Screener AI vs Altys: Company Q&A or Continuous Research System?

Screener AI and Altys solve different parts of the Indian equity-research workflow. Screener AI makes it easier to ask questions of a company’s annual reports, concalls and other documents. Altys is being built as the research system around that reading: source-linked data, point-in-time history, models, screening, portfolio context and continuous company monitoring.

The distinction is more useful than asking which product has the better chatbot. If every serious platform can add a capable language model, the durable advantage moves to the evidence, workflow and memory underneath it.

This comparison reflects the publicly described products as of 27 August 2026. Products change, so investment teams should verify current capabilities directly before making a decision.

What Screener AI does

Screener AI describes itself as an AI assistant with direct access to company documents. An investor can ask a question in ordinary language and use the answer to navigate annual reports, earnings-call material and other company disclosures.

That is a valuable job. Company documents are long, terminology changes and the relevant explanation may be buried in a note or management response. A grounded question-and-answer layer can shorten the distance between “what happened?” and the paragraph that explains it.

Screener.in also surrounds that AI layer with a widely used company-research interface. A public HDFC Bank company page illustrates the broader product: financial statements, ratios, peer comparison, quarterly results, shareholding and company documents in one place.

For an investor researching a company one question at a time, this is a sensible and accessible workflow:

1. open the company page;
2. inspect the financial history;
3. ask the document layer a specific question;
4. follow the source;
5. form a view.

The question is what happens after the answer.

What Altys is designed to do

Altys is designed for investment teams that need the research process to continue after a document has been read.

It connects Indian company filings, financial statements, concalls, management guidance, company-specific KPIs, ownership, valuation, factor and forensic data with the team’s screening rules, models, notes, portfolio exposures and monitoring conditions.

The intended workflow is not simply “ask a question about HDFC Bank.” It is closer to:

• detect what changed in the latest disclosure;
• compare it with what management previously guided;
• update the relevant financial and operating series;
• identify whether an analyst model or scorecard needs attention;
• connect the change with the written investment thesis;
• assess its significance in the context of portfolio exposure;
• preserve the decision and later compare expectations with outcomes.

AI can help read and organise that evidence. Code should calculate the financial metrics. The underlying records should preserve the source, basis and date on which each fact became knowable. The investment team then applies judgment.

A practical comparison

Research need	Screener AI	Altys
Primary job	Ask questions of company documents	Operate an end-to-end research and monitoring workflow
Company financials	Available through Screener.in company pages	Source-linked Indian company financials and derived research data
Document Q&A	Core AI use case	One part of a wider company-intelligence workflow
Point-in-time research	Not the central public positioning	Designed to preserve what was knowable at a historical date
Screening and backtesting	Screener is well known for screens; test the exact historical-data requirement	Connects strategy rules, factors and point-in-time backtesting infrastructure
Financial models	Research inputs can be exported and used separately	Intended to connect evidence, assumptions, forecasts and model revisions
Portfolio context	Primarily company-centric	Company events can be assessed against holdings and related exposures
Continuous monitoring	An analyst returns to ask or inspect	Monitoring rules are intended to surface material company and thesis changes
Best fit	Investors who want fast company research and document answers	PMS, AIF, family-office and research teams building a repeatable research system

This is not a claim that one product is universally better. A well-designed document assistant can be the right tool for an investor who wants to understand a company quickly. A research desk covering dozens or hundreds of companies has a different operational problem.

The difference between an answer and a workflow

Imagine two analysts receive the same earnings release.

The first asks an AI assistant, “Why did margins fall this quarter?” The assistant locates management’s explanation and returns a concise, sourced answer. That is already much faster than reading the whole transcript blindly.

The second analyst’s system also knows:

• the margin range management guided six months ago;
• the analyst’s forecast and original assumption;
• whether this is the first miss or part of a pattern;
• which input cost or business segment usually drives the margin;
• how much of the portfolio is exposed to the same factor;
• what action the investment committee said it would take if the condition persisted.

Both analysts may use the same underlying language model. The second has more context, a defined workflow and accumulated research memory. That surrounding system is the advantage.

This is the argument behind our article on who wins when everyone has AI. The foundation model becomes widely available. Clean data, proprietary context, repeatable workflows and feedback loops remain harder to reproduce.

Why point-in-time data changes the answer

Historical research is especially sensitive to the data layer.

The period a figure describes and the date it became public are different. Companies restate prior periods, recast segments and separate discontinued operations. A research or backtesting system that only stores the latest version of history can accidentally use information that was not available on the decision date.

That is why point-in-time data is not a cosmetic product feature. It determines whether a historical screen, factor study or decision review is reproducible. The right evaluation question is not only “does the product show ten years of financials?” It is “can it show the version an analyst could actually have known on a specified date?”

Our point-in-time financial data buyer’s guide gives investment teams a practical way to test this.

Why monitoring matters after the research is complete

Most investment theses do not fail because the initial memo had too few pages. They fail because the evidence changes gradually and the process does not bring the right change back to the analyst.

A useful monitoring layer should distinguish between activity and significance. “New filing received” is activity. “The company reduced the capacity target that supported the growth assumption, and this is the second revision in three quarters” is a research signal.

Altys company monitoring is intended to connect the new source with the historical series, management commitment, model assumption and portfolio context. The purpose is not more alerts. It is fewer, better investigation prompts.

How an investment team should evaluate both

Run a small, controlled test using companies the team already understands.

Test the answer

Ask five questions whose answers sit in different documents. Check whether the response is faithful to the source, whether the citation supports the exact claim and whether missing evidence is admitted clearly.

Test the number

Choose a metric affected by a demerger, restatement or basis change. Check whether the product identifies the basis and lets the analyst reconstruct what was known at the time.

Test the workflow

Record a thesis condition and a management promise. Introduce a later filing that changes one of them. See whether the system merely stores the filing or connects it with the earlier research.

Test the portfolio context

Use a company event that affects several holdings through the same sector, commodity, currency or factor exposure. Check whether the product can move from a company answer to a portfolio question.

Test the memory

Return three months later. Can the system show what the team believed, which evidence supported it, what changed and whether the original forecast was accurate?

The simplest way to choose

Choose around the job you need done.

If the job is understand this company and question its documents, Screener AI is a useful tool to evaluate.

If the job is build a repeatable research process across companies, models and a portfolio, then keep that process current, evaluate Altys as the system around the AI.

The long-term edge is unlikely to be the ability to generate a polished answer. That capability will spread. The edge starts where the answer becomes evidence, the evidence enters a workflow, the workflow remembers the thesis and the system learns from what happened next.