San Francisco-based Ema, the AI agent startup founded by former Google and Okta employees, today announced it is raising an additional $36 million as part of a Series A fundraising round.
The investment takes the company’s total raise to $61 million and was led by Accel and Section 32.
The company says it’ll use the cash to further develop its proprietary tech allowing enterprises to configure and deploy no-code AI agents — what Ema calls “universal AI employees” — capable of handling various tasks across functions.
“Our goal at Ema is to help automate most of the mundane tasks that human employees perform today and free them up to do more valuable work in the enterprise. We’ve built Ema as a universal AI employee. Ema can morph into taking on any role in the organization — from customer support, employee experience, sales & marketing to legal & compliance,” Surojit Chatterjee, the CEO and co-founder of the startup, told VentureBeat.
Ema emerged from stealth a few months ago and is already seeing significant traction, with its AI employees being deployed across organizations in fintech, legal, healthcare and e-commerce.
The fresh funding marks another vote of confidence in the company’s tech stack, but it’s not going to be an easy win anytime soon. Over the last year or so, several vendors have emerged in this space, tapping the power of foundational models to empower enterprises with ready-to-use AI agents.
What does Ema bring to the table?
Before the rise of OpenAI’s ChatGPT in late 2022, enterprises automated basic tasks like customer support by deploying rigid, flow-based chatbots into their stacks.
The offerings did the job but failed to deliver the required answers customers expected as they had no contextual knowledge and learning.
However, when large language models (LLMs) appeared on the scene, the experience of these chatbots got a whole new upgrade. Eventually, this translated into the idea of powerful AI agents — LLM-powered systems that could not only provide reliable answers but also take complex actions across multiple enterprise applications, working with any kind of data.
Imagine an AI agent for customer support that actually cancels your order upon request, instead of directing you to the page to cancel it.
With the idea of a universal AI employee, Ema is delivering this exact experience and providing enterprises with an agentic system that can take up any role in the organization, right from handling customer service and technical support to providing support for sales and marketing.
No-code agentic platform and AI employee templates
At the core, the company offers a no-code agentic platform, where users can access a library of pre-built AI employee templates.
Once the user chooses an agent for a given use case, the platform runs a guided conversation, allowing them to quickly fine-tune and deploy the finalized AI employee (or Ema persona) for making decisions, creating plans, orchestrating enterprise workflows — while collaborating seamlessly with humans at the same time.
“Ema empowers enterprise customers to create tailor-made personas by specifying goals, resources, and constraints. This level of customization was previously the domain of AI experts and data scientists. Now, with just a few guided conversations and straightforward configuration, enterprises can create and deploy new AI employees customized for specific roles within their organization, faster than ever before. This capability doesn’t just expand the reach of Agentic AI—it democratizes it,” Chatterjee said.
Under the hood, Ema’s agent deployment experience is driven by a Generative Workflow Engine, a small transformer model that generates workflows and associated orchestration code, selecting appropriate agents and design patterns. When configuring the agent, the engine allows users to connect their desired data sources and applications with a library of over 200 connectors.
This way, the deployed AI employee gets contextual awareness (covering documents, logs, data, code and policies) as well as the ability to take actions across systems.
To ensure the agent works accurately after deployment, the company leverages a 2T+ parameter mixture of experts model called EmaFusion. It combines 100+ public LLMs and domain-specific custom models to maximize accuracy at the lowest possible cost.
Chatterjee also confirmed that users have the option to plug in any private custom models, trained on their own data, to guide the behavior of their AI employee. Plus, the company has robust data protection and security algorithms in place to ensure all enterprise data going into the agent remains private and secure.
“We’ve implemented robust systems for safe redaction and de-identification of sensitive data, rigorous copyright violation checks (in document generation cases), end-to-end encryption of data both in transit and at rest, comprehensive audit logging, real-time monitoring, output explainability and frequent penetration testing. We’re also fully compliant with the top international standards,” he explained.
Goal to expand in a competitive market
While the CEO did not share exact revenue or customer specifics, he did note that the number of companies using Ema has tripled since it emerged from stealth in March 2024 – across sectors such as fintech, legal, healthcare, ecommerce and insurance.
“Ema has been hired for multiple roles by enterprises such as Envoy Global, TrueLayer, Moneyview, and in each of these roles Ema is already performing at or above human performance,” the CEO said.
As the next step, the CEO said, the company will use the funding to further develop its technology and expand its go-to-market team with the goal of better meeting the demand from potential customers.
That said, it will be interesting to see how the company continues to stand out in this rapidly expanding market. Other notable players setting up conversational AI agents for enterprises are Decagon, Yellow AI, Cognigy, Rasa and Kora AI.
Even Bret Taylor, who is the director of the board at OpenAI, has ventured into the category with a startup named Sierra. It has raised $110 million from notable venture capital firms and is racing to tap the power of large language models to enable enterprises to build always-available AI agents for their respective businesses.
Author: Shubham Sharma
Source: Venturebeat
Reviewed By: Editorial Team
