How Ema Closes 7-Figure Fortune 500 AI Agent Deals

Before 'AI agent' was a common phrase, Surojit Chatterjee was already building them. His company Ema helped one client cut its HR team from 1,000 to 550 people by automating 65,000 monthly job changes. Here's exactly how he got there.

Key Facts

• Hitachi deployment scale: 55,000 employees across 5 companies on multiple cloud platforms (Surojit Chatterjee)
• Wipro HR headcount reduction: 1,000 people to 550 (45% reduction) after automating 65,000 monthly job changes (Surojit Chatterjee)
• Request resolution time: 5–7 days reduced to seconds-to-minutes for employee HR requests (Surojit Chatterjee)
• Deal size: Large six-figure to seven-figure annual, multi-year enterprise contracts (Surojit Chatterjee)
• Stealth period: Over one year building compliance infrastructure before launching a public website (Surojit Chatterjee)

The PMF Moment: Hitachi's 55,000-Employee Deployment

Ema's product-market fit arrived not at contract signing, but when a full production deployment for Hitachi—spanning 55,000 employees across five companies and multiple cloud platforms—went live and worked. That proof of scalability made every subsequent Fortune 500 conversation significantly easier.

For enterprise AI companies, signing a contract is table stakes. The real milestone is surviving production. Surojit Chatterjee is explicit about this: 'When we deployed and started seeing the outcome—it was really the moment I felt we have hit product market fit in terms of the scalability of the product, the governance, security, and the selling to other clients just became that much easier for us.'

Hitachi's setup was deliberately complex—five companies, shared HR services, sensitive employee data, and multiple cloud platforms requiring agent coordination. Passing that test gave Ema something no pitch deck can manufacture: credibility. As Surojit puts it, 'Credibility is everything in an enterprise context.' Ema has since scaled to even larger production deployments, but Hitachi remains the inflection point.

"When we deployed and started seeing the outcome. I mean signing the contract is only the first step in many of the enterprise deployments." — Surojit Chatterjee

"Credibility is everything in an enterprise context." — Surojit Chatterjee

What Ema Actually Does: AI Employees as the Enterprise Glue Layer

Ema builds teams of AI agents—called AI employees—that connect to enterprise systems like Workday, Salesforce, and ServiceNow, understand policies and processes, and automate the manual handoffs between those systems that currently require thousands of human workers.

Every large enterprise runs on two parallel layers: the software stack and the human layer patching it together. Surojit describes the problem directly: 'Every enterprise today has hundreds of applications, SaaS applications, and thousands of people who are acting as glue between those applications—moving data from one, pushing data somewhere else, creating a report, sending an email.'

Ema's platform replaces that glue with coordinated agent teams. Each agent specializes in a capability; Ema's orchestration layer coordinates them across tools and data sources. The platform is horizontal—deployable across HR, finance, and customer support—but goes deep in each vertical. For Wipro, with 250,000 employees across 65 countries, this meant handling everything from vacation policy questions to full relocation workflows spanning tax changes, compensation adjustments, and management changes.

The result at Wipro: 65,000 monthly job changes automated, request resolution time dropping from 5–7 days to seconds, and HR headcount reduced from 1,000 to 550—while employee satisfaction scores increased because remaining HR staff could focus on actual HR work instead of paperwork.

"We are creating that intelligence and action layer across these applications and all this data—which basically frees up humans to do more value added work rather than doing the same similar mundane work every day." — Surojit Chatterjee

"They have sixty five thousand job changes every month. Within the company, people either getting promoted, moving to another role, moving to another location. Can you automate that?" — Surojit Chatterjee

• Connects to Workday, Greenhouse, ServiceNow, Salesforce, SharePoint, databases, and data warehouses
• Handles questions, workflow orchestration, multi-step planning, and cross-system actions
• Deployed horizontally across HR, finance ops, and customer support verticals
• Full execution lineage: every agent decision is traceable with explanation and feedback loops

Building Before Selling: The Compliance-First Strategy

Ema spent over a year in stealth—without even a public website—building compliance certifications and air-gapped architecture before pursuing revenue. This upfront infrastructure investment became the primary trust signal that unlocked Fortune 500 conversations.

Most startups race to revenue. Ema raced to SOC 2. Surojit made a deliberate bet that enterprise trust is earned through infrastructure, not promises: 'We went and got all our certifications compliance done within the first few months—SOC 2 Type one, Type two, ISO 42001, GDPR, HIPAA, anything and everything.' The rationale was simple: you cannot ask a Fortune 500 to open their most sensitive employee and financial data to a startup that can't prove it won't cause a security incident.

