Ema raises $77 million to push AI agents deeper into enterprise work

Ema, the startup building AI-agent systems for corporate workflows, has raised $77 million in fresh funding as it tries to capture more of the work now split across enterprise software, consulting and IT services. The Series B round underscores how quickly buyers are shifting budget toward AI tools that can automate end-to-end business processes rather than just answer questions or draft text.

The financing, announced on September 23, 2026, was led by Bengaluru-based Creaegis, with existing backers Accel, Section 32 and Prosus also increasing their commitments. Ema said the round was all primary equity and lifted its total funding to $140 million, while its valuation has more than quadrupled since the company’s previous round in 2024. The startup did not disclose the new valuation.

The deal lands at a moment when major AI labs and established software vendors are all competing for the same enterprise spend. For Ema, that backdrop is not a warning sign but the market thesis: if AI can carry out business tasks across HR, finance and IT, companies may buy less software and fewer services over time.

What Ema is building and why it matters

Ema was founded in 2023 by former Google and Coinbase executive Surojit Chatterjee and ex-Okta executive Souvik Sen. The company sells what it describes as “AI employees,” systems that coordinate multiple AI agents to complete business processes across an organization’s existing tools.

That distinction is important. Rather than focusing on one narrow action, such as summarizing a document or answering a customer query, Ema’s platform is built to orchestrate multi-step workflows. That can mean handling a request from start to finish across systems such as HR platforms, finance software, ticketing systems and internal knowledge bases.

Chatterjee argues that this approach eventually reduces dependence on traditional SaaS products. In his view, Ema first sits on top of an enterprise’s current stack, then gradually makes some of those applications less central — and, in certain cases, unnecessary.

“Many of our customers are already on the way to replace [large SaaS applications] completely,” Chatterjee said, describing those systems as increasingly looking more like databases than full business applications.

That is a bold claim, but it captures the broader shift underway in enterprise software. As AI agents become capable of taking action across systems, the value proposition is moving away from standalone interfaces and toward automation layers that can operate in the background.

How the funding round is structured

The $77 million Series B consisted entirely of primary capital, meaning the funds go directly to the company rather than to early shareholders through secondary sales. Ema confirmed there was no debt in the transaction.

That structure suggests the company is still in growth mode and intends to keep building rather than simply providing liquidity to insiders. According to Ema, the latest capital will primarily support go-to-market expansion, with sales and marketing set to receive the biggest share after an early period focused on product development.

The company now has nearly 200 employees and operates from Mountain View, with additional offices in Bengaluru, London and Vancouver.

Why enterprise AI is becoming a bigger battleground

Ema’s funding arrives as AI companies increasingly target enterprise software budgets that have historically gone to SaaS vendors and consulting firms. The competition is intensifying on several fronts at once.

Large AI labs want their models embedded directly into corporate workflows. Software vendors want to add agentic features to protect existing products. Startups such as Ema are trying to own the orchestration layer that sits between models and the business processes they automate.

Anthropic has been expanding its push into enterprise operations, including financial and legal work through Claude. OpenAI has also built out teams of forward-deployed engineers who work closely with customers to get AI into production environments. Ema does not see that as a threat so much as market validation.

Chatterjee said the company benefits from progress in frontier models, adding that more capable models make Ema’s platform more useful rather than less relevant.

According to the startup, its systems can draw on more than 150 models, including both frontier and open-source options. Ema says its competitive edge is not tied to any single model, but to the domain expertise, integrations and orchestration needed to make those models useful for enterprise operations.

How Ema says it differs from model providers

As AI labs race to improve model quality, Ema is betting on a different layer of the stack. The startup’s argument is that enterprises do not need a model alone; they need a system that can reliably complete work inside complex companies.

That means connecting to existing software, understanding business rules, handling approvals and shifting from one step to the next without human handoffs whenever possible. In other words, Ema is selling the operational plumbing that turns AI from a chatbot into something closer to an automated digital worker.

What the company calls “AI employees”

Ema’s branding language reflects that ambition. The startup says its “AI employees” coordinate multiple AI agents to carry out tasks across functions such as HR, IT and finance. The emphasis is on end-to-end execution, not isolated prompts.

For example, a system might help resolve an employee request, pull information from multiple internal systems, trigger follow-up tasks and track completion. The appeal to enterprises is obvious: fewer repetitive tasks for staff, faster turnaround times and less reliance on manual coordination.

Why customers may be receptive

Enterprises are under pressure to do more with fewer resources, and many are already experimenting with AI in back-office functions. Ema’s pitch goes further by suggesting that AI can become the operating layer for those workflows.

The company says customers are already moving beyond trial use cases. More than 90% of its customers, it claims, have expanded after the initial deployment, with some using the platform across dozens of workflows.

