Regional Language Interview Screening: Hire in the Job's Language

Regional language interview screening can go wrong before the first question. The job is selling a personal loan over the phone in Hindi, or working a beat of kiranas in Telugu, or running a till in a Chennai store, but the first-round call is in English because that is the recruiter's script. This article is for the employer who suspects that default is hiding capable applicants and wants to fix it without hiring a recruiter for every language on the map.

For a job performed in Marathi, the screen needs evidence of the candidate working in Marathi. Start with three things: a statement of which language each task needs, a way to review answers when your recruiter does not speak that language, and written scoring anchors calibrated with fluent reviewers. The matrix and rubric below give you a starting point.

What the English screen actually selects for

An English-only conversation tells you how a candidate handles that conversation in English. It does not establish how they would handle the same task in Marathi or Tamil. India's 2011 Census table on bilingualism and trilingualism distinguishes a person's mother tongue from additional languages spoken. That historical record is useful context, but it is not a measure of today's applicants or their job-level fluency. Those need a task-based check.

Consider a worked example. A Hindi telecalling floor posts twenty seats and draws 400 applicants. The recruiter has a day, so the first call is ninety seconds, in English, and the thing being graded is "communication". A candidate who hesitates in English sounds less confident, gets a low mark, and is out. On the floor, in Hindi, that same person might have been the best closer in the batch. The recruiter never finds out, because the test measured the wrong language. The same mismatch can occur on a field beat in Telangana where the retailer speaks Telugu, or at a Chennai store where the customer speaks Tamil and the panel's form is in English.

An international voice process or a role that writes to clients may genuinely need English and should be screened for it directly. The failure is using English as a proxy for polish or education when the task requires another language. Compare your screening scores with later job performance to see whether the test measures the work. It is the same selection-by-logistics problem that can make the bulk hiring process break at screening rather than at sourcing.

Five language skills to separate

The phrase that causes confusion is "good communication skills in Hindi and English". It collapses five different requirements into one line.

The telecaller and field sales question banks already split two of these: the language a candidate can sell in, and the language they could handle an angry customer in. A fluent sales pitch does not establish complaint-handling ability. Understanding, reading and writing need their own checks too, sometimes in a different language.

Speaking is opening a call, pitching a plan, explaining a scheme: the language the customer buys in. Understanding is hearing a fast, irritated customer, possibly using a different dialect, and getting the facts right; assess it separately from speaking. Reading is the script, the offer sheet or the app, which may be in English. Writing is the CRM remark, the daily sales report or the ticket note; check which languages and scripts the receiving team can use. De-escalating is changing pace and tone under stress: staying calm while someone is upset, in a language they understand.

In the example below, an outbound telecaller needs Full-level Hindi for speaking and understanding, Working-level Hindi for routine complaints, and Basic reading and writing. If the job description only says "Hindi and English", a recruiter can spend the screen on English and miss the Hindi tasks entirely.

The language-requirement matrix

Here is the instrument. Score each task at one of three levels, per language.

• Basic (B): gets through a scripted exchange, may need a repeat, can read a line and type a short note.
• Working (W): handles the normal day unscripted, at normal speed, without losing the thread.
• Full (F): handles the abnormal day: fast speech, dialect, ambiguity, an angry counterpart, without losing structure or register.

Then fill the grid for the role you are actually hiring. These are worked examples, not universal requirements or certified proficiency levels. Confirm each cell with the manager who knows the work, the customer languages and the tools.

In several of these examples, English is needed only for reading and writing. A conversational English score cannot establish either. Most regional-language requirements involve spoken tasks, which need a live exercise rather than a fluency checkbox. Collections and nursing also need a separate de-escalation task; a rehearsed pitch will not show that skill.

For role-specific exercises, use the retail staff question bank and the staff nurse question bank. The nurse's language screen sits alongside a separate in-person clinical assessment.

To build your own row:

1. For each of the five tasks, write down who is on the other side: a customer, a retailer, a patient's relative, your manager, an app.
2. Name the language that counterpart actually uses and the level the task needs. If the task is not in the job, write "none" and stop screening for it.
3. Everything marked W or F gets a live task in that language, with difficulty matched to the level.
4. Everything marked B gets one short task matched to that cell: explain a line, type a remark or ask for directions. Check it against the Basic requirement.
5. Whatever you were about to require that is not in the grid, usually "fluent English", is the proxy. Delete it.

