Multilingual AI agents
Agentforce responds in the person's language using the underlying LLM, augmented by detection, glossary enforcement, and translated Knowledge.
Agentforce
Agentforce changes everything
Organizations are no longer translating pages. They are translating conversations. Every assumption about content, governance, and quality has to be re-examined.
Conversation flow
Translation layer applied throughout
- 1
Person
- 2
Agentforce
- 3
Knowledge
- 4
CRM Data
- 5
Human Agent
↕ Runtime translation, glossary enforcement, language detection, human escalation
Six capabilities Agentforce multilingual requires
Runtime translation
Every conversation turn is translated as it happens, not pre-translated like static page text. Latency, quality, and audit must be handled per turn.
Human escalation
When Agentforce hands off to a human, the conversation language has to be preserved so the agent can either converse in that language or use live translation.
Knowledge retrieval
Translated retrieval over Knowledge gives the agent grounded answers in the right language, with attribution back to the source article version.
Language detection
Detection runs on every inbound message, combined with person profile preferences and channel metadata to set the conversation language.
Translation governance
Glossaries, review workflows, and quality scoring keep terminology consistent across millions of generated conversation turns.
Multilingual person engagement
Build a multilingual Salesforce experience that serves every person
Learn from Hyde Housing's journey and accelerate your own multilingual transformation across Experience Cloud, Agentforce, Mobile Publisher, Knowledge, and messaging channels.