Ishan Chhabra has a tidy name for the untidy middle of company-building: the startup desert. His version lasted years. There was a time-management tool for sales teams that buyers could replace with Google Calendar. Then a meeting-guidance system that earned the faintly murderous compliment every founder dreads: neat. The products worked. The urgency did not. Chhabra, an engineer trained to make systems function, had run into a problem that refused to be debugged in code.
His eventual fix was to reverse the work. Instead of starting with product and hoping to discover a market, he would define the buyer, test the pain, create demand and ask for a commercial commitment. Product came last. It sounds obvious when compressed into a sentence. Chhabra says it took him six years to learn.
That admission makes the story of Oliv AI more useful than the familiar tale of a founder spotting a perfect opportunity. Oliv was not found whole. It was argued with, taken apart and returned to the customer in a different order. Today the San Francisco company builds AI agents for revenue teams: software that captures deal context, updates CRM records, prepares sellers for meetings, detects risk and helps managers forecast. The line from the first calendar idea to that platform is crooked, but it has a consistent subject: what salespeople do with their time, and how much of it software should give back.
Before the pitch deck, the research lab
Chhabra graduated from the Indian Institute of Technology Ropar in 2012 with a computer science and engineering degree. His early work already moved between abstract models and stubborn real-world systems. At Aston University, he explored ways to visualize web search results, using phrase extraction and dimensionality reduction to make a crowded information space intelligible. At Bell Laboratories, his research touched privacy-preserving personalization and the computation of user similarity.
Then came Rocket Fuel, the advertising-technology company whose systems made decisions in the tiny interval between an ad request and the response. From 2012 to 2017, Chhabra climbed from software engineer through management roles to director of engineering. The titles chart a change in scale: first make the machine work, then make the organization that makes the machine work.
This is the useful prehistory of Oliv. Search visualization asks how to surface meaning from too much information. Privacy research asks what a system may learn without becoming reckless. Real-time advertising asks whether software can act quickly across a noisy stream of signals. Revenue operations has its own versions of all three questions. A sales organization produces calls, emails, notes, calendar events, CRM fields and off-record recollections. Somewhere inside that pile is the condition of the deal.
“I just love to experiment.”Ishan Chhabra, describing his “chief mad scientist” instinct
Chhabra has worn the title “Chief Mad Scientist & Reluctant CEO,” an excellent bit of self-diagnosis. He once joked that too much experimenting could produce a Frankenstein. The label grants permission to try. The word reluctant admits that a company eventually asks for more than invention. It asks for choices, especially choices about what not to build.
Three products and an uncomfortable mirror
Chhabra co-founded the company now known as Oliv with Venkata Deepankar Duvvuru in 2018. He has written about the early product history with unusual specificity. In 2019 and 2020, the team worked on time management for salespeople, helping them prioritize customer meetings. Chhabra later called the attempt a flop. Buyers did not rank the problem as urgent, and their existing calendars were adequate.
The failure did produce something valuable: conversations with sales teams. Those conversations suggested that representatives struggled to run meetings well. From 2021 through 2022, Oliv became a meeting-guidance system intended to help sellers before, during and after a call. Chhabra estimated that this version had moved from no validation to perhaps 20 percent. People liked it. They could also imagine surviving without it.
The third era began in 2023. Oliv pitched and marketed an AI-powered deal-intelligence concept before pouring in development time. The point was to hear more than praise. Chhabra wanted the sharper signal of willingness to pay. In April that year, the company's legal entity, Instaworks, Inc., reported an offering just under $5 million. Foundation Capital led the documented funding, with Spider Capital also listed among Oliv's backers.
Chhabra's public post about the journey gives founders a clean sequence: product marketing, demand generation, sales, product. It is less a doctrine against building than a rule for earning the right to build. The engineer still gets to engineer. First, the buyer must make the problem expensive enough.
From a copilot to a crew
In May 2024, Oliv presented an AI Copilot built to reduce sales administration and help teams move deals forward. Chhabra's complaint was not that sellers lacked talent. It was that redundant tasks consumed the hours in which that talent should operate. The product used purpose-built models to turn sales conversations into preparation, updates and action.
The company has since widened the frame. Its current platform describes specialized agents with legible jobs: CRM Manager, Deal Driver, Forecaster, Coach, Researcher and Meeting Assistant. They collect context from calls, emails and other systems, map it to accounts and opportunities, and return work to the team through familiar channels. The ambition is execution without another dashboard pilgrimage.
One small example reveals Chhabra's attention to adoption. In a recent public post, he argued that a vendor can make its agents harder to use simply by giving every function a human name. A rep should not need to remember whether Aria enriches accounts while Sage forecasts. Keep a personality where the daily interaction benefits from one; name the machinery by the work it performs. The interface should reduce memory, not recruit it.
The machine Chhabra wants to make quiet
A thinner CRM, a richer memory
The larger thesis is aimed at the CRM itself. Traditional systems preserve a compact human-entered snapshot: stage, close date, value, key contacts and whichever custom fields an operations team can persuade sellers to maintain. Chhabra argues that agents need a different kind of memory. They need the rationale behind a proposal, the appearance of a new stakeholder, the objection buried in an email and the decision trail that links a sale to onboarding and renewal.
In this view, the important record is not merely the object. It is the evolving context around the object, written and consumed continuously by software. Oliv's 2026 writing calls this cross-channel deal intelligence: a story stitched from calls, email, Slack, support conversations and unrecorded updates. The Deal Driver can flag a risk; the Forecaster can rebuild a pipeline view; a CRM agent can propose the field change. Human approval remains part of the described workflow.
Chhabra has said that AI matters when it disappears. The phrase is useful because it refuses the usual theater. A system earns its place when a seller walks into a meeting prepared, leaves without an evening of clerical repair, and notices the regained attention more than the mechanism. The software becomes infrastructure. Its success is measured partly by how little ceremony it requires.
“Six years in, I finally learned to build ‘backward.’”Ishan Chhabra on Oliv's market-first reset
The experiment gets a boundary
There is a pleasing tension in Chhabra's story. The experimenter remains in charge, but the experiments now face a commercial gate. The researcher who once looked for structure in web results is still looking for structure in messy information. The engineering leader who operated fast decision systems is still trying to make software act on live signals. What changed is the order in which the hypothesis meets the market.
He is candid enough to make the failures part of the operating manual. He has even tied company history to domestic comedy, joking that founding Oliv and getting married around the same time meant signing two life-altering contracts back-to-back. In his telling, his wife earned the right to tease him about products nobody wanted. The joke lands because the lesson cost years.
Oliv's next wager is that revenue software will move from passive records and dashboards toward agents that own defined responsibilities. Chhabra calls the emerging category AI-native revenue orchestration. The aspiration is broad, but his best rule is narrow: do not confuse the pleasure of making with evidence that someone needs the thing made.
For an engineer, putting product last can sound almost indecent. For Chhabra, it became a form of respect - for the code, for the team writing it, and for the buyer who must live with the result. The mad scientist did not stop experimenting. He finally gave the laboratory a door, and taught the market to knock before he opened it.