← All case studies
AssetWorks AI · 2024–25 · Co-Founder & CEO

Ten crore new investors, and nobody had built them a research tool

India's demat accounts went from four crore to over fifteen crore in four years. Almost none of those investors had access to research infrastructure worth the name — the professional tools cost more than they would ever invest. This was the thesis I founded a company on, and what I built against it.

Role
Co-Founder & CEO
Sector
Fintech · AI · Investment research
Problem
Research fragmentation and intelligence inequality
Scope
Thesis · Product · Team · Platform
The thesis

An investment revolution that stopped halfway

Between 2020 and 2024, Indian demat accounts grew from roughly four crore to over fifteen crore. Investors under thirty went from around 29% of the total to nearly half. By any measure it was the largest retail investment expansion in the country's history.

What did not arrive alongside it was any way for those investors to make good decisions. They inherited a research stack assembled for somebody else — a different generation, a different asset mix, and a different assumption about who was doing the analysis.

The revolution was in participation. It never reached the intelligence that participation requires.

Diagnosis

Two problems that look like one

It is tempting to describe this as a tooling gap and start building features. The gap has two distinct causes, and a product that solves only one of them fails.

Problem 01
Fragmentation
A single decision meant moving across four or five platforms — one for prices, another for ratios, another for charts, another for news, then a community somewhere to check whether any of it was sensible. Over an hour, and still no confidence at the end of it.
Problem 02
Intelligence inequality
Institutional research tools cost tens of thousands of dollars a year. Quality advisory takes a percentage of assets. Both are structurally out of reach for someone investing a few lakh — the people who most need help are the least able to buy it.
Problem 03
Complexity by design
Personalisation in this category stopped at content. The interface stayed uniformly difficult, forcing every user to adapt to the tool rather than the reverse — which quietly selects for people who already know what they are doing.

Fragmentation is a workflow problem and could be solved with aggregation. Intelligence inequality is an economic one and cannot — no amount of interface design makes a thirty-two-thousand-dollar terminal affordable. Solving only the first produces a tidier version of the same exclusion.

Decision

Build the layer between information and decision

The strategic choice was about what not to be. Not a brokerage, because the moment you take a position in the transaction your research is suspect. Not an advisory service, because the goal was to make individual investors more capable rather than to decide for them. Not another data source, because the problem was never a shortage of data.

What was missing was the intelligence layer: the thing that sits between fragmented information and a confident decision. Platform-agnostic, so it could be honest, and priced for consumers rather than institutions.

The old journey
Four platforms, an hour, low confidence

Prices here, ratios there, charts somewhere else, news elsewhere again, sentiment from a group chat. Subscriptions stacking up across tools that never spoke to each other.

The intended journey
One question, one answer, in seconds

Ask what you actually want to know — across stocks, funds, bonds, ETFs — and get an integrated answer with the reasoning visible.

The cross-asset capability mattered more than it first appears. Real investment questions do not respect product boundaries. Should this money go into a bank stock, a mutual fund, or a fixed instrument, given a ten-year horizon? Every existing tool answered within its own category and none could answer across them — which meant the most common question an Indian investor actually has had no product built to address it.

The product

Three pillars, one of which was the moat

Pillar 01
Intent
Read what the investor is actually asking and return that, rather than returning everything and leaving them to sift. Efficiency of intelligence, not volume of it.
Pillar 02
Visualisation
Turn multi-dimensional analysis into something immediately legible. The bridge between professional-grade output and mass-market comprehension is almost always visual.
Pillar 03
Community validation
Every user makes the system better. Individual insight compounds into collective judgement — and unlike the other two, this is a defence that strengthens with scale.

The third pillar was the strategic bet. Intent parsing and good visualisation are buildable by anyone with capital and time; they are features, not defences. Validated collective judgement is not, because it requires users you already have. In a category where large technology companies can ship a competing feature in a quarter, network effect was the only durable position available.

There was a discipline underneath it that mattered as much: AI needed productising, not just deploying. Raw model capability is not a product. What makes it useful is validation, workflow integration, and adapting to how someone actually decides — the unglamorous work between a capable model and a tool people trust with money.

From the manifesto

We're drowning in data, yet starving for insight. Quality research tools belong to everyone, not just institutions. Structure brings order to chaos. Collective intelligence trumps individual genius.

Written early, and deliberately. In a category where the temptation is to build whatever the technology newly permits, a stated set of beliefs is what keeps a product from drifting into being clever rather than useful.

Execution

What got built

The first version of the business was manual — insights delivered by hand to an engaged community of investors. That was deliberate. It proved people wanted the thing before anything was automated, and it taught us what they actually asked for rather than what we assumed they would.

It also had an obvious ceiling. Manual curation does not scale, and knowing that early is more useful than discovering it at volume. The pivot was to automate the delivery of intelligence rather than the judgement inside it — AI as the mechanism for scale, with the human validation kept in place.

Alongside it: a founding team drawn from CoinDCX, Captain Fresh and other product-led companies, with a co-founder who had built and led product at IPO-stage businesses. And a working MVP — the intent engine, the visual layer and the core analysis tools built and shipped, not described in a deck.

Why it's interesting

The hardest part of an AI product is the part that isn't AI.

Founding this taught me something that has held up everywhere since. When a technology becomes newly capable, the instinct is to build what it can now do. But capability is not a product, and it is not a position either — anything buildable purely from a model is buildable by everyone with the same model. The work that matters is productisation: validation, workflow, and the interface between raw capability and how a person actually decides.

The related lesson is about which problem you pick. Fragmentation and intelligence inequality look like one gap from the outside, and only one of them is solvable with better software. Aggregating fragmented tools makes a tidier product for people who could already afford the tools. Changing who can access institutional-grade analysis is a harder problem and the one worth building for — but it requires deciding, early and explicitly, which of the two you are actually in business to solve.

Facing a version of this problem?

Tell me where you are and what is stuck. If I am not the right person, I will point you to who is.