Ask for mutual funds in plain English: how it works

The problem it solves

A fund screener speaks in fields: "volatility < 15 AND expense ratio < 1". Most people think in outcomes: "something safe that does not cost much". The Ask page bridges the two. You describe the funds you want, an AI writes the screener query, and the screener does the rest. The answer comes from computed data, not from the AI's opinion.

How to use it

1. Go to Ask for funds in plain English, or use the same box on the home page. You need to be signed in with a free account.
2. Type into "Describe the funds you want…", up to 300 characters, and press Find funds. The three "Try:" chips under the box run sample requests: "Safe large cap funds with low fees", "Consistent small cap funds over 5 years" and "Cheap index funds from HDFC".
3. You land in the query builder with the query filled in, a one-line note on how your words were read, and the results below.
4. Edit any condition and press Run query. With an account you can also Save screen.

The builder has the same feature built in. Type a sentence instead of a query and an "Ask AI" button appears beside "Run query".

What the AI is told to do

The translation follows written rules, the same each time:

The model can only use fields and values the screener knows: real categories, real fund houses. What it returns is checked by the screener's own parser before it runs. If it does not parse, you get an error, not a guess.

Three requests, run on 15 October 2026

The model's wording can vary slightly between two phrasings of one request, so the queries below are what the rules above produce. Each was run in the builder on the NAV of Wednesday 15 October 2026.

"Safe large cap funds with low fees" becomes sub category = Large Cap Fund AND volatility < 15 AND expense ratio < 1 AND plan = Direct. 12 funds matched. Ranked by 3-year CAGR the first was Invesco India Large Cap at 12.82%, with volatility of 14.48% and an expense ratio of 0.91%. Every fund on the list had a 1-year return below zero, which no part of the request asked about.

"Cheap index funds from HDFC" becomes sub category = Index Funds AND amc = HDFC AND expense ratio < 0.3 AND plan = Direct. 13 funds matched, and the top six by 3-year return were all target-maturity government bond index funds, led by HDFC Nifty G-Sec Apr 2029 Index Fund at 7.51%. "Index fund" covers debt indices too. Someone wanting an equity index fund has to say so, or add name = nifty 50 in the builder.

"Funds with only gold and silver" has no category of its own, so the rules match fund names: name = gold AND name = silver AND plan = Direct. That exact pair is one of the AI's written examples. 6 funds matched. Two had 3-year records: Edelweiss Gold and Silver ETF FoF at 38.62% a year and Motilal Oswal Gold and Silver Passive FoF at 35.85%. The other four are too young for a 3-year figure. The gold funds guide covers how these fund-of-funds hold the metal.

What it will not do

• No OR and no sorting. "Large cap or mid cap" becomes the single closest category, Large & Mid Cap Fund. "Top 10 by Sharpe" becomes a threshold, such as Sharpe above 0.6, not a top 10.
• No advice. It does not know your age, goal or tax bracket, and it will not choose between the funds it finds.
• Not every sentence is a search. A greeting or a general question about investing returns no query, with a message saying so.
• Not unlimited. Each account gets 30 translations per 15 minutes. Repeating a recent question uses a stored translation.

Where to go from here

If a translation lands close but not exact, edit it; that is the point of showing it. If you would rather start from someone else's filter, the ready-made screens gallery has 36. And once you have a list, shortlisting a fund with the screener shows how to cut it down using rolling returns and worst falls, while Consistent compounders reads one screen in depth.

All figures are computed from AMFI NAV history; the methodology page has the formulas. A fund passing a filter is not a recommendation to buy it.