Stack AI’s co-founders, Antoni Rosinol and Bernardo Aceituno, were in the final stages of their PhD programs at MIT in 2022, right as large language models began gaining prominence. Well before ChatGPT made its debut at the year’s end, they identified a significant issue within companies integrating data with models without much specialized knowledge – a challenge they aimed to address.
Upon completing their doctorates, they relocated to San Francisco and participated in Y Combinator’s Winter 2023 cohort. It was here that they officially launched Stack and honed their concept. Today, Stack AI offers a low-code workflow automation tool designed to assist companies in creating AI-driven automations like chatbots and AI assistants. To date, the startup has secured $3 million in funding.
“Our platform enables users to establish workflows that necessitate the integration of various tools. We emphasize connecting data sources and LLMs, facilitating the creation of robust workflow automations. Additionally, we provide a suite of tools and functionalities to automate intricate business processes,” Aceituno explained to TechCrunch. Although the product has been live for just six months, it already boasts over 200 customers.
The platform operates through a drag-and-drop interface on a workflow canvas. Typically, this involves a data source such as Google Drive, a language model, and additional workflow components like triggers and actions, enabling users to build generative AI applications with minimal coding. While the actual coding isn’t AI-driven, the workflow tasks are, occasionally necessitating some manual coding for smoother operations.
A significant portion of their initial clientele comes from the healthcare sector. Aceituno highlights the importance of caution when dealing with applications involving medical professionals and patients, particularly when internal data sources might be unreliable or conflicting.
In such sensitive scenarios, human expertise is crucial. Doctors must evaluate answer quality, and to aid in this, Stack AI includes source citations with every answer, allowing healthcare professionals to verify information before accepting it.
“That being said,” Aceituno acknowledged, “the system’s output quality can only be as good as the input data, meaning if you input flawed data, the citations will also be flawed. This is why these AI assistants shouldn’t completely replace the human decision-making process.”
Reflecting on their journey from MIT to launching a startup, Rosinol noted that their experiences at Y Combinator were instrumental in grasping the business aspects and refining their startup idea through customer interaction.
“Our initial API version was more developer-focused, aimed at automating tasks like RFP responses or sales. However, through working with customers, we realized the primary challenge: effectively querying and connecting data sources to language models was more crucial than merely training a model.”
Currently, the company has a team of six, with plans to expand by hiring engineers and sales and marketing professionals.
The $3 million funding round, completed about a year ago, saw participation from Gradient Ventures, Beat Ventures, and True Capital, among others, including LambdaLabs, Y Combinator, Soma Capital, and Epakon Capital.
