Bilingual is not a checkbox: how we assess and maintain language capability

Almost every outsourcing provider in the Caribbean and Latin America describes itself as bilingual. Very few will tell you what that means.

The word covers an enormous range. It can mean an agent who handles a complex complaint in a second language with complete fluency, and it can mean an agent who can follow a script until the customer says something unexpected. Both get described the same way in a proposal.

This is a real commercial problem, not a pedantic one. Language capability is the single most common point of failure in multilingual outsourcing, and it usually fails in a specific way: capability that was adequate for the pilot proves inadequate for the actual work, and by then the team is built.

The framework we use

We assess against the Common European Framework of Reference for Languages, the CEFR. It is the standard used across European education and employment and it defines six levels from A1 to C2. For our purposes the relevant range is B1 to C2.

B1 handles routine, predictable interactions. The speaker can deal with most situations likely to arise in familiar contexts and can describe experiences and give reasons. B1 is adequate for structured transactional work with a defined scope.

B2 handles complexity and abstraction. The speaker can interact with a degree of fluency and spontaneity that makes regular interaction with native speakers possible without strain on either side. B2 is our minimum for most customer-facing voice work.

C1 handles nuance, implicit meaning and difficult conversations. The speaker can express ideas fluently without obvious searching for expressions and can use language flexibly for social, academic and professional purposes. C1 is our requirement for team leaders, trainers, escalation handling and any client-facing role.

C2 approaches the capability of an educated native speaker. Required for a small number of specialist roles including linguistic quality review.

What we assess, and what we do not

Assessment covers four dimensions: speaking, listening, reading and writing. They are recorded separately because they do not move together. An agent can be C1 in listening and B2 in writing, and a chat or email role weights those differently than a voice role does.

We assess register and appropriateness alongside raw proficiency. This is the part that generic language testing misses, and it is frequently what causes multilingual engagements to disappoint. A speaker can be C1 by any grammatical measure and still sound wrong to the customer, because register, directness, formality and idiom vary between regions in ways that proficiency scales do not capture.

For a Canadian insurer serving Quebec, Metropolitan French delivered at C1 is not the same product as Canadian French delivered at C1. For a United States retailer serving Hispanic customers, Castilian Spanish is not the same product as Latin American Spanish. We assess for the variant the engagement requires, not for the language in general.

Assessment at hire

Every candidate for a language-dependent role is assessed before an offer, not after. The assessment includes a structured conversation with an assessor, a written exercise, a listening comprehension component and a role-relevant scenario.

The scenario matters. General proficiency testing tells you whether someone can hold a conversation. It does not tell you whether they can hold a difficult conversation with an angry customer about a billing dispute while navigating three systems. The scenario component tests the language in the shape the work will actually take.

Candidates are recorded at the level they demonstrate. We do not round up. A strong B2 is recorded as B2.

Maintenance in role

Language capability is not static. It improves with use and degrades without it, and an agent who joined at B2 three years ago is not necessarily B2 today, in either direction.

Proficiency is reviewed periodically in role. Where an agent has improved, the record is updated and it becomes relevant to progression, since C1 is a requirement for team leader and trainer positions. Where capability has drifted, targeted coaching follows.

Quality monitoring includes a language dimension separately from a process dimension, so a scorecard distinguishes between an agent who followed the wrong process and an agent whose language let them down. Those require different interventions and conflating them fixes neither.

What this means for a buyer

Three practical implications.

First, ask any provider what level they are staffing to, on what scale, assessed how. A provider that cannot answer in specific terms is telling you something.

Second, specify the level you need rather than asking for bilingual. Different parts of one engagement often need different levels, and paying C1 rates for B1 work is as much a failure as the reverse.

Third, ask about the variant, not just the language. If your customers are in Quebec, Mexico or the Dominican Republic, the regional register is part of the specification.