One Year of Building DigiIQ

A year ago DigiIQ's AI layer was still closer to a conviction than a product, and I've spent the year in the weeds building it.

The conviction was simple: brands with real ambition kept getting stuck behind commerce platforms and content pipelines that could not move at the speed the ambition demanded. Everyone could see it. Almost nobody was closing the actual gap between AI and the systems brands already ran on. That's the gap I put my head down and started closing.

There was no anniversary planned for this post. I only noticed a year had passed when I sat down to write about something else and realized the list of things we'd shipped had gotten long enough to need its own post.

What the year actually looked like

For Mahindra Electric, we shipped a voice assistant that answers real questions about range, charging, and features across the entire SUV lineup, day or night, no waiting on a dealership to open. We built it to sit alongside the same answer-engine-optimized content that makes the site legible to AI search in the first place, so a customer gets the same answer whether they ask the assistant on the site or ask a model somewhere else entirely. Consistency between those two channels turned out to matter more than either one individually.

For Bajaj, we built Pulsar Underground, a mobile-first microsite that treats the bike as a creative partner instead of a product to be photographed. Creators pick a challenge across music, dance, art, style, or stunts, then build their entry three ways: fully automated AI generation, a hybrid mode that blends AI with their own choices, or a fully manual upload for anyone who wants total control. The AI toolkit, powered by Google Gemini, includes a stunt video generator that drops a rider into styles like Manga and Urban Samurai, an album cover generator, and a mashup engine, with a live leaderboard and direct Instagram integration turning every entry into something shareable instead of a submission nobody else would see. We expected manual upload to be the fallback for people who didn't trust the AI tools. It wasn't — across the six-week window, close to 1,900 of the roughly 2,200 total entries used the AI toolkit, meaning creators treated it as the actual instrument, not a shortcut around making something themselves.

We built jadau.com for Surana Jewellers, a storefront, an AI product imagery workflow, and a voice chatbot designed together from day one instead of bolted on after, for a catalogue where every single piece is genuinely one of a kind.

We built BourbonIT 2.0 for Britannia, an assistant that listens, sees, and speaks in seven languages. Alongside that, campaign work for Yamaha, Alan's India, and Invisalign.

None of it shipped as a pilot. All of it is still live.

The AI layer has to be a day-one decision, not a feature request

The earliest builds I shipped treated the AI layer as something you bolt onto a storefront once the storefront is already considered finished. A feature request that shows up after the "real" project is done.

It never worked as well, and it always cost more time than doing it the other way.

Every platform we've shipped since has been designed with the AI layer and the commerce layer as one decision made on day one, not two decisions made a year apart. I learned that the hard way, on our own build, before I started insisting on it for anyone else's.

Our own website is the clearest proof of what I learned

Search is turning into an answer, not a list of links, and I spent a good part of this year internalizing what that actually changes for a storefront. Before recommending any of it to a client, we applied it to ourselves: structured schema for our organization and every case study, a plain llms.txt file that tells any AI system directly who we are, and every fact checked against every other page that mentions it.

digiiq.com today is not the site we started the year with. It's the one built with everything we'd learned building everyone else's, which is a strange but honest way to describe a homepage.

What I'd tell myself a year ago

Build the discipline you're selling into your own work first, or a client will eventually notice you haven't.

That's the only rule from this year I expect to still be repeating a year from now.

Written with Vox.