AI Translation for Public Services: When 200 Expert Interpreters Are Not Enough
- Apr 20
- 3 min read
Updated: 7 days ago
ISM Interprétariat came to us with a problem that many mission-driven organisations recognise: their clients were finding workarounds. And those workarounds were creating risk.
ISM is one of France's leading professional interpretation services, with around 200 highly trained interpreters working across healthcare, social services, and humanitarian contexts. Their work is excellent. But the need in the market is vast — and when clients couldn't get an interpreter quickly enough for a quick exchange, they were turning to generic AI tools: Google Translate, consumer chatbots, whatever was to hand.
For ISM, this was both an ethical concern and a strategic one. Low-quality, uncontrolled translation in sensitive settings … between a nurse and a patient, a social worker and a family in crisis … carries real risk. And it meant ISM's expert interpreters were being bypassed for exactly the moments where their presence matters most.
The brief ISM brought to Mission AI
The ethical challenge
Clients were using uncontrolled consumer AI tools for sensitive conversations — with no safeguards, no escalation path, and no awareness of the risks.
The capacity challenge
The need for language support far exceeds what 200 interpreters can provide. ISM needed a way to extend quality access without compromising standards.
Mission AI's role was not simply to build a tool. It was to understand ISM's problem deeply, design a solution that was technically sound and ethically defensible, and then help ISM tell that story in a way that could attract the funding needed to bring it to scale.
How Mission AI works
From problem to funded solution
Mission AI specialises in taking a client's real operational challenge, designing an AI-enabled solution that fits their context and values, and building the branding and donor narrative needed to fund it. Passerelle is that process in action.
The solution: a hybrid model that protects quality
The answer wasn't to replace ISM's interpreters with AI — that would have replicated the very problem they were trying to solve. Instead, Mission AI designed Passerelle as a purposeful hybrid: AI handles what is genuinely routine, and human interpreters take over the moment things become sensitive, complex, or unclear.
The goal was to give clients a quality-controlled path for simple exchanges — so that when an interpreter is needed, the conversation is already in good hands.
Client opens Passerelle and selects the context
Medical, social, or administrative — the right communication flow is triggered automatically. The person being supported hears consent spoken aloud in their language before anything begins.
AI supports the routine exchange
Simple questions, basic checks, and administrative intake are handled in real time across 20 languages. No waiting. No workaround.
Passerelle detects when to step aside
The tool monitors for warning signals — words linked to distress, extended pauses, repeated confusion. When they appear, the worker is prompted to connect directly to an ISM interpreter.
An ISM interpreter takes over
Sensitive conversations — health disclosures, emotional moments, anything requiring professional judgment — are always handled by a trained human. The AI steps back completely.
What this means for ISM's interpreters
Passerelle doesn't reduce the role of ISM's interpreters, it sharpens it. Instead of being called for a quick appointment reminder or a basic intake form, they are freed to focus on the work that genuinely requires their expertise: nuanced clinical conversations, trauma-sensitive exchanges, situations where a mistranslation could cause real harm.
The hybrid model also extends ISM's reach. Clients who previously turned to consumer AI now have a quality-controlled, ethically designed alternative — one that keeps them connected to ISM's professional network rather than bypassing it.
Does your organisation face a similar challenge?
Mission AI works with nonprofits and public service organisations to identify where AI can genuinely help, design solutions that fit their values and context, and build the case for funding. If you're navigating a gap between what your team can deliver and what your clients need, we'd like to hear about it.
Comments