Surojit Chatterjee is the founder and CEO of Ema. Beforehand, he guided Coinbase by means of a profitable 2021 IPO as its Chief Product Officer and scaled Google Cell Advertisements and Google Procuring into multi billion greenback companies because the VP and Head of Product. Surojit holds 40 US patents and has an MBA from MIT, MS in Pc Science from SUNY at Buffalo, and B. Tech from IIT Kharagpur.
Ema is a common AI worker, seamlessly built-in into your group’s present IT infrastructure. She’s designed to reinforce productiveness, streamline processes, and empower your groups.
Are you able to elaborate on the imaginative and prescient behind Ema and what impressed you to create a common AI worker?
The aim for Ema is evident and daring: “rework enterprises by constructing a common AI worker.” This imaginative and prescient stems from our perception that AI can increase human capabilities quite than exchange staff completely. Our Common AI Worker is designed to automate mundane, repetitive duties, releasing up human staff to concentrate on extra strategic and beneficial work. We do that by means of Ema’s revolutionary agentic AI system, which might carry out a variety of advanced duties with a set of AI brokers (referred to as Ema’s Personas), enhancing effectivity, and boosting productiveness throughout numerous organizations.
Each you and your co-founder have spectacular backgrounds at main tech corporations. How has your previous expertise influenced the event and technique of Ema?
Over the past 20 years, I’ve labored at iconic corporations like Google, Coinbase, Oracle and Flipkart. And at each place, I puzzled “Why will we rent the neatest individuals and provides them jobs which can be so mundane?.” That is why we’re constructing Ema.
Previous to co-founding Ema, I used to be the chief product officer of Coinbase and Flipkart and the worldwide head of product for cell adverts at Google. These experiences deepened my technical data throughout engineering, machine studying, and adtech. These roles allowed me to establish inefficiencies within the methods we work and find out how to clear up advanced enterprise issues.
Ema’s co-founder and head of engineering, Souvik Sen, was beforehand the VP of engineering at Okta the place he oversaw knowledge, machine studying, and units. Earlier than that, he was at Google, the place he was engineering lead for knowledge and machine studying the place he constructed one of many world’s largest ML methods, centered on privateness and security – Google’s Belief Graph. His experience, significantly, is a driving power to why Ema’s Agentic AI system is very correct and constructed to be enterprise prepared by way of safety and privateness.
My cofounder Souvik and I believed what should you had a Michelin Star Chef in-house who may prepare dinner something you requested for. You could be within the temper for French at present, Italian tomorrow and Indian the day after. However no matter your temper or the delicacies you want, that chef can recreate the dish of your goals. That’s what Ema can do. It could tackle the function of no matter you want within the enterprise with only a easy dialog.
Ema makes use of over 100 massive language fashions and its personal smaller fashions. How do you guarantee seamless integration and optimum efficiency from these different sources?
LLM’s, whereas highly effective, fall brief in enterprise settings resulting from their lack of specialised data and context-specific coaching. These fashions are constructed on common knowledge, leaving them ill-equipped to deal with the nuanced, proprietary info that drives enterprise operations. This limitation can result in inaccurate outputs, potential knowledge safety dangers, and an incapability to supply domain-specific insights essential for knowledgeable decision-making. Agentic AI methods like Ema deal with these shortcomings by providing a extra tailor-made and dynamic strategy. In contrast to static LLMs, our agentic AI methods can:
- Adapt to enterprise-specific knowledge and workflows
- Leverage a number of LLMs based mostly on accuracy, value, and efficiency necessities
- Keep knowledge privateness and safety by working inside firm infrastructure
- Present explainable and verifiable outputs, essential for enterprise accountability
- Constantly replace and be taught from real-time enterprise knowledge
- Execute advanced, multi-step duties autonomously
We guarantee seamless integration from these different sources through the use of Ema’s proprietary 2T+ parameter combination of specialists mannequin: EmaFusionTM. EmaFusionTM combines 100+ public LLMs and plenty of area particular customized fashions to maximise accuracy on the lowest potential value for large number of duties within the enterprise, maximizing the return on funding. Plus, with this novel strategy, Ema is future-proof; we’re continuously including new fashions to forestall overreliance on one know-how stack, taking this danger away from our enterprise clients.
Are you able to clarify how the Generative Workflow Engine works and what benefits it provides over conventional workflow automation instruments?
We’ve developed tens of template Personas (or AI staff for particular roles). The personas will be configured and deployed rapidly by enterprise customers – no coding data required. At its core, Ema’s Personas are collections of proprietary AI brokers that collaborate to carry out advanced workflows.