How to write a language requirement that is honest

The grid turns into a job description line almost mechanically: one sentence per language, stating the task and the level, must before helpful.

Before: "Excellent communication skills. Fluent in English and Hindi."

After, for the Hindi telecalling floor: "Must: sell and handle complaints over the phone in Hindi. Basic: read an English offer sheet and type short call notes in English or Roman Hindi. Hinglish is fine."

After, for the Telugu FMCG beat: "Must: sell to and negotiate with kirana owners in Telugu. Basic: read an English price and scheme sheet and file a daily report on the app."

After, for the Chennai store: "Must: serve and handle complaints in Tamil. Helpful: basic Hindi and English for customers who do not speak Tamil. No writing beyond the billing screen."

Three habits follow. Share the vacancy in the language and channels your intended applicants use, wherever the posting format allows it. Say that code-mixing is acceptable when the work permits it. And never write "preferred" next to English when you mean "we will decline you without it".

Running a screen in a language your recruiter does not speak

Suppose a Bengaluru recruiting team is hiring for Kannada, Tamil, Telugu and Hindi floors in the same quarter, and the team speaks two of those well. Here are four ways to cover the gap, and the work each requires:

The first option's cost is easy to underestimate: sixty Tamil applicants waiting on one Tamil-speaking recruiter can form a bottleneck even while another language queue moves quickly. Run a separate scenario for each language in the time-to-hire calculator, using the recruiter capacity available to that queue.

Whichever you pick, the protocol is the same, and it is the protocol that makes the score trustworthy, not the method:

1. Review translations before the hiring batch. Have fluent reviewers check that the core tasks, difficulty and scoring anchors are equivalent. Keep follow-ups tied to the same criteria.
2. Run the Working- and Full-level spoken tasks in the job language, live. The sixty-second pitch, the summarise-this-complaint task, the angry-customer role-play. They are performed, not described.
3. Run reading and writing checks separately. Ask the candidate to read a short offer sheet and explain it, then type a sample call note. A phone conversation alone cannot test reading or typing.
4. Score against written anchors while listening, either on the call or from its recording, rather than from memory afterwards.
5. Double-score a sample. As a starting check, a second fluent reviewer re-scores one call in ten per language. Repeated gaps of more than a point call for a review of translations, audio quality and scoring anchors.
6. Keep the evidence. Explain recording and obtain permission before recording a call. Save the relevant answer and scoring reason so another reviewer can check the decision.

The split that makes this workable is between what was said and how it was said. From a checked translation, a reviewer who speaks no Telugu can assess content: a number with a unit, a specific outlet named, the structure of the pitch, whether the question was answered. Translation errors can still change those facts, so uncertain or decisive passages need a fluent review of the original. Pace, tone, comprehension at conversational speed and spoken de-escalation require listening to the audio. Do not infer them from a translated transcript or assume an AI transcript score measures them.

Using one fluency rubric across languages

Scoring "fluency" as an impression leaves room for accent and confidence to outweigh whether the candidate completed the task. Start with five observable behaviours, then calibrate their translated anchors with fluent reviewers in each language. A shared rubric is a starting point; it does not make scores comparable automatically.

Weight it from the grid. A collections seat with Marathi F in de-escalation puts more weight on register switch and comprehension. A Basic-language check can put more weight on clarification and recovery. If noise, a poor connection or an unsuitable test format prevents a fair assessment, repeat the task in a suitable format instead of treating the interruption as a language failure.

Score whether the intended customer can understand the answer, rather than whether the accent resembles the interviewer's. Accept code-mixing when the customer's language and the task allow it. At Basic level, judge whether the candidate gets the meaning across accurately; do not silently raise the bar to polished grammar. Confidence by itself is not a language skill.

The Council of Europe's CEFR provides an external reference: six proficiency levels, A1 to C2, described through what a person can do. The Basic, Working and Full labels and the five behaviours above are a proposed hiring rubric, not CEFR levels or a validated substitute for a language assessment.

The fairness problem: different interviewers, different languages

Multiple languages add another source of variation to the differences between interviewers. Imagine one panel scoring strictly and another generously: two shortlists for the same role can look different because the bar lives in different heads. Translations and task difficulty can introduce differences too.

