AI tools for Indian stock fundamental research: a checked workflow

Which kind of AI research tool fits the job?

An assistant, a data connector and a source document serve different roles. The assistant writes and compares; the connector retrieves structured fields; the filing lets you check what the company reported. A broker connector adds account context, which a public company-data service does not have.

DalalOS publishes this guide and provides one of the services discussed. Use the dated provider comparison for feature and pricing claims. This page explains how to check a financial workflow rather than rank vendors.

Choose by task, then check the provider’s current documentation

How do you build a five-period Infosys table?

Run get_financials with query="INFY", period_type="annual", limit=5. Save the response before writing a narrative. Our 2 November 2026 run returned consolidated statements in INR, with bse-xbrl as the source. The response was stale; its cache timestamp was 25 October 2026 at 21:02:11 UTC.

The following values are a dated tool snapshot, not a fresh issuer-verified five-year series. Only the FY2024 revenue cross-check against Infosys’s annual report was performed here. Recheck later periods against their filings before using them in a research decision.

get_financials({ query: "INFY", period_type: "annual", limit: 5 })

Observed Infosys consolidated tool output; INR crore, rounded to whole crore; stale response

How should you calculate growth?

Five annual observations contain four year-to-year intervals. For the displayed revenue endpoints, calculate CAGR as (178650 / 121641)^(1/4) - 1. Do not label that result a five-year CAGR. The tool’s separate growth block can use a longer cached history than the five displayed rows.

Check whether both endpoints are positive before using this formula. A loss base, changed accounting basis, restatement or a short reporting period can make a percentage misleading. Preserve any not-meaningful reason in the tool result instead of turning it into zero.

What do PE, PB, ROE and ROCE leave out?

PE divides a dated share price by earnings per share. State whether earnings are trailing, annual or estimated, and check share adjustments. A negative earnings base does not support a normal positive PE comparison.

PB divides price by book value per share. A low book value or a business built on assets absent from the balance sheet can limit its use. Do not compare ratios that mix standalone equity with consolidated earnings.

ROE relates profit to equity; say whether the denominator uses closing or average equity. ROCE depends on the provider’s definitions of operating profit and capital employed. Check those formulas before comparing providers, and use the financial-institution template for banks rather than force an industrial-company metric onto them.

No ratio alone establishes why a company changed or what its shares should cost. Keep the calculation separate from a proposed explanation, then look for filing evidence that supports the explanation.

How do you keep the workflow auditable?

Save the input arguments, raw response, cache date, stale flag, period basis and units. Link issuer reports for checks you actually performed. A vendor description proves what a service claims, not the accuracy of a particular returned number.

Check both provider and upstream-source terms before storing, sharing or redistributing data. Public access does not by itself grant a commercial redistribution licence. This is a research method, not investment advice or an investment recommendation.