Ema raises $77 million to replace layers of enterprise software

What Ema is selling

Ema calls its technology “AI employees.” The product is designed to run multi-step business processes rather than answer a single request inside one application:

The company’s pitch is not another application-specific assistant. It is an orchestration layer that sits across the software a company already uses.

Chatterjee believes that layer could eventually reduce dependence on traditional software, including SaaS. Ema first “wraps” existing enterprise applications with an additional layer of automation. Once customers depend on that layer, they may reduce their use of some underlying products or replace them entirely.

According to Chatterjee, many Ema customers are already moving toward replacing major SaaS applications, which gradually become little more than databases that can be discarded.

Ema did not disclose a new valuation. The round contained only primary capital: no debt and no secondary transactions.

Traction, with one important caveat

Ema says its commercial footprint already includes:

Its customers include:

Revenue has grown 50 times over the past two years, while booked contracts have exceeded $150 million. Chatterjee clarified that the figure includes the full value of multi-year agreements, including two- and three-year contracts; it is not annual recurring revenue. He did not disclose Ema’s current annual revenue run rate.

More than 90% of customers have expanded their use beyond the initial deployment scenario, with some rolling Ema out across dozens of workflows. Net dollar revenue retention is about 180%, meaning existing customers materially increase their spending over time.

That distinction around the $150 million matters. The number signals demand, but it does not tell us how much revenue Ema is currently collecting each year or how quickly those contracts convert into operating results. I think the expansion and retention figures are more revealing than the headline bookings number: they suggest Ema is becoming embedded after the first workflow works.

The larger target is services

Ema is also aiming at the IT-services work that often accompanies enterprise software: implementation, integration and consulting. Chatterjee says AI can take over part of that work, while many service companies already working with Ema are changing their business models because a model centered on human labor may fit the future market less well.

The company says it maintains gross margins of about 80% despite automating tasks once performed by software and services providers. As the AI systems learn from new deployments, Ema expects to need less human support, which should improve margins over time.

Its pricing is tied neither to the number of software seats nor to AI-token consumption. Customers pay based on completed tasks and achieved business results.

The announcement is quiet about the operational boundary between those claims and the product’s current reality: how much of each deployment runs without human support, how much work still sits with Ema’s team, and which SaaS replacements are already complete rather than aspirational. That gap matters because outcome-based pricing can make automation look simpler from the outside while leaving substantial delivery work underneath.

The next bet is distribution

Ema plans to use a significant share of the new funding to build its go-to-market operations, especially sales and marketing. For its first years, the company focused mainly on product development.

Ema has nearly 200 employees, with offices in Bengaluru, London and Vancouver. It has so far focused mainly on customers in the United States and Europe, and plans to enter more markets over the next year, particularly in the Asia-Pacific region, South America and parts of the Middle East.

The rise of larger AI companies may help that expansion rather than threaten it. Anthropic is pushing Claude into core enterprise functions such as finance and legal work, while OpenAI has formed engineering teams that work directly with customers on production deployments. Chatterjee does not view those labs as direct competitors because Ema can use more than 150 models, including frontier and open models, while concentrating on domain knowledge, integrations and end-to-end coordination.

That makes Ema’s business less dependent on winning the model race than on turning better models into reliable enterprise workflows. Its $77 million round is therefore a bet on distribution and execution: the more AI improves, the more valuable Ema’s coordination layer may become—but also the more directly the large labs can enter the same corporate budget.