Pankaj Kulshreshtha and the bank that couldn’t see him

Pankaj Kulshreshtha went into his bank for a corporate credit card. He had a long relationship with the institution and a business he thought would make the conversation straightforward. The employee kept looking at her screen. Ten days later came a rejection. He turned to an online startup and, he says, received approval in five minutes. Twenty years of banking history had lost to an afternoon on the internet.

Retelling the episode when Scienaptic launched iCUE, he reserved his criticism for the system rather than the employee. Her attention was being claimed by the machinery around her. That distinction matters to the founder: he wants software to give a lender enough context to notice the person sitting opposite. His complaint has the precision of someone who knows how the machinery gets built.

An academic career takes a detour

Before he was a founder, Kulshreshtha was an engineer. His career began at Larsen & Toubro in 1991; his education took him from an engineering degree at Visvesvaraya National Institute of Technology to doctoral work at the Indian Institute of Management Bangalore. The ingredients of his later business were coming together well before there was a company to put them in.

His doctoral thesis concerned fuzzy preference relations, choice functions and rationality conditions. IIM Bangalore lists his doctorate among its 1999 awards. It is a wonderfully demanding preparation for a career spent asking why institutions make the decisions they do. The thesis title would not fit comfortably on a business card. Its central preoccupation fits a lending business rather well: the complicated business of choosing.

In his own account of that period, he expected to become an academic. He later described his research as spanning AI, social choice theory and behavioural economics. The attraction of actual financial data pulled him toward industry. An abstract decision problem acquired customers, balances and consequences. That change of setting would turn out to be a lasting one.

The smaller team that taught him more

At GE, Kulshreshtha built models and then managed people who built them. By 2004 he was leading roughly 250 people. He moved into risk at GE Money in the UK, where the team numbered eight. He regarded that apparent contraction as a valuable move: it brought him close to the consequences of analytics in a working lending business. Headcount had become a particularly unhelpful career compass.

He returned in 2008 to the analytics operation that had become part of Genpact. By October 2013, it had grown past 5,000 people. He was now discussing the discipline of deploying analytics across a business: connecting functions, understanding a process and judging technology by what it changed. The problem had outgrown the question of whether someone could produce an impressive model.

That experience helps explain the kind of founder he became. In a 2013 discussion, his emphasis was already on business processes and the conditions under which analytics could alter them. A good piece of analysis had to travel through an organization before it could become useful. There were departments, handoffs and competing priorities along the route. Anyone selling intelligence would eventually have to deal with all three.

A startup with an unusually quiet office

Scienaptic began in 2014. Kulshreshtha saw banks buying data infrastructure and investing in analytics while still struggling to make customer interactions more intelligent. The early platform, Ether, was designed to work with infrastructure a bank already possessed. Credit decisions, fraud management and customer engagement belonged in the same practical conversation. The bank had accumulated information; the work was to get it into decisions.

Building the company supplied a less mathematical set of problems. In January 2015, the young business had recently moved out of an incubator. A prospective lead engineer, whom Kulshreshtha calls Jo in his account, arrived after several interviews. Everyone else happened to be away. She found the founder alone. He had expected an offer meeting. The scene, by his telling, suggested to her that the previous interviewers might have been actors.

He presents that last thought as his reconstruction, not her recorded verdict. She did not join then, and returned about two and a half years later to explore a role again. It is an unusually funny recruiting story because the founder is the person left looking faintly implausible. An experienced executive, a serious offer, a room with nobody in it: the evidence could support a disappointing conclusion.

A co-founder’s departure also forced a harder decision about continuing. Kulshreshtha has described having to commit to running the business himself. Hiring senior people subsequently became something he approached with greater care. The company’s personnel decisions were part of its survival, every bit as consequential as the quality of its analytics.