Our patent-pending Generative Workflow Engine™, a small transformer mannequin, generates workflows and orchestration code, choosing the suitable brokers and design patterns. Ema leverages well-known agentic design patterns, equivalent to reflection, planning, device use, multi-agent collaboration, language agent tree search (LATS), structured output and multi-agent collaboration, and introduces many revolutionary patterns of its personal. With over 200 pre-built connectors, Ema seamlessly integrates with inner knowledge sources and might take actions throughout instruments to carry out successfully in varied enterprise roles.
Ema is utilized in varied domains from customer support to authorized to insurance coverage. Which industries do you see the best potential for development with Ema, and why?
We see potential throughout industries and features as most enterprises have lower than 30% automation in processes and use greater than 200 software program purposes resulting in knowledge and motion silos. McKinsey & Co. estimates that generative AI may add the equal of $2.6 trillion to $4.4 trillion yearly in productiveness good points (source).
These points are exacerbated in regulated industries like healthcare, monetary providers, insurance coverage the place a lot of the final a long time technical automations haven’t occurred because the know-how was not superior sufficient for his or her processes. That is the place we see the most important alternative for transformation and are seeing a whole lot of demand from clients in these industries to leverage Generative AI and know-how like by no means earlier than.
How does Ema deal with knowledge safety and safety issues, particularly when integrating a number of fashions and dealing with delicate enterprise knowledge?
A urgent concern for any firm utilizing agentic AI is the potential for AI brokers to go rogue or leak non-public knowledge. Ema is constructed with belief at its core, compliant with main worldwide requirements equivalent to SOC 2, ISO 27001, HIPAA, GDPR, NIST AI RMF, NIST CSF, NIST 800-171 To make sure enterprise knowledge stays non-public, safe, and compliant, Ema has carried out the next safety measures:
- Automated redaction and protected de-identification of delicate knowledge, audit logs
- Actual-time monitoring
- Encryption of all knowledge at relaxation and in transit
- Explainability throughout all output outcomes
To go the additional mile, Ema additionally checks for any copyright violations for doc technology use instances, lowering clients’ likelihood of IP liabilities. Ema additionally by no means trains fashions on one buyer’s knowledge to profit different clients.
Ema additionally provides versatile deployment choices together with on-premises deployment capabilities for a number of cloud methods, enabling enterprises to maintain their knowledge inside their very own trusted environments.
How simple is it for a brand new firm to get began with Ema, and what does the standard onboarding course of appear to be?
Ema is extremely intuitive, so getting groups began on the platform is sort of simple. Enterprise customers can arrange Ema’s Persona(s) utilizing pre-built templates in simply minutes. They will wonderful tune Persona conduct with conversational directions, use pre-built connectors to combine with their apps and knowledge sources, and optionally plug in any non-public customized fashions educated on their very own knowledge. As soon as arrange, specialists from the enterprise can practice their Ema persona with only a few hours of suggestions. Ema has been employed for a number of roles by enterprises equivalent to Envoy World, TrueLayer, Moneyview, and in every of those roles Ema is already acting at or above human efficiency.
Ema has attracted vital funding from high-profile backers. What do you consider has been the important thing to gaining such sturdy investor confidence?
We consider traders can see how Ema’s platform permits enterprises to make use of Agentic AI successfully, streamlining operations for substantial value reductions and unlocking new potential income streams. Moreover, Ema’s administration workforce are specialists in AI and have the required technical data and ability units. We even have a powerful monitor file of enterprise-grade supply, reliability, and compliance. Lastly, Ema’s merchandise are differentiated from anything available on the market, it’s pioneering the newest technical developments in Agentic AI, making us the go-to selection for any enterprise wanting so as to add next-generation AI to their operations.
How do you see the function of AI within the office evolving over the subsequent decade, and what function will Ema play in that transformation?
Ema’s mission is to remodel enterprises and assist each worker work quicker with the assistance of simple-to-activate and correct brokers. Our common AI worker has the potential to assist enterprises execute duties throughout buyer assist, worker assist, gross sales enablement, compliance, income operations, and extra. We’d like to remodel the office by permitting groups to concentrate on probably the most strategic and highest-value tasks as an alternative of mundane, administrative duties. As a pioneer of agentic AI, Ema is main a brand new period of collaboration between human and AI staff, the place innovation thrives, and productiveness skyrockets.
Thanks for the nice interview, readers who want to be taught extra ought to go to Ema.
