Debjani Deb’s case for leaving the shopper alone

The customer is already reaching for the checkout. The retailer, eager to help, offers a discount. A sale follows, everyone applauds, and somewhere in the accounts a little margin disappears. The coupon has received credit for a decision it may never have influenced. Debjani Deb has spent years asking businesses to look more closely at moments like this.

It is an oddly restrained proposition for an artificial-intelligence company: work out whether an intervention is useful before making one. Session AI, which Deb co-founded and now serves as President, builds software around that judgment. Its current pitch distinguishes shoppers likely to buy, shoppers an offer might persuade, and visitors who are not in the market. A discount need not be the greeting at every door.

The thought has a long history in her working life. Before the software could respond to a shopper’s behavior, another company she co-founded helped businesses listen to what people were saying. Between the two sits a career concerned with the distance between information and a decision.

Before the checkout, the conversation

Deb’s professional background crosses several worlds that technology companies usually put in separate departments. She worked at AT&T Bell Laboratories, went into strategy consulting at Booz Allen Hamilton, and held strategy roles at Mayan Networks and ONI Systems. She also served as worldwide chief strategy officer at GCI, a WPP company. Research, enterprise technology and communications all belong in the account.

Her education likewise combines technical work and business: computer science at Rutgers, a master’s in computer science from Stanford, and an MBA from MIT Sloan. Those credentials describe the route. The more revealing continuity is the kind of problem she kept choosing: how an organization turns what it knows into something it can use.

EmPower Research gave that question a particular setting. Social media was producing conversations that brands wanted to understand. The company’s work included monitoring media, measuring it, tracking brands and researching the effects of events. Consumer perceptions were becoming business material, provided someone could extract meaning from the accumulating chatter.

The service could be quite hands-on. An EmPower partnership with the social-analytics company NetBase let customers work with trained analysts to get their listening programs started. Buying access to a dashboard did not automatically supply the judgment needed to use it. Software could collect the conversation; people still had to decide what mattered.

In September 2011, Genpact announced its agreement to acquire EmPower. The company had approximately 360 employees across locations including New York, Bangalore, San Francisco and London. Genpact completed the acquisition on October 3. Deb had helped build a business substantial enough to become part of a larger analytics operation. The next venture would shorten the interval between a signal and a response.

Three founders, one library

ZineOne began in March 2014 with Deb, Arnab Mukherjee and Manish Malhotra. Their early planning happened at a public library, where they whiteboarded the platform they wanted to build. For a company concerned with immediacy, the setting has a pleasing lack of fuss. A useful idea can survive without a reception desk.

The work itself required patience. Deb later described two and a half years building the enterprise product with no incoming revenue. She had already experienced an exit, but the next company still needed time, money and customers willing to trust a new platform. Experience could inform the bet; it could hardly make the engineering arrive by return post.

The original ambition reached beyond the online shopping basket. Brands needed to respond as consumers moved between digital services and physical places. Mukherjee brought experience from call-center software. Deb and Malhotra had both built social-media businesses. Together, they were looking at customers whose expectations had moved faster than many companies’ systems.

“The fundamentals are revenue, profit, margins, product, people.”

Debjani Deb, 2019

The clock becomes the product

Consider an example Deb used in 2020: a person walks into a mall with a retailer’s app on their phone. There is a sale, and the shopper has reward points available. A timely message could connect the two. The same information delivered after the visit has much less to recommend it, however immaculate the formatting.

In that illustration, entering a location triggered a check for useful information. The aim was to surface it selectively. Deb’s explanation included a warning about spamming the user. Even when a system can send a message, it still needs a reason to interrupt.

The example also shows how her work has changed over time. That earlier engagement platform discussed location, loyalty information and multiple industries. The more recent Session AI proposition concentrates on interpreting a visit’s behavior, including visits where the shopper’s identity is unknown. The recurring interest is context; the information available to establish it can differ.

In September 2022, the company announced a $28 million Series C led by SignalFire, with participation from Norwest and other existing investors. Deb described plans to improve the platform and expand functions including data science, engineering, sales and customer success. A round of capital meant more capacity to build and deliver the system.

In March 2023, ZineOne became Session AI. The name put the company’s unit of attention on the sign outside: the session, that bounded interval when a visitor is actually present. It also made the business easier to explain. The useful question was what a retailer could do while there was still someone there to respond.

The economics of an unnecessary favor

By 2024, Deb was writing directly about a problem that can hide inside encouraging sales figures: promotions may help revenue while eroding profit. Retailers need competitive offers, yet a sitewide discount gives the same concession to customers with very different reasons for visiting.

Her proposed response was an onsite system that could distribute incentives more selectively. Current behavior could help determine whether a visitor needed an offer, assistance or no promotional encouragement at all. She described the work in unusually personal terms: “Building this engine is what drives me.”

There is a useful distinction here between persuading someone and rewarding a purchase that would have happened anyway. Both can appear as a completed order. Only careful measurement can establish what the promotion contributed. Deb’s argument asks retailers to care about the difference before celebrating the transaction.

In her writing on customer identity, she has also questioned how far a stored profile can take a brand. People use multiple screens and arrive with changing intentions. What someone did previously offers context, but it does not resolve every decision about the present visit. The website needs to notice the person currently using it.

The people behind the prediction

Building a company around those decisions also requires decisions about people. Deb has spoken about growing up in India, moving to the United States at a young age, and reconsidering hierarchical ideas of leadership. She favors motivating people to pursue a shared purpose and taking care over who joins the team.

One small detail makes the philosophy tangible. In 2021, she recalled sending her leadership team Trillion Dollar Coach, the book about Bill Campbell. Coaching interested her because the leader’s work changes as a company grows. Habits suited to five people need revisiting when the organization becomes much larger.

Her caution about hiring follows from that experience. Rapid growth can create pressure to recruit quickly; a poor appointment creates costs of its own. There is a recognizable connection to the product she describes: the ability to act quickly becomes more valuable when someone has first taken care over the decision.

The public recognition arrived alongside the building work. Deb was named among The Software Report’s Top 50 Women Leaders in SaaS for 2020 and selected as a Silicon Valley Business Journal Woman of Influence in 2021. In March 2021, women made up more than 40 percent of the company’s management team. The dated detail gives her leadership record some substance beyond an individual award.

In 2023, Session AI announced new leaders in finance, marketing and customer delivery. Enterprise software needs all three. A model must be sold, implemented and supported inside another organization, often by people who have demanding responsibilities of their own. The company’s work extends well beyond making a prediction.

A quieter kind of ambition

Deb’s 2024 observations from Shoptalk returned to selection. Retailers were discussing more carefully curated merchandise assortments, and she connected that approach with choosing incentives more thoughtfully. A shopper can face too many products and too many promotions. Attention places a limit on both.

The following year, the CommerceNext Growth Show program paired her with Minted marketing leader Suruchi Shukla for a session about profitable conversions through behavioral AI. That pairing brought the conversation back to an operating business and a concrete question: how does a retailer grow without casually giving away the economics of the sale?

Today, Deb is President and a board member at Session AI, where Sanjeev Kulkarni is CEO. Her company has acquired a new name and her responsibilities a new title. The thread through the work remains fairly consistent: understand the situation closely enough to make a useful choice.

For the shopper at the checkout, that choice may be almost invisible. No extra coupon. No unnecessary interruption. Just the purchase they were already prepared to make. It is an undramatic ending for an AI demonstration, and a rather satisfactory one for the margin.