Manav Agarwal

We'd built an HR attendance and task tool for IT companies. But IT companies already have this solved — a dozen enterprise tools do it, often for free. The real unmet need was in FMCG and field operations, where teams are semi-literate, mobile-first, and have zero digital infrastructure for accountability. We'd talked to the wrong users and validated the wrong assumptions.

The lesson isn't that we failed. Lots of startups fail. The lesson is that user interviews only tell you whether people like your idea — not whether your idea is for the right people. Those are two completely different questions, and we confused them. Real validation is the person who says: "I need this right now and I have no alternative." If I started WorkWave again tomorrow, I wouldn't start by building. I'd start by finding the person who couldn't sleep because of the problem.

What was harder to sit with was the question underneath it: why did this problem exist in the first place? The data existed. The databases were maintained. But it was practically inaccessible to anyone without a specific technical skill. Because in large organizations, technical capability is often used as a gatekeeping tool — consciously or not. Knowledge gets siloed. Access gets restricted. That dependency becomes structural.

AI can remove the technical barrier. A natural-language layer means anyone can ask questions. But it cannot remove the organizational incentive to keep information siloed. When you deploy AI in an enterprise, you don't just make things more efficient — you make the existing power structures visible. That's either the most valuable thing about it, or the most uncomfortable, depending on who's in the room.

So I made a deliberate decision: I would not spend significant time memorising Quant formulas. The ROI wasn't there. Quant has a very specific ceiling — you either know the formula or you don't. Grinding formulas for weeks to move from 4 to 25 wasn't worth the cost to everything else I was managing. I scored 4 in Quant — around the 52nd percentile. But 99.57 in DILR and 96.66 in VARC pulled me to 97.60 overall. The prep I didn't spend on Quant formulas went into the sections where it actually moved the needle.

This isn't a flex about CAT. It's about resource allocation. Most people optimize by effort — work harder on the thing you're weak at. I optimize by ROI — figure out which inputs move which outputs, and cut the ones that don't move the needle relative to the cost. The highest-leverage thing I did before CAT wasn't studying. It was deciding what not to study. That same thinking shapes every project I work on.