ThesisModelVenturesTeamInsightsBuild With Us
All Articles

Your Portfolio Now Has
a Portfolio Manager.
It's Powered by Claude.

For decades, senior portfolio manager-level analysis was a privilege of institutional money. Diverss changed that by building StockSense — an AI engine that wires Anthropic's Claude API to a proprietary five-dimension scoring framework. Here's exactly how it works.

4Analytical pillars per stock
25%Weight on ROCE/PE quality — highest in the framework
ClaudeAnthropic API reasoning across the full framework
LiveSignal layer + Correlation Engine for portfolio-level intelligence

There's a question every retail investor in India eventually asks themselves, usually after a bad quarter: how is it that a fund manager at a large AMC can look at the same stock I'm holding and see something completely different?

The answer isn't secret information. It isn't insider access. It's analytical depth — the ability to assess a stock not on one dimension (the price chart, a tip from Twitter, a quarterly earnings beat) but across many dimensions simultaneously, with a framework that holds all of them in tension. A senior portfolio manager doesn't ask "is this stock going up?" They ask: is the fundamental health sound? Is the momentum confirming or diverging from the technicals? Is the sector headwind manageable? Is institutional behaviour suggesting accumulation or quiet exit? And critically — how does this holding interact with the rest of the portfolio?

That kind of thinking takes years to develop and is impossible to scale. Until now.

StockSense by Diverss — a TEN Labs venture — was built to close that gap. You upload your portfolio as a simple CSV (symbol, quantity, average price), and StockSense runs it through the Diverss proprietary rules framework powered by Claude — returning stock-wise verdicts with strengths, concerns, hope factors, and a concrete next action. Not a rating. Not a chart. A briefing, the way a senior analyst would give it to you in a Monday morning meeting.

"A stock screener tells you what. A good analyst tells you why, what it means, and what to do about it. StockSense is built to do the latter — at scale, for everyone."

Why Claude — and Not Just Any AI Model

Before getting into the architecture, it's worth explaining a choice that shapes everything: why Claude specifically?

The answer lies in what StockSense is actually asking an AI model to do. It's not asking it to retrieve data. It's not asking it to run a calculation. It's asking it to reason — to take a structured scoring framework with five analytical dimensions, apply it to a real company's real data, hold all five signals in tension with each other, and then synthesise that into an explanation that a retail investor can actually act on.

That requires three things that not all models handle equally well. First, instruction-following fidelity — Claude follows the Diverss scoring framework precisely, not approximately. When the framework specifies how momentum should be weighted against fundamentals in certain market regimes, Claude adheres to that logic rather than defaulting to its prior training assumptions about markets. Second, long-context coherence — portfolio-level analysis means holding multiple stocks, their interconnections, and their collective risk profile in a single reasoning session. Claude maintains coherence across that full context without the analytical drift that plagues shorter-context models. Third, explanation quality — the output has to be readable. A retail investor doesn't benefit from correct analysis delivered in jargon. Claude's ability to translate complex multi-factor reasoning into clear, actionable language is not incidental — it's the product.

~73% of retail investors cite lack of quality research as a key reason for poor returns (SEBI Study, 2023)
10M+ New demat accounts opened in India in FY24 — most with no access to institutional research
₹40L+ Minimum AUM typically required to access dedicated portfolio advisory in India

The Four Pillars Diversss Built Into the Framework

The foundation of StockSense is a proprietary multi-factor scoring framework built by the Diverss team — research-backed, weighted by conviction, and calibrated specifically to the Indian market. Every stock in a user's portfolio is evaluated across four analytical pillars, each with defined sub-factors and percentage weights. Claude reasons across all four simultaneously to produce a unified verdict.

01
ROCE / PE Quality
Is the stock cheap relative to its return on capital? ROCE/PE Mapping carries the highest weight in the framework at 25% — because quality at a fair price is the most durable source of long-term return. Claude contextualises this against the stock's sector peers and market cap range.
02
Revenue & Profit Growth
4-quarter revenue and net profit momentum versus the sector. At 15% weight, this pillar asks whether the business is actually growing — not just whether it's cheap. Claude reads the trajectory, not just the latest quarter, and flags when recent growth diverges from the multi-quarter trend.
03
Technical Composite
Trend, RSI, and MACD combined into a single momentum score. Rather than presenting three separate indicators for the user to interpret, StockSense collapses these into one technical signal — and Claude explains whether that signal is confirming or contradicting the fundamental picture.
04
FII / DII & Mutual Fund Flow
Are institutions and mutual funds buying or quietly exiting? Foreign and domestic institutional flow, combined with MF position changes, is one of the most reliable leading indicators available — and one of the hardest for retail investors to act on without a framework. At 10% weight, it functions as a conviction signal: when institutional flow confirms the quality and momentum reading, Claude can say so with greater certainty. When it diverges, Claude flags it explicitly as a risk to monitor.

Beneath these four pillars sit additional weighted factors — Market Cap (10%), Price/Book (10%), Liquidity (10%), and Event Catalyst (5%) — that add further precision to the score. No single factor dominates; the framework is deliberately multi-factor so that no stock can score well by being exceptional on one dimension while hiding weaknesses on others.

How Claude API Is Wired Into StockSense

The architecture is worth understanding, because it's where the real innovation is. StockSense doesn't use Claude as a chatbot that happens to know about stocks. It uses Claude as a structured reasoning engine — fed a precise context and expected to produce analysis within a defined framework.

