Why TwoDots exists.

The night before Black Friday.

Sometime in 2016, I (Sunil Kumar, founder) was sitting in a war room at Kohl's headquarters in Milpitas, California — the night before Black Friday. I'd spent the better part of that year building a recommendation engine for their website and stores. Millions of shoppers were about to hit it.

The room was full of serious people making serious decisions. Inventory placement. Promotional pricing. Which SKUs to push, which to pull. Real-time calls under real pressure. And I watched senior retail executives — smart, experienced operators — staring at dashboards and still flying half-blind. Even with the data. Even with the models. Even with everything we'd built.

The gap between data and decision was enormous. And almost nobody was bridging it.

What I realised that night wasn't technical. It was structural. The problem wasn't that good AI was impossible to build. We'd just built it. The problem was that even inside a $17B retailer, with a full data engineering team and a multi-million dollar budget, the AI barely made it from development to production — and almost never made it from production into the decisions that mattered. The people making calls at midnight on Black Friday eve were not looking at model outputs. They were looking at experience and instinct and hoping for the best.

If this was the reality inside Kohl's, what was it like for everyone else?

Coming home.

When I came back to India, I saw two things simultaneously. Something worse, and something better.

Worse: because the online sellers I met — the Shopify stores, the Amazon businesses, the B2B wholesalers — had nothing. No recommendation engine. No demand forecasting. No automated reconciliation. They were navigating the same complexity that kept Kohl's executives up at night, but with spreadsheets and gut feel and no safety net. Stockouts cost them 15–20% of potential revenue. Returns ate margin without any predictive model to reduce them. Payment reconciliation was manual and months behind. They were operating a modern business with 1990s information systems.

Better: because they had three things the big retailers didn't — speed, ownership, and nothing to lose. A D2C brand doing ₹10 crore a year can try an AI experiment on Monday and see a result by Friday. There's no 9-month procurement cycle. There's no legacy system politics. The founder is the product manager and the P&L owner and the person who answers Slack. When the AI ships, it ships into a business that can actually use it.

This was the opportunity I couldn't walk past.

What we built first.

We spent seven years building HappySellers — a platform that today handles order processing, inventory management, and payment reconciliation for 250+ Indian e-commerce businesses. Over 6,000 sellers have registered on it. Every day, we see the AI use cases that move real numbers. Every feature we've shipped has been validated on real GMV, real returns, real margin impact.

HappySellers is not a consulting portfolio. It's a live laboratory. When we recommend a demand-forecasting approach to a client, it's because we've already run variations of it across hundreds of sellers. When we say "returns prediction reduces your reverse logistics cost by 18%," we have the data behind that number.

This is the TwoDots moat that no agency can replicate. Not credentials. Not case studies. A live, operating platform that tells us every week what AI is actually doing for the businesses we care about most.

The playbook exists. It just hasn't been compressed.

Here is what I've learned across 15 years: the AI that makes a $10B retailer 3% more efficient also works for a $5M Shopify brand. The maths are the same. The data patterns are the same. The implementation discipline is the same.

The difference is access. Enterprise brands paid McKinsey and Accenture and Deloitte to build this over years and millions. The $5M brand either gets told "you're too small" or gets sold a SaaS tool that's one feature of one use case wrapped in a nice dashboard.

Neither option actually ships working AI into their business.

TwoDots is the third path. We take the enterprise playbook and compress it into a 4–12 week implementation that actually goes into production. No decks. No "recommendations." No "proof of concept that requires another phase." Working AI, in your stack, moving a KPI you can measure.

"We don't sell AI. We ship outcomes."

What I believe.

I don't believe AI will replace retail operators. I've sat in enough war rooms to know that the human who understands the SKU, the season, the customer, and the margin will always be irreplaceable. What I believe is that AI will finally give retail operators the tools they've always deserved — the tools that Fortune 500 brands spent hundreds of millions to build, now available to a 20-person team running a Shopify store.

That's the world TwoDots is building. Not hype. Not slides. Not another pilot that never sees production.

Just AI that ships.