An old working relationship becomes a new bet

One connection carried across from the corporate years. Pramod Bhasin, the founder of Genpact, became an investor in Scienaptic. The two had worked together while building the earlier analytics business. For Kulshreshtha, the new venture was a departure from a large organization that still drew on relationships formed inside it. A career can provide both the reason to leave and the people willing to back the departure.

The distinction matters in a field that often presents entrepreneurship as a sudden flash of youthful inspiration. Kulshreshtha arrived with experience of both supplying analysis and owning lending risk. His company’s proposition rested on joining those experiences. He could talk about the technology, then continue the conversation into the workflow where a lender would actually have to use it.

Beware the machine that promises easy money

His writing contains a useful antidote to uncomplicated enthusiasm. In a 2019 essay, he describes a fictional, illustrative lending operation built around an apparently self-improving money machine. It combines lessons from his own and colleagues’ experience. Early lending results invite celebration; losses arrive later. He explicitly labels the narrative illustrative, so its characters and outcomes should stay in the realm where he placed them.

The point of the fable is operational. Response to an offer becomes visible quickly, while credit losses take longer to reveal themselves. Models can interact badly, and changes in the way a business runs can undermine their predictions. His enthusiasm for AI sits alongside a risk manager’s insistence on experience and judgment. The champagne in a lending story may be opened several chapters before the ending.

“But the responsibility remains ours.”

Pankaj Kulshreshtha · September 2026

Getting information across the desk

By 2022, Kulshreshtha was describing a familiar disconnect: financial institutions could possess sophisticated technology without delivering an intuitive customer experience. Records, customer-facing systems and decision tools needed to connect. He also argued for continuing adaptation. A model left alone after deployment could not be expected to keep up with changing customers and conditions. Software needed an ongoing relationship with the business it served.

Those ideas put some substance behind his interest in human attention. A person speaking with a lender should benefit from the institution’s accumulated knowledge. A staff member should be able to use that knowledge without disappearing into a succession of screens. It is an ambition with a modest outward appearance: a better conversation, a more relevant offer, a decision that takes account of the available picture.

In conversations with Peter Renton, Kulshreshtha has put the product’s outcomes ahead of a prolonged argument over the AI label. He talks about approval rates, automation and what happens to a lender’s portfolio. The vocabulary is revealing. His audience has to live with the resulting loans. A demonstration may earn a meeting; what the decisions do over time determines the value of the system.

The cost of seeing the full picture

His concerns extend to the price of the information itself. In a December 2025 essay about open banking, he argued that expensive access to financial data could make cash-flow underwriting harder for smaller institutions to provide. He framed the question as a choice about infrastructure. A technically available insight can still be commercially out of reach. For a company built around richer lending decisions, that is a consequential limitation.

This adds a practical condition to his aspiration for wider credit access. Better algorithms require usable inputs, and those inputs have an economic cost. Kulshreshtha’s argument connects the experience of the individual applicant to the arrangements between financial institutions and technology providers. A rejected borrower sees the answer at the end of that chain. The founder spends much of his time thinking about what happened earlier in it.

A founder’s complaint, twelve years on

In 2026, Scienaptic’s agenda includes iCUE, which combines predictive intelligence with large language models and agentic AI. The company says its credit union service organization now has 17 strategic investors. Its September announcement for ELEVATE, a gathering in Nashville, put implementation choices and lending strategy on the program. Kulshreshtha’s emphasis remained on leaders choosing useful problems and accepting responsibility for the outcome.

The customer announcements give that agenda a working setting. Financial Center First Credit Union went live on the platform in September 2026. In October, Department of Commerce Federal Credit Union selected it to streamline lending and expand access for members. These are institutional decisions about how lending gets done, the territory Kulshreshtha has occupied since his years inside GE and Genpact.

His bank-branch story supplies a personal measure of the work. The customer should be seen; the employee should have the information to see him. After a career spent around models, teams and credit portfolios, he still returns to the desk where those systems meet a person. It is a small place to locate a technology ambition, and a demanding place to prove one.