectively querying data sources and connecting to these language models.”
The company currently has six employees but is hiring engineers, sales and marketing specialists.
The $3 million investment closed about a year ago. Investors include Gradient Ventures, Beat Ventures, True Capital, LambdaLabs, Y Combinator, Soma Capital, and Epakon Capital.