VB Transform 2024 returns this July! Over 400 enterprise leaders will gather in San Francisco from July 9-11 to dive into the advancement of GenAI strategies and engaging in thought-provoking discussions within the community. Find out how you can attend here.
In the rush to train AI in a company’s knowledge library, the reality is that some important information may inadvertently be left out. It’s easy to upload images, videos, documents and spreadsheets to educate models, but what about other unstructured data types like invoices, emails, and PDFs? Though it is possible to convert those files into something searchable, it’s done manually and is time-consuming—not to mention that engineering teams may not have the machine-learning capabilities necessary.
However, one company believes it has a solution organizations are looking for. Founded by Antonio Bustamante and Upal Saha, Bem is an AI data interface company that provides an API developers can use to convert any input, structured or unstructured, into any data shape. The startup has raised $3.7 million in seed funding to support this mission, with the round led by Uncork Capital. Lookout’s founder Kevin Mahaffey, Roar Ventures, and angel investors such as Y Combinator’s Garry Tan also participated.
“Engineering teams spend about 44 percent of their time building and then monitoring and maintaining their data pipelines,” Bustamante writes in a statement. “At our core, Bem is an AI data interface, redefining the scope of interoperability and integration to save time and frustration for these teams so that they can focus on what truly matters to their product. Our goal is for engineers to not worry about integrations or data ingestion pipelines ever again – we want them to focus on valuable work for their users.”
Building Structured Data as a Service
Targeted at engineers, Bem is billed as easy to use, requiring no training or configuration. Developers can use the company’s API to specify their desired data shape or schema before sending over their information. Bustamante compares it to how Stripe started with a pure API that’s simple to implement.
VB Transform 2024 Registration is Open
Join enterprise leaders in San Francisco from July 9 to 11 for our flagship AI event. Connect with peers, explore the opportunities and challenges of Generative AI, and learn how to integrate AI applications into your industry. Register Now
It’s not a one-off service, either. Bem, which stands for “brilliant enterprise magic,” is a data pipeline transmitted from an organization to Bem and then back again. “The customer sets up a pipeline and sends us thousands and thousands of pieces of data every minute, every hour. It’s a continuous stream of data.”
In a way, it’s providing Structured Data as a Service (SDaaS) for engineering teams. It’s something Bustamante agrees with. “That’s a great way of putting it. All of these inputs [companies are] sending us are just incredibly messy,” he shares with VentureBeat in an interview. All our users need is a structured version of all those inputs that look like their internal data shape and schema. And so, at the end of the day, we’re offering them just a transformation-structuring service.”
Bem believes there’s enormous potential for its service. Besides a lack of in-house expertise, companies are too busy to develop their own solution to what Bustamante describes as a “hair on fire” problem.
“We work with software companies building products in logistics, supply chain, health care insurance. Most unstructured to structured systems out there are more made for document review workflows. If you have an insurance company that needs to review a bunch of documents, we are what we call a primitive—we’re very much a product building block. So, if you want to integrate this type of capability directly into your end product and expose it to your end users, you would use us as opposed to another provider that is more apt for having a back office process where you review financial statements and things like that.”
Bem is built atop a combination of foundation and open-source models that the startup has fine-tuned and trained. In addition, the models gain intelligence the more frequently customers use their pipeline. However, Bem doesn’t train global models on customer data, so everything is isolated from each other. “These tiny models that they get, they get trained. They only get trained for each of our customers. We never share data across different models.”
Raising capital to go to market
Bem is live but accessible through a private beta. Bustamante shares that the company has 10 early customers using the product, and the startup is targeting organizations between the Series C and public stages. He feels that when a company reaches Series C, that’s when this problem becomes “incredibly evident.” Furthermore, Bustamante highlights those in logistics, supply chain, health care and insurance as four industries that “suffer this pain the most.”
The $3.7 million the startup received will go towards further developing Bem’s platform. Specifically, it plans to invest in its engineering team, research and development and product improvement.
“I love companies who are taking the work that every company has to do and generally does it badly, and then productive it and make it available as a service,” Andy McLoughlin, Uncork Capital’s managing partner, tells VentureBeat. “Every company I’ve ever worked with has always had some issue with messy data ingestion. For some companies, if they’re working in insurance, logistics or agriculture, it’s even more pronounced, but every company has this. And right now, the status quo has been whatever comes in and somebody has to manually review it. Having something that just kind of automatically does it for you and it just works is such a no-brainer.”
He believes what Bem is doing is the democratization of data and the power of data. For the most part, only big technology companies int he world have had the infrastructure and capabilities to transform data like Bem is offering. “What we’ve seen from every kind of developer infrastructure, developer tool and company we’ve invested in is when they make that leap from selling to companies that look a little bit like them, to sell it to companies your parents would have heard of, that’s when you know you’re beginning to nail it because all of these companies, they want to do better, they want to build better, they want to make software in the same way as Facebook does. But they need the tooling in order to do so. And that’s what we love to invest in.”
Prodded to list its competitors, Bem’s CEO called out Unstructured. He hedged, though, saying Unstructured focuses on a different customer base and deals only with documents. But perhaps Bem’s biggest competitor is the status quo, companies trying to build this capability in-house. “Fortunately for us, a lot of our customers don’t actually want to do that. And some of the ones that do start to build this themselves realize it will be way more expensive than they initially thought.”
Bem has yet to formalize its pricing, saying that currently, it’s done on a “case by case basis,” and the company measures each piece of data sent and designs a cost plan around it. Bustamante says over the next few months, he hopes to standardize the pricing structure to ensure it’s “very tangible and visible” and deliver an ROI of at least 25 to 30 times “because this is such an expensive problem to solve.”
No valuation was disclosed.
