Grounding techniques: merge fields, flows and retrievers

What the exam expects

• Choose the grounding technique for a scenario: record merge fields, related list merge fields, a template-triggered prompt flow, Apex, or a Data 360 retriever.
• Explain why logic such as filtering, sorting or totals belongs in a flow or Apex resource, never in instructions asking the model to do it.
• Recognise unstructured, high-volume or frequently changing content as a retrieval-augmented generation (RAG) job.
• Wire a retriever into a template: create and activate it in AI Models, insert it, set search text, output fields and number of results, and use a dynamic pre-filter when the filter value is only known at run time.
• Key terms: grounding (adding relevant data to a prompt); merge field (a placeholder resolved with Salesforce data); template-triggered prompt flow (the flow type that returns prompt instructions to a template); retriever (searches a Data 360 search index and returns relevant content); dynamic pre-filter (a retriever filter whose value the template supplies at run time).

Key facts

Check yourself