Four checks catch it:

• One rubric, translated anchors. The 1, 3 and 5 descriptions exist in each interview language, in writing, and every interviewer has read them before the first call.
• Compare score patterns. As an initial review, inspect the first fifty calls per language, or all calls in a smaller batch. A gap between groups is a reason to inspect tasks, panels and applicant experience; it is not proof of bias or different ability.
• Cross-score the same content criteria. Compare a fluent reviewer's content score with a second review of a checked translation. Investigate gaps in the original audio, translation and scoring anchors. A gap alone does not establish that someone scored accent.
• Keep the reason. Every decline should identify the job-related task and the anchor it missed. "Did not sound confident" does not explain which requirement was unmet.

A structured screen in the job's language gives reviewers relevant evidence to check. Its fairness still depends on the tasks, translation quality and how decisions are made. The automated phone screening guide explains what to look for when evaluating tools for this stage.

How EasyInterview screens in eleven languages

EasyInterview supports live AI phone interviews in eleven languages: English, Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Marathi, Gujarati, Punjabi and Odia. Set the allowed call languages for the job and check that the candidate's selection matches the task you need to assess.

You share the job's apply link through your sourcing channels. The apply form records consent and runs the configured knockout rules; a recruiter can review an exclusion and use "Include anyway". Existing applicants can also enter through CSV import, which records existing consent rather than collecting it and does not run the apply form's knockout checks. The Naukri applicant screening workflow covers that distinction.

The recruiter schedules eligible applicants with consent on file. Calls run within the workspace's calling window, using the selected language and the job's question set. For spoken-language tasks, pilot the translated questions with a fluent reviewer before using the results to make hiring decisions. Reading and writing checks still need a separate exercise.

Reports link scored answers to transcript evidence, alongside the transcript and recording. The scoring dimensions depend on the interview preset: the communication rubric includes clarity, listening relevance, answer structure, concision, thought process and adaptability. These are transcript-based scores, not a certified spoken-fluency assessment. A fluent person needs to listen to the recording for pace, tone and the language behaviours in this article's rubric.

The hiring team reviews the evidence and decides whom to advance. Outcome emails are optional at handoff and require an email address. Start with one live role alongside your existing screen: compare task scores and reviewer disagreements first, then follow later job performance and retention. The sample report gives a preview of the reporting format; ask to review a pilot call in each language you plan to use.

Frequently asked questions

Should I interview candidates in English or in their regional language?

Use the language required for each job task, then check other languages separately at the level the grid specifies. For the Hindi telecalling example, run the pitch, comprehension task and complaint role-play in Hindi. Check reading and note-writing with separate written exercises in the languages the team uses.

How do I assess a candidate's language skills if I don't speak the language?

Split content from spoken delivery. You can assess facts and answer structure from a checked translation, with a fluent reviewer resolving uncertain passages in the original. Pace, tone and spoken de-escalation require a fluent reviewer to listen to the call or recording. A translated transcript alone cannot establish them.

Is Hinglish or code-mixing a problem in a screen?

Judge it against the actual customer and task. If the customer understands the mixed-language answer and the role permits it, switching one word need not reduce the score. If the task requires a particular language throughout, state that requirement in advance and test it consistently.

What level of English does a telecaller actually need?

Check the floor's actual scripts, customer queue and note-writing requirements. A regional-language floor may need only basic English reading and writing, while an English-language voice process needs live speaking and comprehension checks too. Set each requirement from the work rather than the job title.

Which Indian languages can an AI phone screen run in?

EasyInterview runs live phone interviews in eleven languages: English, Hindi, Bengali, Tamil, Telugu, Kannada, Malayalam, Marathi, Gujarati, Punjabi and Odia. If your market speaks a language outside that list, the protocol in this article still applies with a human speaker on the call.

Screen the job, not the schooling

An English screen for a regional-language job is a decision, even when nobody made it. Fill in the grid for the role, write the requirement it gives you, and test the required tasks in those languages. Calibrate the written anchors with fluent reviewers, then check whether the screen predicts the work you hired people to do.

Book a demo to see EasyInterview take one regional-language req from an apply link to a ranked, evidence-scored shortlist.