Harder still was the architectural decision to support fully air-gapped deployments—where Ema's platform runs entirely within a customer's private cloud, accessing no external services during execution. 'That's very, very hard to do,' Surojit acknowledges. 'Architecturally, a lot of investment is needed and very hard to do later on in your life cycle as a company.' Building it first meant Ema never had to retrofit security into a product designed for speed.

The stealth period also included building a full governance layer—auto-redacting PII and PHI, creating complete execution lineage for every agentic decision, and enabling automatic behavior adjustment from feedback. These aren't features; they're enterprise table stakes that competitors who moved faster to market are now scrambling to add.

"For more than a year we were under stealth. We didn't even build a website. Because we wanted to build the product the right way and spend a lot of time doing all of this infrastructure." — Surojit Chatterjee

"Air-gapped means it cannot access any external service while it's running. It has to be fully contained inside their cloud. That's what they need to know, they need to see to be sure you are not going to create a security issue for them." — Surojit Chatterjee

Enterprise Sales Tactics: Dinners, Partners, and Cold Emails

Ema's most effective enterprise sales motions are small executive dinners, partnerships with major consulting firms like PwC and KPMG, and targeted thought leadership in analyst reports and publications like HBR—not large conferences where C-suite buyers rarely attend.

Counterintuitively, one of Ema's largest partners arrived via cold email—no warm introduction, no conference handshake. Surojit credits consistent visibility through analyst relationships (Gartner, Everest Group, HFS Research) and published thought leadership as the signal that made cold outreach land: 'Creating that credibility is super important—you are an enterprise player, you have built technology ahead of others, you won't steal my data.'

For in-person sales, Ema runs small dinners with CHROs and other executives in targeted cities. The explicit goal is not to sell. 'Our sales process is to not sell,' Surojit explains. 'We are trying to get people to think about agentic business transformation—and through that conversation, we get to know them, build relationship, and then that turns into a commercial relationship over time.'

Large conferences, by contrast, have been largely unproductive. 'If you go to a very large conference, twenty thousand people showing up—you as a small company, probably not noticeable at all. A lot of the attendees are not necessarily decision makers.' Ema's deals require C-level buy-in, and those executives aren't walking conference floors. Partnering with PwC and KPMG became a faster wedge—these firms already advise C-suites on transformation and carry the credibility Ema was still building.

"One of our largest partners today came through a cold email." — Surojit Chatterjee

"Our sales process is to not sell. We are trying to get people to think about agentic business transformation and through that conversation, we get to know them, build relationship, and then that turns out into a commercial relationship over time." — Surojit Chatterjee

• Small executive dinners with CHROs in targeted cities—focused on transformation, not sales pitches
• Partnerships with PwC and KPMG to reach C-suite buyers through trusted advisors
• Analyst presence in Gartner, Everest Group, and HFS Research to establish category credibility
• Thought leadership in HBR and industry publications as inbound trust signals

The Real Definition of Product Market Fit for Enterprise

For enterprise founders, product-market fit is the moment your average salesperson can close deals without the founder in the room. Until then, you're selling founder credibility, not product value—and that doesn't scale.

Surojit's PMF definition is deceptively simple: 'You have product market fit when your average salesperson can go and sell your product without you being in the room.' For enterprise software, this is a high bar. Early deals often close because a founder's reputation—Coinbase CPO, Google VP—opens doors. Real PMF means the product and its proof points can stand alone.

The Hitachi deployment was the evidence that made this possible. Once Ema could point to a live, production deployment across 55,000 employees with measurable outcomes, any trained salesperson could lead with that proof. The story sells itself. Before that, every deal required Surojit in the room to provide the credibility the product couldn't yet demonstrate independently.

He draws a parallel to his mobile advertising days at Google, when even senior CMOs doubted that anyone would buy a dress on a phone. The early believers who built infrastructure for that future captured disproportionate value. The same dynamic applies to agentic AI—the window for building trust-first, before competitors arrive with shinier demos, is closing fast.

"You have product market fit when your average salesperson can go and sell your product without you being in the room. I am continuously looking at that and saying, can I train my sales team to go and sell without me being in the room?" — Surojit Chatterjee

"I had a discussion, I remember early days, with the CMO of Macy's. I'm like, mobile advertising will change the world. And he was like, I like your enthusiasm, but I don't think anybody will buy a dress on a mobile phone." — Surojit Chatterjee

Ema's Enterprise Sales Tactics: What Works vs. What Doesn't