Ema, a ‘Universal AI employee’, emerges from stealth with $25M

Generative AI well and truly has a grip on public technology discourse these days. A new startup called Ema out of San Francisco thinks it’s a lot more than just a passing fancy. It’s emerging from stealth today, with a product of the same name, that it believes will open a new chapter in how AI, and specifically generative AI, will change how we work.

“Our goal is to build a universal AI employee,” Surojit Chatterjee, the CEO and co-founder, said in an interview. “Our goal is to automate on the mundane tasks that employees do on a day to day basis in every enterprise… to free them up to do more valuable and more strategic work.”

The company, and investors, are putting money and revenues where its mouth is: it’s already raised $25 million from an impressive list of backers, along with customers that it quietly amassed while still in stealth, to blow away any accusations of vaporware, include Envoy Global, TrueLayer and Moneview.

As for what Ema can do, these businesses are using it in applications that range from customer service — including offering technical support to users as well as tracking and other functions — through to internal productivity applications for employees. Ema’s two products — Generative Workflow Engine (GWE) and EmaFusion — are designed to “emulate human responses” but also evolve with more usage with feedback.

As Chatterjee describes it, it’s not just robotic process automation (that is so 2010’s) and it’s not just AI to accelerate certain tasks (that’s going back even further), and it’s not just another gen AI accuracy fail waiting to be lampooned on social media.

Chatterjee says that Ema — which is an acronym for “enterprise machine assistant” — taps into more than 30 Large Language Models, he said, and combines that with its own “smaller, domain specific models” in a patent-pending platform “to address all the issues you have seen with accuracy, hallucination, data protection and so on.”

This early round is adding a lot of names to Ema’s cap table. Accel, Section 32 and Prosus Ventures are co-leading, and Wipro Ventures, Venture Highway, AME Cloud Ventures, Frontier Ventures, Maum Group and Firebolt Ventures also participating. On top of this there are also some big-name individual backers: Sheryl Sandberg, Dustin Moskovitz, Jerry Yang, Divesh Makan, and David Baszucki among them.

There are already dozens, maybe hundreds, of companies building GenAI tools for enterprises at the moment, both those working on solutions for particular verticals or use cases, as well as ambitious home-run style swings like Ema’s. If you’re wondering why this particular GenAI startup is capturing attention from these investors, some of that might be because of the fact that they’re already drumming up business. But it’s also because of some of the background of the team.

Previous to Ema, Chatterjee was chief product officer of Coinbase leading up to its IPO. Before that, he was VP of Product at Google in both its mobile ads and shopping businesses. He himself has some 40 patents to his name in areas like machine learning enterprise software and ad tech.

The other co-founder, Souvik Sen, who is Ema’s head of engineering, has some equally impressive experience. Most recently, he was VP of engineering at Okta where he oversaw data, machine learning and devices; and prior to that he was also at Google, where he was engineering lead for data and machine learning where he focused on privacy and safety. He himself also has 37 patents.

The combined experience of these two both gives a weight to the company’s ambitions and likelihood of being able to execute on them. But it also drops lots of details that may well figure in how it evolves.

For example, consider Chatterjee’s expertise in e-commerce and adtech. Given that these are such cornerstones of how so many businesses interact with customers today, it feels inevitable that they will figure in how Ema might evolve if it flies.

On the other hand, having a founder who has previously had to build in and account for data protection and privacy potentially gives the startup a better chance of not making a mess of these. Or at least we can hope! It is AI after all, and this is a Silicon Valley startup that will ultimately be focused on business at hand and how to use technology to achieve it.

For the moment, it’s notable to see ambitious startups working to build products that cut across different LLM silos to achieve more advanced results. It is perhaps an early sign of how the LLMs are more interchangeable than you might assume over time, and more commoditized, too.

And the ability to cut across different use cases gives the startup a potential diversification that could help grow its business and usefulness overall, investors say.

“Most point Gen AI solutions provide high value for specific use cases but are either hard to expand across use-cases or even adjacent use-cases and more importantly, large enterprises are worried about fragmentation and access to their sensitive data by so many different applications,” Ashutosh Sharma, head of investments for Prosus Ventures in India, told TechCrunch. “Ema can solve for these problems and deliver high accuracy with optimal return on investment.”