Tijori Stack vs Altys: Two Different AI Research Systems for India

Tijori Stack and Altys are both India-focused financial research systems that try to ground AI in real evidence. Tijori Stack is strongest as a suite for fast company understanding through concalls, disclosures, generated reports and company-grounded questions. Altys is built as a broader research operating system in which source-linked evidence also feeds deterministic calculations, point-in-time history, screeners, models, backtests and continuous monitoring.

That is the important correction to simplistic comparisons: hallucination control is not unique to either product. Tijori publicly describes Atlas as answering from company material rather than the open web. Altys also grounds answers, then applies a stricter system contract around financial numbers: code performs the calculation, material claims carry citations, and unsupported answers fail closed. The difference is not “grounded AI versus ungrounded AI.” It is the research workflow built around that grounding.

What Tijori Stack is and who it serves

Tijori Stack sits on top of Tijori’s long-running database of Indian company data and applies AI to the research workflow rather than bolting a chatbot onto a price screen. Based on its public materials and Zerodha’s coverage of the launch, it is made of four products.

Concall Monitor. A few minutes after an earnings call ends, you get the full transcript plus an AI summary that pulls out pricing, margins, growth, demand, guidance, capex, and risks. Its standout move is a management-consistency check: it compares what management said in the past against what they are saying now, so you can see whether a company followed through, changed the story, quietly postponed a target, or acted as if an earlier comment never happened.

Report on Demand. You pick a company and choose what you want, and Tijori builds a deep-dive from filings, disclosures, and financials. It offers a Risk Probe that surfaces red flags and key risks, a Management Credibility report that compares what management said against what actually happened, and an AI-assisted five-year revenue and EBITDA estimate grounded in the company’s own data.

Radar. You define a metric or a risk you care about, such as client concentration, promoter pledging, working-capital stress, capex delays, receivables build-up, dollar revenue exposure, or margin pressure, and Radar keeps scanning disclosures and alerts you when it shows up. It turns a standing thesis question into a monitor.

Atlas. A company-grounded question-and-answer tool that answers from a company’s own filings, disclosures, financials, and concalls rather than scraping the open web. Tijori frames its edge as roughly fifteen years of structured and unstructured filings processed into machine-readable form, which is what keeps the answers anchored to source material.

Tijori Finance is the parent behind all of this. It has been building Indian equity data since 2016 and is known for going beyond the three financial statements into segment, operational, and alternative data.

Where Tijori Stack is genuinely strong

This is a thoughtful, capable stack, and it deserves real credit. A few things stand out.

• Concall speed and consistency checks. A transcript and a structured summary minutes after a call is a real workflow win, and comparing past guidance against current commentary is exactly the kind of judgment work analysts otherwise do by hand across quarters.
• Report generation on a clean base. Building the risk probe, credibility check, and multi-year estimate on top of a long, cleaned filings history is a sensible way to make an AI report more rigorous than a generic model working from whatever it can find.
• Grounded answers. Constraining Atlas to a company’s own disclosures is the right instinct for research, where an answer that cannot be traced to a filing is worth little.
• Standing monitors. Radar fits how professionals actually think, where the question is not just what is true today but whether a specific risk is developing over time.

If your work is reading businesses through their disclosures and concalls quickly and well, Tijori Stack is a strong tool on its own terms. Reading an earnings call closely is a discipline in its own right, and it pairs with the way we describe how to read a concall like an analyst.

Grounding is not a binary feature

Calling a product “grounded” is only the start of the evaluation. A model can retrieve a real filing and still make a wrong-period, wrong-entity or wrong-unit claim. It can cite a document that contains a nearby fact but not the sentence being asserted. It can also calculate a ratio incorrectly after retrieving both inputs.

A professional team should test five separate controls:

1. Retrieval: did the system find the right primary source?
2. Claim support: does the cited passage support the exact statement?
3. Calculation: was the number computed from compatible inputs?
4. Time: was the evidence available on the date being studied?
5. Failure behaviour: what happens when the source does not contain the answer?

Tijori Stack’s public materials establish the first principle clearly for Atlas: answers are constrained to company filings, disclosures, financials and concalls. Altys treats the five controls as one pipeline. The answer is grounded, the calculation is deterministic, the history is point-in-time, and missing evidence becomes an unavailable state rather than a completed paragraph.

Where a professional research desk may need a broader system

There is honest overlap here. Tijori Stack does concalls and reports, and so does a professional desk platform, so this is about emphasis rather than one tool being better than the other. None of the following is a knock on Tijori Stack. They are simply different jobs.

Point-in-time history. An AI summary and a report describe a company as it looks now, with today’s restated and reclassified numbers. For a research process or a backtest, what matters is what a company had actually reported and what was knowable on a given past date, before those revisions. Judging a past decision with today’s tidied-up figures is how lookahead bias creeps in, and it is why point-in-time data matters for anyone testing a process.

