Ema raises $77M as AI starts eating into enterprise software and services

Ema, a Mountain View startup that coordinates teams of AI agents across HR, IT, and finance workflows, raised $77 million in a Series B led by Bengaluru’s Creaegis, with Accel, Section 32, and Prosus increasing their stakes. Total funding hits $140 million; the company says the round more than quadruples its 2024 valuation, though it declined to name the new number.

The money arrives as AI startups, frontier labs, and incumbent software vendors fight over budgets that used to go to enterprise SaaS and IT services. Ema’s pitch is blunt: wrap existing apps first, then help customers treat those apps more like databases — and, in some cases, replace them.

We’re watching the enterprise stack get renegotiated in public. Founders Surojit Chatterjee (ex-Google, Coinbase) and Souvik Sen (ex-Okta) are selling “AI employees” that orchestrate multi-step processes across a company’s existing tools rather than one-shot chatbots. Chatterjee told TechCrunch that many customers are already on a path to remove dependency on large SaaS applications. That is the thesis investors just paid up for.

The traction numbers, if they hold under diligence, are loud: more than 50 active enterprise deals, over a million active enterprise users, more than five million actions and queries handled, customers including NTT DATA, Hitachi, ADP, PwC, Google, KPMG, Wipro, and Microsoft. Revenue up fifty-fold over two years; bookings past $150 million — with the important caveat that bookings mean multi-year contract value, not ARR. Net dollar retention around 180% and over 90% of customers expanding past the first use case suggest the land-and-expand motion is real.

Useful AI earns credit here when it actually completes tasks and prices on outcomes instead of seats or tokens. Ema claims gross margins near 80% and less human support as systems learn from deployments. If true, that is the services industry’s nightmare and the buyer’s leverage: implementation and consulting work that used to bill humans gets absorbed into the product. Chatterjee even says services firms are partnering while rewriting their own models.

The risk side is just as clear. Replacing SaaS with agent orchestration concentrates operational dependency on a young vendor’s reliability, security, and model routing across 150-plus models. Frontier labs pushing Claude and forward-deployed OpenAI engineers into the same enterprises means Ema’s complementary story only lasts until a lab ships good enough orchestration. Bookings that include two- and three-year deals can look richer than cash collected. And “AI employees” language can paper over accountability when an automated HR or finance path fails.

I’m watching whether Ema discloses ARR alongside bookings as it spends this round on sales and marketing, whether Asia-Pacific and Middle East expansion repeats the U.S./Europe land pattern, and whether large customers actually churn core SaaS or just wrap it. The trajectory if this becomes normal is software priced by completed work — and a thinner middle layer of seat-based SaaS and billable implementation hours.

TechCrunch by Jagmeet Singh, Sep 23, 2026. Series B primary equity only; headcount nearly 200 with offices in Bengaluru, London, and Vancouver; near-term GTM expansion after product-heavy early years.