Surojit Chatterjee, founder of Ema, says he knows product-market fit has arrived when his average salesperson can close deals without him in the room.
Ema's breakthrough came with Hitachi, where deploying to production HR data, not signing the contract, was the milestone that counted.
Ema's AI employees, coordinated teams of agents, cut Wipro's HR staff from about 1,000 to 550, a 45% reduction, while speeding requests from days to minutes.
AI employees are making custom enterprise software cheaper than SaaS
Ema bets that enterprise software is moving from off-the-shelf SaaS to custom teams of AI agents. AI makes bespoke work cheaper and faster, so agents can take over the glue work between systems in HR and finance. Hitachi and Wipro are the proof, and the sales cycle is long but the contracts are large.
Takeaways
Custom enterprise software is becoming cheaper than SaaS. Chatterjee compares ready-made software to off-the-rack shirts, which are cheap and fast but never fit exactly. Custom builds once cost tens of millions and needed firms like IBM. AI now cuts that cost and time, so some SaaS will die and others will evolve.
Ema sells teams of agents as AI employees. Each AI employee is a team of forty to a hundred agents run by Ema's orchestrator and control plane. They operate tools such as Workday, Greenhouse, ServiceNow and Salesforce, and take over the human glue between systems. Ema is horizontal but goes deep in HR and finance operations.
Wipro shows the labor impact in HR. Wipro has about a quarter million employees in 65 countries. HR requests that took five to seven days now take seconds to minutes. The HR team fell from about 1,000 people to 550, a 45% reduction.
Production deployment, not a signed contract, marked the breakthrough. Hitachi wanted HR operations automated across five companies and about 55,000 employees through a shared service. Chatterjee says touching production data, not a pilot, was the moment that counted. His PMF test is that an average salesperson can sell without him.
Early versions failed before HR and finance worked. Chatterjee left Coinbase in early 2023 and treated LLMs as a platform that can plan, use tools and remember. An analytics product for spreadsheet clutter failed. The team moved to HR, where onboarding can touch twenty or thirty systems.
Most work is routine, which is the opening. Chatterjee says about 90% of enterprise work is updating things and creating reports, and about 60% of talented people's time just keeps the lights on. He saw this at Google and Coinbase, and it is the work agents can absorb.
Buyers mostly want efficiency, not layoffs. Chatterjee says most enterprises want efficiency and relief for overburdened staff, citing nurses who spend half their time on paperwork. He expects customer support to see a more radical shift than other functions.
How Ema sold into large enterprises
Showed early prospects demos on proxy data before touching customer data, then ran contained proofs of concept.
Built compliance first with SOC 2, ISO 42001, GDPR and HIPAA. The product runs on any cloud or air-gapped in a private cloud, with a governance layer that redacts PII and PHI and traces every agent decision.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Stayed in stealth for over a year with no website. Customers came through relationships, and one of the largest partners arrived by cold email.
Earned analyst coverage from firms such as Gartner and HFS Research. Everest Group has placed Ema in its leaders bracket for three years, and Harvard Business Review has run articles.
Hosts small dinners for CHROs and similar leaders instead of 30,000-person conferences, where C-level buyers are hard to reach.
Expects about nine months for very large seven-figure, multi-year deals, and partners with PricewaterhouseCoopers and KPMG.