Backtesting a strategy. A five-year AI estimate and a risk report help you evaluate one company today. Neither tells you how a rule, a screen, or a factor tilt would have behaved across past cycles using only the data available at each step. That needs a point-in-time backend and an explicit backtest, which is a different kind of tool.

Auditable source-linking. When a number feeds an investment committee memo, an analyst needs to trace it to the supporting filing, period and basis. Tijori’s filings-grounded approach gives generated work real footing. Altys makes traceability a system-wide contract for material research claims and calculated figures, including the explicit unavailable state when a source cannot support the answer.

Exportable calculation trails. A cited answer can be checked in a document. A scored universe also needs its inputs, eligibility rules, weights and calculations. Altys strategy screens and scorecards export as formula-native Excel workbooks, and GenGrid exports an Excel-ready CSV with persisted citations. This gives an investment team a second verification surface outside the product.

One institutional workflow. PMS firms, AIFs, and family offices tend to want screening, modelling, forecasting, and backtesting in one place on a shared, auditable data layer, rather than a set of AI research tools sitting alongside their model. This is the shape of the institutional equity research workflow.

How Altys approaches the same problem

Altys Labs is an equity research and fundamental analysis platform for Indian stocks (NSE and BSE) and Indian mutual funds, built for professional users: PMS firms, AIFs, family offices, and MFDs. It currently runs one-month pilots for investment teams. A few points on how it approaches the work, stated as focus rather than as any claim of superiority:

• India-first and India-deep. It brings India’s market data into one place: company filings, concall transcripts, management guidance, shareholding patterns, 80-plus global and domestic macro series, FII and DII flows, factor scores, and mutual-fund data.
• Point-in-time by design. It keeps data as it stood on each past date, so research and backtests reflect what was actually knowable then rather than restated history.
• Source-linked and auditable. Every figure links back to its source document, line, and date.
• Calculated, not guessed. Numbers are computed from a company’s own filings, and forecasts come from explicit statistical methods rather than a language model guessing a growth rate.
• Fail-closed research answers. A material answer without valid supporting evidence resolves to error or unavailable rather than being published as confident prose.
• Exportable audit paths. Strategy screens and scorecards compile to formula-native Excel with inputs, formulas, weights and provenance; GenGrid and analytical tables export as Excel-ready CSV; alerts retain rule versions and historical replay.
• Tools on top of the data. You can screen, build financial models, forecast, and backtest a strategy in one place.

Altys is not a broker, not a tip service, and not a SEBI-registered research analyst or adviser. It does not tell you what to buy or sell. It is software for doing your own research.

Which fits whom

Dimension	Tijori Stack	Altys Labs
Best fit	Fast company discovery, concall reading, generated reports and disclosure monitoring	Professional research teams connecting evidence to models, screens, backtests and portfolio monitoring
Research grounding	Atlas is publicly described as answering from company filings, disclosures, financials and concalls	Answers use source-linked financial data and cited document evidence
Numeric calculations	Public product materials describe estimates and reports grounded in company data; the method depends on the workflow	Ratios, factors and forecast outputs run through explicit code-owned methods
Citation behaviour	Filings-grounded answers; confirm the in-product citation level for the workflow being tested	Material claims carry citations to supporting evidence
Missing evidence	Test the chosen workflow directly, because generated-report handling can differ from Q&A	Unsupported material answers fail closed as error or unavailable
Exportable audit path	Confirm the export and calculation detail required by your workflow during a trial	Formula-native Excel for strategy screens and scorecards; Excel-ready CSV for GenGrid and analytics
Point-in-time history	Not described publicly as the core product contract	Core design principle across historical research and backtesting
Strategy backtesting	Not described publicly as a core workflow	Built into the research system
Monitoring	Radar monitors defined metrics and risks in disclosures	Company, thesis, factor, guidance and portfolio monitoring share the same data layer
Access	Tijori Finance product suite	One-month pilots for investment teams

The honest summary: if you want fast, well-grounded AI reads of a business through its concalls, disclosures and reports, Tijori Stack is a serious tool worth testing on its own terms. If the binding constraints are point-in-time history, reproducible calculations, source-linked research, backtesting and continuous monitoring across one institutional workflow, that is the system Altys is built to provide.

Neither product should be reduced to a chatbot. The useful buying test is to take one company you know well and ask both systems for an obscure sourced fact, a derived metric and an undisclosed item. Inspect the evidence, reproduce the calculation and see how each handles the missing answer. Our broader reliability checklist explains seven tests for AI stock analysis, and the Altys implementation is described in How Altys prevents hallucinated financial numbers.

If you are weighing Tijori’s wider product line, our note on Tijori Finance alternatives covers the segment and alternative-data app. For a broader survey of the landscape, see our roundup of the best AI research tools for Indian stocks.

If Screener.in is also on the shortlist, use our workflow-level comparison of Screener.in, Tijori Finance and Altys.