As AI makes copying easier, this product leader says trust is the real edge

A young composer built a backyard studio in 1989 and did everything himself: composing, producing, engineering. By 1992, his first film soundtrack was a nationwide hit. Decades on, he is still refining what film music sounds like. That composer was A.R. Rahman.

Kunal Shrestha, VP of Product at Responsive, opened his session at DevSparks Chennai 2026 with this story, using it as a frame for developers watching AI make software cheaper and faster to build, and by extension, easier to replicate. His session, titled ‘The new moat for developers: moving beyond the feature’, set out to answer where differentiation goes once features stop being defensible on their own.

Redefine or get replaced

Shrestha says questions posted on the coding forum Stack Overflow have collapsed to a fraction of what they once were, a drop he attributed to developers who entered the workforce after 2021 turning to AI tools instead of forums.

"If you don't redefine, you get replaced," he said, arguing that the pressure cuts both ways: software companies risk losing customers to AI-built in-house alternatives, while internal engineering teams must prove their own AI systems hold up over time. He pointed to a recent McKinsey survey: 32% of organisations have decided against buying a software product because it could be built internally with agentic coding tools.

Building trust before building smarter

Responsive, a company that helps enterprises respond to requests for proposals, security reviews, and vendor questionnaires, shipped its first AI feature in 2024. Shrestha said the company does not build or fine-tune its own models; instead, the early priority was closing what he called a "trust gap" between customers and AI-generated output. 

"What started as 10% of our customer base that used AI within our products is now 70% who are using it," Shrestha said. He added that Responsive serves 2,000 customers globally, including more than a quarter of Fortune 500 companies, several of which have the engineering talent to build comparable tools in-house but have not.

Once trust was established, Shrestha said the company moved to automation, launching agents to help customers handle more requests faster, and then to intelligence, adding reasoning capabilities once speed alone stopped being enough of a differentiator.

Three pillars of differentiation

Shrestha outlined three areas where he believes companies can still build a defensible edge: trust, intelligence, and distribution.

On trust, he said AI systems need more than a confidence score attached to their output. Users should be able to trace any answer back to its source, and every step an AI system takes toward a decision should be logged, both for internal support teams and, eventually, for users who want to know why a system did what it did.

On intelligence, Shrestha pointed to agentic RAG, a method where an AI system pulls information from a company's own data rather than relying only on what it was originally trained on, combined with memory of a customer's history so systems do not start fresh with every interaction. 

He also described learning from patterns across many customers, using anonymized onboarding data, for instance, to speed up onboarding for new customers without exposing any individual customer's data.

On distribution, Shrestha said companies should not expect users to come to their application. AI tools should be built into the platforms customers already use daily, through direct integrations or the Model Context Protocol, a standard that lets AI systems connect to external tools and data sources. 

"You need not be hung up on your application," he said. The right interface, he added, depends on the task, and a genuine omni-channel experience lets a user start in one interface and switch to another without losing context.

How the work itself is changing

Shrestha contrasted how Responsive built software two years ago, with a product manager writing a specification, a designer creating mockups, and an engineer implementing them as three separate handoffs, with a newer approach the company now uses for simpler projects. 

In that approach, he explained, one person, often a developer, prompts an AI coding tool connected to the company's design system, then reviews and publishes the result. "For simple projects, only one person can do this from start to finish," Shrestha said, adding that complex initiatives still "involve people from all the multiple functions." 

Shrestha closed by returning to curiosity, telling developers that staying relevant means continuing to question whether the problem being solved is the right one, not only how well it is solved.