AIMI Tier 1 Agent
The AIMI Tier 1 Agent is the orchestration agent at the center of MIC AI. Working from an Interaction, it coordinates four subagents, one for each kind of work, to produce the recommendations needed to handle the request:
- Interaction Management Agent, recommends an Interaction summary.
- Request Management Agent, recommends the Request(s) to create and the documents to fulfill them.
- Adverse Event Management Agent, evaluates the Interaction for a potential adverse event.
- Product Quality Complaint Management Agent, evaluates the Interaction for a potential product quality complaint.
The Tier 1 Agent does not create records on its own. Each subagent writes its recommendation to an AIMI Response Log (MAI_AIMI_Response_Log__c) record, which a human Medical Information specialist reviews and approves in the AIMI Recommends panel. Specialists can also run individual subagent prompts on demand through Ask AIMI.
The assembled Tier 1 Agent is configured in your org; it is not delivered ready-made by the package. The AIMI package ships the building blocks (the subagents, their prompt templates, and their flows) as reusable assets in the Agentforce Builder Asset Library; an administrator assembles them into the Tier 1 orchestration agent. See Set up MIC AI and the AIMI Tier 1 Agent Builder script.
How orchestration works
Each subagent pairs a GenAI prompt template, which produces the recommendation, with flows that read the Interaction and write the result. The orchestration agent runs a subagent, takes the structured JSON the prompt template returns, and its flows map that JSON onto the fields of an AIMI Response Log record. Prompt templates ground their output in the Interaction's own data and are told how to handle missing or ambiguous information, so recommendations stay grounded rather than improvised.
The order the subagents run in is driven by MIC, through the ordered utterances and prompt-option configuration it sends the agent, rather than by the agent planning the sequence itself. For example: recommend the Interaction summary, then the Request(s), then evaluate for an adverse event, then for a product quality complaint. The active subagents, their prompt templates, and their order are configured through MAI_AIMI_Prompt_Option__mdt records (MAI_Config_Type__c = Agent). See Set up MIC AI.
The Tier 1 Agent skips re-running on an Interaction that already has response logs, so changing context (such as Country) on an already-processed Interaction does not duplicate recommendations.
In this section
Each subagent is documented on its own page (linked above). For the fields each one writes and what gets created on accept, see Recommendation types.