Spanish-language content moderation and classifier training for a Latin American platform

The client operates a social and user-generated content platform with a substantial user base across Colombia, Mexico, Argentina, Chile, Peru and Central America, alongside a Brazilian Portuguese segment. Content volume is high and predominantly Spanish, with strong regional variation in slang, humour and cultural reference.

The platform's automated moderation performed poorly on Latin American Spanish. The toxicity classifier, trained largely on English and Castilian data, achieved 72.0% precision and 61.4% recall on Latin American content. Low recall meant harmful content persisted; low precision meant benign content was removed, generating complaints and accusations of arbitrary censorship. The failure was concentrated in exactly the content that matters most. Regional slang, coded language and context-dependent insults were systematically missed, while vocabulary benign in one country and offensive in another was handled inconsistently. Response time was compounding harm, with median time from report to action at 45 minutes and exceeding four hours during spikes. Moderator wellbeing had become a retention crisis, with annualised attrition above 70% under the previous arrangement, producing a permanently inexperienced workforce and degraded decision quality.

Corpshore Colombia proposed Cali for its Latin American demographic mix and its growing services labour market, and proposed a moderation model built around regional decision routing rather than uniform staffing.

The wellbeing framework was decisive. Corpshore's documented approach, mandatory rotation off high-intensity queues, scheduled decompression, on-site counselling and peer support, was assessed as materially more developed than competing bids, and the client had concluded its attrition problem was the root cause of its quality problem. Corpshore also proposed pairing human moderation with classifier training in a single engagement.

Sixty-six moderators in Cali drawn from across Latin American nationalities, eight AI training specialists on classifier improvement, five trust and safety leads and two wellbeing counsellors on site. Coverage is 24 hours on rotating shifts. All moderators complete a six-week onboarding covering policy, regional context, decision frameworks, escalation and wellbeing practice before handling live content, with separate training and rotation limits for the most difficult categories.

Regional decision routing. Content routes to a moderator familiar with the originating region's language and cultural context rather than a generic Spanish queue, addressing the inconsistency uniform staffing cannot solve.

Moderator decisions as training data. Every decision, with its regional context and reasoning code, feeds the classifier training set. Six retraining cycles were completed, each incorporating the prior cycle's human decisions.

AI-augmented triage, not replacement. The classifier pre-scores and prioritises rather than acting autonomously above a narrow high-confidence band, keeping human judgement on ambiguous content while removing routine volume, which produced the response-time improvement without a precision cost.

Wellbeing as an operating requirement. Rotation limits, decompression, on-site counselling and peer support are scheduled and enforced. Attrition on this account is 21% annualised against the client's prior 70%, and the client attributes the decision-quality improvement primarily to tenure.

Spanish-language classifier precision and recall

Classifier precision rose from 72.0% to 94.0% and recall from 61.4% to 93.1% across six retraining cycles. The precision improvement reduced wrongful removals, the source of the platform's public criticism. Median time from report to action fell from 45 minutes to 4 minutes, sustained through spikes because AI triage absorbs routine volume that previously queued behind ambiguous cases. Moderator attrition fell from over 70% to 21%, and decision accuracy against trust and safety lead review rose from 79% to 96%.

The regionally annotated training corpus and reasoning-code taxonomy are client-owned and will support classifier development independent of any vendor. The wellbeing framework has been adopted as a Corpshore Colombia standard for accounts handling distressing content. The engagement has extended to Brazilian Portuguese moderation and to red-teaming the client's generative content features.