Ema raises $77M to replace SaaS busywork with teams of AI agents

Ema, a startup that runs teams of AI agents inside large companies' HR, IT and finance operations, has raised $77 million in a Series B round. The round is all primary equity with no debt. It was led by Bengaluru-based Creaegis, and existing backers Accel, Section 32 and Prosus increased their stakes.

The new money takes Ema's total funding to $140 million.

What Ema does

Ema doesn't sell another app for employees to learn. It deploys groups of AI agents that work through multi-step business processes across the software a company already has, such as approving a leave request, resolving an IT ticket or reconciling an invoice.

That pitch puts Ema in direct competition with the per-seat SaaS model. If agents can do the work across existing systems, companies may need fewer specialist tools and fewer seats in them.

Co-founder and CEO Surojit Chatterjee, formerly an executive at Google and Coinbase, started the company in 2023 with Souvik Sen, previously at Okta. Ema is headquartered in Mountain View, has offices in Bengaluru, London and Vancouver, and has close to 200 employees.

The numbers Ema is sharing

The customers it names include NTT DATA, Hitachi, ADP, PwC, Google, KPMG, Wipro and Microsoft.

Why frontier models don't worry Ema

A common question for agent startups is what happens when the big model labs ship agents of their own. Chatterjee's answer is that better models help Ema. The company draws on more than 150 models and competes on domain knowledge and on orchestration, meaning how agents are chained together, checked and connected to enterprise systems.

What the money is for

Ema plans to scale sales and marketing and to expand into Asia-Pacific, South America and the Middle East.

Why it matters for SaaS buyers

Ema's growth numbers are another sign that enterprise budgets are moving from "software people use" to "software that does the work". Before renewing a stack of point tools, it's worth asking which of those workflows an agent platform could run end to end.