Last updated:
Ema builds enterprise AI agents: coordinated teams of agents that run entire HR, IT, and finance workflows for large companies, and investors just put $77 million behind the bet that this works at scale. At Wipro, a systems integrator with 240,000 employees across 65 countries, those agents now handle 2.9 million employee queries a year and have cut IT support tickets in half, turning response times that used to take days into a matter of seconds. That is not a pilot. That is a production system carrying real headcount.
The $77 million Series B, announced September 23, 2026 and led by Creaegis, brings Ema’s total funding to $140 million and more than quadruples its valuation from the prior round. For a business owner watching AI agents move from novelty to back-office infrastructure, Ema’s numbers are a useful stress test of how far this has actually gone.
Why do AI agents that close IT tickets matter to your business?
Ema, based in Mountain View and founded by CEO Surojit Chatterjee (formerly Chief Product Officer at Coinbase and before that a senior executive at Google) with Souvik Sen (formerly of Okta), does not sell a chatbot. It sells coordinated teams of AI agents that plug into more than 250 enterprise applications and carry out multi-step processes end to end, across HR, IT, and finance, without a human re-typing the output into another system.
That distinction is the whole pitch. As Chatterjee put it in the company’s funding announcement, “Enterprises do not need more software. They need work to get done.” For a mid-sized company drowning in HR tickets or finance approvals, the practical question is not whether a chatbot can answer a question, but whether an agent can actually close the loop.
How fast can enterprise AI agents actually go live?
Speed of deployment is the detail most funding stories skip, and it is the one that matters most to a business owner. Hitachi, the global industrial and technology conglomerate, took Ema’s agents from concept to production in under four weeks, wiring them into more than 20 internal systems. Ema reports a 70% gain in employee efficiency and 30% fewer support tickets there.
A separate customer, a top-five global business process outsourcer, now runs the bulk of its IT service management tickets through Ema’s agents across more than a dozen languages, escalating only a small share of cases to a human. Between Wipro, Hitachi, and that BPO, the common thread is not the technology demo. It is that none of these deployments stayed a pilot for long.
What do Ema’s growth numbers reveal about AI agents in enterprise automation?
The clearest signal in Ema’s business is not the new money, it is what existing customers do afterward. Ema reports a 180% net dollar retention rate, meaning that existing customers, taken together, spend about 80% more each year as they extend agents into new departments. Alongside that, the company has been signing large multiyear contracts and growing revenue 50-fold over the past 24 months, with contracted bookings above $150 million, according to TechCrunch’s reporting on the round.
Chatterjee has gone further, telling TechCrunch that “many customers are already replacing large SaaS applications completely” because those systems are “mostly becoming like a database.” That is a striking claim, and one worth treating with caution: Ema has not publicly named which enterprise software products its customers have actually retired, so the claim is the company’s own framing rather than an independently confirmed fact.
What could go wrong when AI agents run HR, IT, and finance?
Nearly all of Ema’s headline metrics (the retention rate, the revenue growth, the customer results at Wipro and Hitachi) come from the company’s own press release and investor-aligned coverage, not from independent audit. That does not make them false, but it means a buyer should treat them as a starting claim, not a verdict.
The bigger risk sits outside any one vendor. Gartner has warned that more than 40% of agentic AI projects will be cancelled by the end of 2027, citing rising costs, unclear business value and inadequate risk controls. Surveys also suggest that only around one in five enterprises has a mature model for governing what its AI agents are allowed to do. Agents that touch payroll, employee records, and financial approvals are exactly the kind of sensitive, compliance-bound workflows where a governance failure is expensive. Ema says it builds in human approval gates for sensitive actions, but that infrastructure has to be configured and enforced by the buyer, not assumed.
We run our own small company on six AI agents and publish the real numbers. Two things we learned the hard way: an agent’s bill depends more on how often it wakes up than on the model, and before an agent touches anything sensitive you need an answer to who vouches for it and what it is allowed to do.
How should a business start testing AI agents this month?
Do not start with finance approvals or payroll. Start where Hitachi did: pick one high-volume, well-documented workflow, such as IT ticket triage or a single HR request type, and set a four-week deadline to get an agent handling it end to end with a human approval gate on anything irreversible.
That mirrors the actual pattern across Ema’s customers: narrow scope, fast production timeline, then expansion once the agent proves itself, which is what the 180% retention number is really describing. A pilot that takes six months to plan is already behind a competitor that shipped one in four weeks.
Before expanding beyond that first workflow, write down who signs off when the agent gets something wrong, and what happens to the sensitive data it touches along the way. That single decision, made early, is what separates the customers scaling past a pilot from the 40% Gartner expects to be cancelled.
The $77 million is not really the story here. A customer base that keeps spending more, quarter after quarter, on software that does the work instead of just describing it: that is the story, and it will not wait for your department to finish its evaluation.
Frequently Asked Questions About AI Agents in Enterprise Automation
What does Ema actually do?
Ema deploys coordinated teams of AI agents that execute multi-step HR, IT, and finance workflows end to end across more than 250 enterprise application integrations. Unlike a chatbot, its agents are built to take action and complete tasks, not just answer questions, according to the company’s September 2026 funding announcement.
How much did Ema raise and who led the round?
Ema raised a $77 million Series B, announced September 23, 2026, led by Creaegis with participation from existing investors Accel, Section 32, and Prosus. The round brings Ema’s total funding to $140 million and more than quadruples its valuation from the previous round.
What results have Ema’s customers reported?
Wipro reported 2.9 million employee queries handled annually across 240,000 staff in 65 countries, with IT support tickets cut in half. Hitachi went from concept to production in under four weeks across 20-plus systems, reporting a strong efficiency gain and fewer support tickets.
What is the biggest risk with agentic AI in HR, IT, and finance?
Gartner expects more than 40% of agentic AI projects to be cancelled by the end of 2027, because of rising costs, unclear business value and inadequate risk controls. Sensitive HR and finance workflows need clear approval gates and data rules before agents touch them.