AI-201 Tests Which Agent Layer to Change — Not Louder Prompt Instructions
For builders who configure agents, prompt templates and grounding in Salesforce. When an agent misbehaves, the exam wants the right layer fixed: topic description, filter, scripted step, retriever or permissions, rather than more wording.
Check Your Readiness →Most candidates understand Salesforce Certified Agentforce Specialist (AI-201) concepts — and still fail. This exam tests how you apply knowledge under pressure.
Most items describe an agent or template producing the wrong result and ask what to change. Weights come from secondary reproductions of the Spring '26 guide; one source puts Governance at 5%. The older five-section Summer '25 outline with Development Lifecycle at 20% is superseded.
Activate it again or use the Agent API
Website deployment needs a Messaging for In-App and Web channel, an Omni-Channel flow routing to the agent and an embedded deployment. The Agent API is for custom apps with no standard channel.
Open the Command Center dashboard
Aggregates cannot explain one conversation. Session tracing shows the topic chosen, actions run, inputs, outputs and errors. The Trust Layer audit trail covers prompts, masking and toxicity, not action execution.
Clear the cache and wait for it to update
Only the active version runs. Edits create a new version that must be activated, and after deployment you confirm activation in the target org; previewing against several records first verifies grounding.
Use A2A for both integrations
The calculator is a tool, so expose it through MCP or a standard action under Agentforce's control. The partner agent owns its reasoning and multi-step planning, which is what Agent2Agent (A2A) is for.
Redeploy the agent metadata alone
Agent metadata depends on flows, Apex, prompt templates and permission sets deploying with it. The agent user is data, so create and permission it in production, and move Data 360 indexes and retrievers with data kits.
Instructions are interpreted by the model and can be ignored. Rules that must always hold belong in filters on variables, which gate topics and actions, or in deterministic steps written in Agent Script for hybrid reasoning.
If the retriever returns the wrong passages, any model answers from them. Fix retrieval: adjust chunk size, use hybrid search for SKUs and part numbers, add retriever filters on metadata, or refresh the search index.
Field Generation binds to one field on a Lightning record page with Dynamic Forms. Flex has custom inputs and no UI placement, so it is called from flows, Apex or agent actions. Sales Email and Record Summary are standard types too.
Zero data retention stops model providers storing or training on prompts. Data masking replaces sensitive values before the prompt leaves Salesforce and restores them in the response. A personal-data exposure requirement points to masking.
The agent user should get least-privilege permission sets covering only each action's objects and fields. Also review system context flows, which bypass the agent user's restrictions, before adding them as actions.
Strengthen weak areas with exam-style practice questions and detailed explanations.
Simulate the real exam experience and assess your readiness under timed conditions.
Review key concepts, objectives, and exam topics in one place.
Get personalized explanations, learning recommendations, and instant answers.
Follow a structured learning path designed to help you prepare efficiently.
AI-201 asks which agent layer to fix. See whether you pick topics, filters or retrievers correctly.