Ema also says its net dollar retention is around 180%, indicating that existing customers are spending materially more over time. In subscription software, that is generally a sign that the product is expanding within accounts rather than stalling after initial adoption.

How strong is Ema’s traction?

Ema is presenting several growth metrics to support its case that it is becoming an important enterprise platform. The startup says it has more than 50 active enterprise deals, over 1 million active enterprise users and more than 5 million actions and queries handled so far.

Its customer list includes major names such as NTT DATA, Hitachi, ADP, PwC, Google, KPMG, Wipro and Microsoft. Those logos suggest that Ema is not pitching only to early-stage AI adopters, but to large organizations with existing technology complexity and substantial workflow overhead.

Over the past two years, the startup says revenue has increased 50-fold, while bookings have topped $150 million. Chatterjee stressed that the bookings figure refers to the total value of multiyear contracts, including two- and three-year agreements, rather than annual recurring revenue.

Chatterjee declined to share the company’s current annualized revenue run rate, but said the growth in bookings reflects broader enterprise adoption.

The distinction matters for investors and customers alike. Bookings can be an encouraging indicator of future revenue, but they are not the same as realized annual recurring revenue, which is often the standard measure of software business health.

What Ema’s pricing model signals about the market

Ema is also experimenting with a pricing approach that differs from the standard software-seat model. Instead of charging by user licenses or by tokens consumed, the company says it prices around task completion and business outcomes.

That model is strategically important. If AI systems are meant to do work rather than merely support workers, then pricing based on work completed aligns more naturally with the value customers receive. It also mirrors the shift from selling software access to selling automation results.

For buyers, outcome-based pricing can be attractive because it links cost to measurable output. For vendors, it creates pressure to deliver reliable automation, since billing is tied more closely to performance than to usage volume.

How services firms fit into Ema’s strategy

Ema’s ambitions extend beyond software replacement. Chatterjee said AI can also absorb some of the implementation, integration and consulting work that companies normally pay IT services providers to handle around enterprise software.

That is an especially consequential claim for large services firms, which have long profited from deploying and customizing business systems for corporate customers. If AI can streamline those projects, then the economic model of enterprise implementation could change just as much as the software layer itself.

Chatterjee said many services companies are already working with Ema and that they recognize the need to adapt as human-centered service delivery becomes less central.

This is where Ema’s story becomes bigger than a single startup. It points to a broader restructuring of enterprise technology spending, where AI tools do not merely sit inside existing systems but also reshape how software is bought, deployed and supported.

Why margins matter in an AI-agent business

Despite handling work that usually involves human support, Ema says it has maintained gross margins of close to 80%. That is notable in a category where some companies struggle with heavy implementation costs or labor-intensive customer support.

Chatterjee said margins improve as the company’s systems learn from deployments and require less human intervention over time. In theory, that means each additional customer can become more efficient to serve once workflows and integrations are established.

Margin performance will be watched closely because enterprise automation often looks promising on paper but becomes expensive in practice if every deployment requires extensive customization. Ema is arguing that its platform becomes more scalable as it matures.

What comes next for Ema?

The new capital will support a bigger sales push and a wider geographic footprint. Ema has so far focused mainly on customers in the United States and Europe, but it plans to expand into Asia-Pacific, South America and parts of the Middle East over the next year.

That expansion strategy makes sense given the company’s leadership, investor base and existing office network. It also reflects the reality that enterprise AI demand is becoming global, especially in regions where large employers want to modernize operations without hiring proportionally more staff.

The startup’s next challenge will be proving that its early traction translates into durable category leadership. In enterprise software, the jump from pilot programs to widespread adoption is often where many vendors stall. Ema is betting that agentic automation will be different because it can deliver direct operational savings and eventually replace multiple layers of software and services.

Timeline of Ema’s rise

Why investors are paying attention now

Ema’s fundraising is part of a larger wave of capital flowing into companies that can show practical enterprise AI adoption rather than just model development. Investors are increasingly looking for businesses that can turn generic AI capability into specific outcomes inside organizations.

The startup’s appeal lies in several intersecting trends: rising enterprise AI budgets, pressure on SaaS spend, automation demand in finance and HR, and the desire among large companies to integrate AI without rebuilding their tech stacks from scratch.

With a large customer base, rapid reported growth and a pitch that spans software and services, Ema is positioning itself as more than another AI wrapper. It is trying to become an operating layer for business work.

Whether that vision translates into long-term market leadership will depend on execution, product reliability and the pace at which corporate buyers are willing to let AI systems take on real operational responsibilities. But the size of the new round makes one thing clear: investors are willing to fund the bet.