Learning Objectives
After completing this unit, you'll be able to:
- Describe how the reasoning engine plans a response.
- Explain grounding and why it matters.
- Describe how the Einstein Trust Layer protects data.
How an Agent Reasons
When a request comes in, the reasoning engine follows a loop: understand the request, pick the right topic, choose actions, check the results, and respond. If the first action doesn't answer the question, the agent can try another.
Order the Steps
Put the steps of the agent's loop in order.
1Choose and run actions
2Understand the request
3Respond to the person
4Pick the matching topic
5Check the results
Grounding in Your Data
A model trained on the internet doesn't know your customers. closes that gap: the agent pulls relevant records, knowledge articles, or Data 360 profiles into the prompt so the answer reflects your business.
Where grounding data comes from
- Records the agent queries with an action.
- Knowledge articles that answer common questions.
- Data 360 profiles that unify data from other systems, including:
- purchase history
- service interactions
- marketing engagement
Tip
Better grounding beats a longer prompt. If answers drift, check what data the agent can reach before you rewrite instructions.
The Einstein Trust Layer
Every request passes through the Einstein Trust Layer. It masks sensitive data before it reaches the model, checks responses for toxicity, and keeps an audit trail. Your data isn't used to train the model.
Resources
Quiz
To complete this unit, you need to answer all the quiz questions correctly.
+100 Points
1Why ground an agent in your data?
2What does the Einstein Trust Layer do?