You have a system that can’t afford to break, and everyone agrees AI belongs in it somewhere. Or you’re about to commit the next tranche of budget and want a read on it before you do.
Four questions to ask before you spend
Most AI programmes fail on something structural that shows up long before anyone writes code. I watched the identical pattern run through enterprise blockchain from the inside, and I hold ten patents in it — which is why I’m willing to say most of it didn’t need to be blockchain at all.
These cost you nothing:
- Who has the power to insist — and will they actually spend it?
- What does the person you’re forcing to change get out of it?
- Who here is powerful enough to refuse? If nobody can, you’ll get compliance on paper and sandbagging in practice.
- Strip out the technology. Would a boring version deliver the same result?
The fourth is the one nobody asks, because most people advising you are paid more when the answer is no.
Where this comes from
Airtel Digital, as Principal PM. Before any of the AI, I sat on customer calls and visited Airtel Centres, and took what customers were actually saying to leadership in their own words. Then I specced and shipped the system that industrialised it — intent classification, agent self-training, moderation — running inside a twenty-year-old stack rather than beside it.
Most people arrive with a model looking for a problem. The order was the point.
Now building LEXI Platform, AI-native trade and payments infrastructure for Indian exporters out of GIFT City — its subsidiary LexiMoney holds in-principle approval as a Payment Service Provider under IFSCA. And MoneeFlo, an RBI-licensed forex company. Earlier: Reliance Jio and Ericsson. The full timeline is on the about page.
If the four questions didn’t settle it
Send me a line on LinkedIn about what you’re deciding, and I’ll come back with three questions. If the answers say I can’t help, I’ll tell you.