The pitch is replacement, not augmentation
Most enterprise AI agent vendors position themselves as a productivity layer sitting on top of existing software, helping employees work faster inside the tools they already have. Ema's CEO Surojit Chatterjee is explicit that this is not what is actually happening with his customers. Many customers, he says, are already replacing large SaaS applications completely, removing dependency on the underlying software until it becomes mostly like a database that Ema's agents read from and write to.
That is a meaningfully more aggressive claim than the standard agent-as-copilot pitch, and it is the kind of claim that should get more scrutiny than most vendor statements, because if true it implies a direct and near-term threat to incumbent SaaS revenue rather than a complementary tool. The fact that Ema's own investors include companies it also counts as customers, Google and Microsoft among them, adds a layer of alignment that cuts both ways: it lends credibility to the adoption claims, but it also means those companies have their own strategic reasons to want this narrative to be true.
The numbers behind the pitch
Strip away the framing and the underlying metrics are genuinely strong for a company at this funding stage, strong enough to draw scrutiny on their own merits regardless of the SaaS-replacement narrative wrapped around them by the company itself. Revenue has grown fifty-fold over the past two years, bookings have surpassed $150 million in multiyear contract value, and net dollar retention sits at 180 percent, meaning existing customers are expanding their spend well beyond initial contract value rather than merely renewing at the same level year over year. Gross margins near 80 percent are healthy for a company whose product involves running compute-intensive agent workflows rather than pure software margins, and suggest the compute cost of running those workflows has not eaten into the business model the way early skeptics of agent-based products widely expected it eventually would.
The round itself was structured entirely as primary equity, no debt, no secondary sales, with Creaegis leading and Accel, Section 32, and Prosus all increasing their existing stakes rather than new investors coming in cold off a hot headline. That pattern, existing backers doubling down rather than new capital discovering the story for the first time, is generally a more reliable signal of genuine conviction than a round dominated by first-time investors chasing whatever category happens to be fashionable that quarter.
Why HR, IT, and finance specifically
Ema's chosen beachhead, HR, IT, and finance departments, is a deliberate choice rather than an accident of where early customers happened to come from. These are functions built around well-defined, repeatable multi-step processes, onboarding, ticket resolution, reconciliation, that are exactly the kind of workflow AI agents currently handle most reliably, as opposed to open-ended creative or strategic work where agent reliability remains inconsistent even in the most capable current models.
It is also where the SaaS replacement thesis is most plausible in the near term, for structural reasons that go beyond agent capability alone. HR information systems, IT service management tools, and finance close software are frequently licensed per-seat for functionality that a well-orchestrated set of agents can approximate without needing the full underlying application, particularly for mid-market companies that never fully used the more sophisticated features they were paying for in the first place and were effectively subsidizing enterprise-grade complexity they did not need.
Frontier model progress helps Ema more than it might seem
Chatterjee made a point of noting that progress in frontier models is actually very beneficial to Ema's business, which is a more interesting claim than it first appears given how the category usually talks about itself in fundraising pitches aimed at wary investors. Companies building narrow point-solution AI products sometimes worry that frontier lab progress will eventually swallow their feature set entirely as models absorb more of the stack. Ema's bet runs the other direction: as underlying models get more capable and reliable at multi-step reasoning, the orchestration and enterprise integration layer Ema has built becomes more valuable rather than less, because the hard part of enterprise AI deployment increasingly is integration and workflow design rather than raw model capability itself, a gap models alone do not close on their own.
That is a reasonable read of where the actual bottleneck in enterprise AI adoption currently sits, and it runs against a lot of conventional wisdom in the space. Most enterprises are not capability-constrained by the underlying models available to them today; they are constrained by the unglamorous work of connecting those models reliably to internal systems, proprietary data, and approval workflows that vary by department. If that assessment holds, orchestration-layer companies like Ema benefit disproportionately from every frontier model release going forward rather than being threatened by it the way a thinner point solution might be.
What CIOs should take from the customer list
The presence of PwC, KPMG, and Wipro as both customers and, in some cases, implicit distribution partners matters more than the funding number itself for enterprise buyers trying to gauge whether this category is real. Professional services firms adopting a vendor's tooling for their own internal operations, and then plausibly recommending similar approaches to their own enterprise clients, is a distribution pattern that can scale adoption faster than direct sales alone.
The practical question for a CIO reading this is not whether to trust Ema's specific numbers, which are self-reported and unaudited, but whether the underlying trend, agents displacing rather than merely augmenting narrow SaaS applications, is showing up in their own vendor landscape yet. If procurement and IT teams are not already fielding this exact pitch from at least one vendor in HR, IT, or finance tooling, they should expect to within the next renewal cycle.
