Before and after AI.

Sriram and I go back to a time when AI was a research paper, not a job description. He came into Travelopia as a consultant — helping us understand what DORA metrics actually meant, how to make software delivery legible to a business, how to know if what we were building was moving any needle at all. The word he uses for that body of work is impact intelligence, and it is a useful frame even now — maybe more useful now than it was then.

What struck me most in this conversation was how specific Sriram has gotten about where AI is actually landing, versus where the headlines say it is. Coding assistance is real and measurable — he gives himself a 4x efficiency multiplier on work he used to bill ten hours for. But the more interesting territory is upstream: the spec, the epic, the story. One of his clients is scanning hundreds of stories every few days with AI, scoring each against weighted quality criteria before it reaches development. Not to replace human judgement — to make the spec ready for the agent. Better documentation helps the agent. The agent only works as well as the context you give it. That is the loop most AI-in-engineering conversations miss.

His older thesis — don't organise by function, organise by outcome — has become urgent rather than merely interesting. Functional decomposition creates handoffs. AI compresses handoffs. He'd start fresh with a small, tight-knit team per product vertical, roles fluid rather than fixed, everyone expected to cover adjacent competencies that AI now enables them to cover. The argument is the same as it was a decade ago. The case is now overwhelming.

The question I most wanted to ask was about friction. We both see it. The ask is obvious; the movement is slow. Sriram's answer: career anxiety. Not tech resistance. Junior engineers worry about identity — am I a developer or a jack of all trades? Senior leaders worry about reporting lines — if I dissolve functional departments, who do my VPs become? Neither reaction is irrational. Both are slowing things down.

We ended on the question I have been sitting with since my own build work over the last few months: what do you tell an engineering student right now? Sriram landed on Jevons paradox as a reason for cautious optimism — the cheaper software gets to produce, the more we will demand of it. I landed somewhere slightly different: dreaming skills, not just coding skills. And whatever field you are in, learn to communicate. Fluently, in a language that gets you heard. Language is the new differentiator.

— SREE