Here's how a single stock analysis request flows through the system:

Data Layer
Live market data, historical OHLCV, fundamentals, FII/DII flows, and news signals are fetched, cleaned, and scored against each of the five framework dimensions. This produces structured numerical and categorical outputs for every stock.
Scoring Engine
The Diverss proprietary scoring engine processes the raw data inputs and produces dimension-level scores. These are not simple averages — they reflect the rules, weights, and interaction effects the Diverss investment team has encoded into the framework.
Claude API
The dimension scores, key data points, and the full scoring framework are passed to Claude via the Anthropic API. Claude's role: reason across all five dimensions, identify the most significant factors, surface contradictions or confirmations between dimensions, and produce a portfolio-manager-quality narrative.
Output Layer
Claude's analysis is structured and surfaced to the user — a clear verdict, a plain-language rationale, key risk flags, and where applicable, the portfolio-level interaction of this holding with other positions.

The critical design choice here is the separation of concerns. The Diverss framework provides the rules — what to look at, how to weight it, what constitutes a signal versus noise in the Indian market context. Claude provides the reasoning — the ability to hold all those rules in mind simultaneously, apply them to real data, and explain the conclusion. Neither component could produce the result alone.

What the output actually looks like
A user holding Tata Motors receives not a number, but a structured verdict. Strengths: ROCE/PE quality score is in the top quartile for the auto sector; FII flow is net positive for a third consecutive quarter. Concerns: Revenue growth momentum has decelerated over the last two quarters versus sector peers; technical composite is signalling mild RSI divergence. Hope Factors: JLR margin recovery is a potential re-rating catalyst; a global commodity tailwind could improve margins. Next Action: Hold with a price alert — the technical signal warrants monitoring before adding. That structure — four dimensions of judgement, rendered in plain language — is what StockSense delivers on every holding, every time.

Live Signal: When Breaking News Hits Your Portfolio

The four-pillar framework scores your portfolio at a point in time. But markets move continuously — earnings drop, RBI policy shifts, sector news breaks, FII flows reverse overnight. StockSense's Live Signal layer solves the most anxiety-inducing problem in retail investing: does this news actually affect me?

StockSense correlates breaking news, earnings releases, and macro events directly against your specific holdings — showing which stocks in your portfolio are most affected right now, and how. Earnings impact, policy changes, sector moves, FII/DII flow shifts: each is mapped to your actual positions, not to the broad market in the abstract.

Claude's role here is synthesis. The signal layer surfaces the events; Claude reasons about their portfolio-specific significance — whether an earnings miss is already priced into the technical score, whether a macro event creates new correlation risk, whether a sector move changes the weight the framework should assign to FII flow for a particular holding. The result is not a news alert. It's a portfolio briefing that updates as the world changes.

"Most retail investors find out a stock was affected by news three days after the move. Live Signal closes that gap — not with noise, but with relevance."

Correlation Engine: Are You Actually Diversified?

The deepest insight StockSense delivers isn't about any individual stock — it's about what your portfolio actually is, beneath the names.

Most retail investors believe they are diversified because they hold twelve stocks across four sectors. What they often don't see is that eight of those twelve stocks have a correlation above 0.7 to the same underlying macro factor — say, domestic consumption recovery, or global commodity prices. In a drawdown, those stocks fall together. The diversification was nominal, not real.

StockSense's Correlation Engine maps the hidden relationships between every holding in your portfolio — sector overlap, risk clusters, and true diversification exposure. It answers the question no screener asks: do you know if you're truly diversified, or unknowingly concentrated in the same risk? Claude then articulates what the correlation structure means: where over-concentration exists, which positions are genuinely uncorrelated, and where the portfolio's actual risk lies versus where the investor thinks it lies.

Correlation Engine in practice
A user holding HDFC Bank, ICICI Bank, Bajaj Finance, and SBI Cards believes they're diversified across private and public banking. The Correlation Engine surfaces that all four have a rolling 90-day correlation above 0.78 to the Nifty Financial Services index — they're four bets on the same trade. Claude flags this as a risk cluster, explains the shared factor exposure, and identifies which positions would need to change to achieve meaningful diversification within the financial sector. That's the difference between owning four stocks and understanding a portfolio.

What This Changes for Retail Investors

The access gap in Indian investing has always been invisible to the people suffering from it. Retail investors don't know what they're missing because they've never experienced what genuine multi-dimensional portfolio analysis feels like. They know the outcome — underperformance, ill-timed exits, concentration errors they didn't see coming — but not the cause.

StockSense makes the gap visible, then closes it. For the first time, a retail investor opening an app sees the kind of analysis that previously required a relationship with a private wealth manager — not dumbed down, not simplified into a traffic light, but genuinely articulated and genuinely reasoned. The Claude API is what makes that possible at scale: the same quality of reasoning, for every user, every stock, every time.

The institutional edge in investing has never been purely informational. In the age of the internet, information is available to everyone. The edge has been analytical — the ability to think about information rigorously, across multiple frameworks, faster than the market prices it in. StockSense, powered by Claude, gives that analytical edge to anyone who wants it.

At TEN Labs, we build companies because we believe the most important problems have solutions that technology can now make possible. StockSense is one of them. The gap between how institutional money and retail money think about their portfolios is not a natural law. It's a product problem — and we've built a product that solves it.

Diverss · StockSense

See StockSense analyse your portfolio

StockSense by Diverss is live. Add your holdings and see what five-dimension AI analysis looks like on your own stocks — in plain language, with a verdict you can act on.

Try StockSense