The Same Old Mix: Infosys's SVP on Why Agents Are No Different From Any Other Stack Decision

This piece draws from a fireside chat and AMA session, "Failures, Surprises, Advice," at AISRENext Bengaluru. The panel had already been asked, and answered, a build-versus-buy question from the audience.

Naresh Choudhary, SVP at Infosys, agreed with what came before him, then reframed the decision as one Infosys has made many times over, long before agents existed.

Decades of Open Source and Custom Tooling Set the Pattern Already

“For us, the approach has always been a thoughtful blend. Over the years, we have used open source, commercial off-the-shelf solutions (COTS), and custom or homegrown platforms depending on the need—and the same principle applies here as well.” 

Infosys's technology stack has never been a single answer. It's been three categories running in parallel: 

• Open-source components, adopted where mature ecosystems have already addressed the problem effectively
• Commercial and vendor-backed tools, selected where specialized capability, support, or speed justifies the investment
• Custom-built solutions, reserved for business-specific needs that cannot be adequately met through off-the-shelf options

Agents are landing in the same three-way split, not a new decision framework built from scratch. 

Building Small Language Models Alongside the LLMs Everyone Buys 

“While we are leveraging large language models, we are also developing our own small language models for specific use cases. The future will require that kind of balanced approach.” 

This is the mix in practice. Infosys uses large language models where scale, breadth, and general-purpose capability are required, while developing smaller, purpose-built models for targeted use cases where greater control, specificity, and efficiency matter. Neither approach is inherently superior; each has a role when applied to the right problem. 

The Two Filters Behind Every Agent Decision 

Naresh’s guidance begins with caution before moving into a practical decision framework 

“Avoid a strategy that locks you into a single path.” 

He then outlined the two considerations that should shape the decision: one for solutions that are purchased and one for those that are built. 

1.  ROI, for anything being purchased. The decision should be grounded in the value delivered for the investment, rather than a broad preference for buying over building. 
2.  Purpose, for anything being built. Custom development should be justified by a clear business objective, not by the appeal of building internally simply because the capability exists. 

Why Purpose-Driven Teams Land on Hybrid by Default 

“When teams are purpose-driven and focused on measurable business outcomes, a hybrid strategy typically becomes the most pragmatic path forward.” 

Naresh's answer doesn't split the difference between Sanjeev's argument for buying and Suvesh's case for building selectively. It explains why both show up on the same roadmap. A team anchored to business outcomes rather than a fixed build-or-buy posture ends up buying where a vendor's expertise clears the ROI bar and building where the goal is specific enough that nothing generic gets them there. 

At StackGen, Aiden is designed to fit into that kind of hybrid stack rather than compete with it, handling the governance and reliability layer so teams can still build custom agents for the parts of the business that need them. Try Aiden for SRE Community Edition for free today.