Ema, a startup building teams of AI agents for HR, IT and finance processes, has raised a $77 million Series B led by Creaegis. Existing investors Accel, Section 32 and Prosus also participated.
The all-primary financing brings Ema’s total funding to $140 million and values the company at more than four times its 2024 valuation, though Ema did not disclose the current figure. The company’s premise is consequential for enterprise buyers: AI may increasingly be bought to complete an outcome across existing systems rather than to add another interface, seat license or point tool.
From copilots to process ownership
Founded in 2023, Mountain View-based Ema calls its systems “AI employees.” The product coordinates multiple agents across a customer’s existing applications to carry out multi-step workflows, rather than addressing one task at a time.
That distinction is central to the company’s pitch. Ema initially wraps around an enterprise’s installed software, according to co-founder and CEO Surojit Chatterjee. Over time, he argues, customers may reduce their dependence on some large SaaS products when the AI layer can manage the workflow and the underlying software functions primarily as a system of record.
For operators, this is a more demanding proposition than deploying a chat assistant. End-to-end automation requires reliable access to systems, well-defined handoffs, controls for exceptions and clear accountability when an action has financial, security or employee consequences. The potential value is also larger, because it targets both software spend and the implementation and support labor surrounding it.
A new competitor—and partner—for services firms
Ema says it is also pursuing work traditionally performed by IT services and consulting firms: implementation, integration and process execution. That does not necessarily make services firms simple losers. Chatterjee said services companies are working with Ema as they adapt their own delivery models.
That relationship may prove important. Large enterprises often need domain expertise, integration capability and change-management support before automation reaches production. Services firms can distribute AI platforms into established accounts, while AI vendors can help those firms deliver more work with fewer billable human hours. The commercial tension is obvious: productivity gains may pressure time-and-materials revenue models even as they create new advisory and managed-automation opportunities.
Ema reports more than 50 active enterprise deals and over 1 million active enterprise users. Its named customers include NTT DATA, Hitachi, ADP, PwC, Google, KPMG, Wipro and Microsoft. The company said revenue has grown 50-fold over two years and that bookings have exceeded $150 million, but it cautioned that bookings include the total value of multiyear contracts rather than annual recurring revenue. It did not disclose its current revenue run rate.
The startup also says more than 90% of customers expand beyond their first use case and reports net dollar retention of about 180%. Those are company-reported figures, but they point to the key adoption question: whether a successful initial workflow can become a repeatable automation platform across departments.
Model choice becomes an architectural decision
Ema says its platform can use more than 150 frontier and open-source models. Its differentiation, it argues, is the orchestration, integrations and business-specific knowledge built around those models—not the base model itself.
That positioning puts Ema in a crowded layer of the market. Anthropic and OpenAI are expanding enterprise efforts, while incumbent software vendors are adding agents into their own suites. A model-agnostic approach can give buyers flexibility on cost, performance and data requirements. But it also raises the bar for the orchestration layer: it must demonstrate dependable governance, observability and measurable workflow results regardless of which model powers a step.
What to watch next
Ema plans to use much of the new capital for sales and marketing, after focusing its early years on product development. With nearly 200 employees and offices in Bengaluru, London and Vancouver, it also plans geographic expansion beyond its U.S. and European base into Asia-Pacific, South America and parts of the Middle East.
The test is not whether enterprises will experiment with agents; they already are. It is whether vendors such as Ema can prove that outcome-based pricing, reportedly tied to completed tasks rather than seats or tokens, produces auditable savings without creating new operational risk. The winners will need to show not merely that an agent can act, but that it can be trusted to run a business process at scale.
