AiThority Interview with Binny Gill, CEO of Kognitos

What challenges do enterprise tech teams face when deploying AI agents and tools today? Binny Gill, CEO of Kognitos weighs in with some observations in this AiThority interview:

____________

 Hi Binny, what inspired Kognitos?

Kognitos really came from a moment of frustration, not inspiration in the traditional sense. During the pandemic, my son was trying to build a simple game and it struck me that, despite decades of progress, programming was still far too hard for most people. We’ve made computers incredibly powerful, but we still expect humans to adapt to them instead of the other way around.

That led me down a deeper line of thinking. Why do we assume software has to be written in programming languages at all? Why can’t systems understand intent expressed in natural language and handle the complexity themselves?

Kognitos is an attempt to rethink that relationship. It’s less about automation as we know it and more about making computers meet humans halfway.

How is enterprise AI changing today, and what is dominating the ecosystem?

Right now, the ecosystem is dominated by large language models and the race to make them bigger, faster, and more capable. That has unlocked a lot of creativity, especially in areas like content generation and copilots.

But in the enterprise, the conversation is shifting. It’s no longer just about what AI can generate, it’s about what it can reliably execute. That’s a very different bar.

We’re starting to see a move from experimentation to accountability. Enterprises are asking harder questions around traceability, correctness, and governance. That’s pushing the ecosystem toward architectures that combine probabilistic models with more deterministic layers.

What are some of the top challenges that enterprise teams face when deploying new AI features to help power business functions?

The biggest challenge is trust. Not in a philosophical sense, but in a very practical one. If an AI system is going to touch billing, compliance, or customer data, teams need to know exactly what it did and why.

A second challenge is handling edge cases. Real business processes are full of exceptions, and most AI systems don’t deal with those well. They either fail silently or produce outputs that look plausible but are incorrect.

The third is operationalization. It’s one thing to demo an AI capability, it’s another to embed it into a system that runs reliably every day without constant human oversight. Real enterprise environments are full of legacy systems, inconsistent data, and processes that were never designed with automation in mind. Bridging that gap takes far more than the model itself.

Also Read: AiThority Interview with Glenn Jocher, Founder & CEO, Ultralytics

What is the highest value customer use case Kognitos can solve?

The highest value tends to come from complex, multi-step business processes that are currently manual, error-prone, and involve a lot of decision-making.

Things like claims processing, compliance checks, or financial operations workflows are good examples. These are areas where small mistakes can have real consequences, and where traditional automation struggles because of the variability in inputs and rules.

More broadly, the real value is in enabling systems that don’t just execute steps, but can handle exceptions and adapt to real-world conditions without breaking. The real differentiator isn’t just automating the happy path, it’s building systems that can handle the messy middle: the exceptions, the edge cases, the judgment calls that traditionally required a human. That’s where meaningful value lives.

What about the current state of AI most interests you and what about it makes you most weary?

What interests me most is that we’re finally at a point where machines can engage with human language in a meaningful way. That’s a profound shift and opens up entirely new ways of interacting with software.

What makes me wary is how quickly we’re moving to deploy these systems in high-stakes environments without fully addressing their limitations. There’s a tendency to treat impressive outputs as reliable outcomes, and those are not the same thing.

If we don’t build the right guardrails and architectures around these systems, we risk creating a lot of hidden fragility in places where precision actually matters.

Five top myths around the future of AI you’d like to bust in this conversation?

The most persistent myth is that AI progress means making machines more sophisticated. Some of the most important shifts come from the opposite direction, making it simpler for humans to express what they actually need. For decades, people have had to think like machines to make them useful. We’re finally approaching the inversion of that

One myth is that bigger models will solve all the hard problems. Scale helps, but it doesn’t address fundamental issues like reasoning, traceability, or correctness

Hallucinations are an engineering problem that will eventually be solved. In probabilistic systems, uncertainty isn’t a bug to be patched, it’s a structural property. The real question is whether you’ve designed your system to contain and surface that uncertainty, or to obscure it.

AI will replace developers entirely. What’s more likely is a shift in what developers are responsible for, or everyone becomes a developer. Less implementation, more intent. The craft moves up the stack, not out the door.

The AI adoption is mainly a technology problem. In reality, it’s just as much about process, governance, and how organisations build trust in these systems.

Also Read: ​​The Infrastructure War Behind the AI Boom

[To share your insights with us, please write to psen@itechseries.com ]