AI Memories

Every business has its own unique “language” and rules. AI Memories allow you to “train” the AI on these specifics so it can answer questions like a seasoned employee.

---

What is an AI Memory?

An AI Memory is a piece of context injected into the AI’s conversation. They represent business logic that isn’t stored in the database columns themselves.

Examples of Memories

• Business Rules: “A order is only considered ‘Successful’ if the status is ‘Delivered’ and the payment is ‘Completed’.”
• Acronyms: “When I say ‘MQL’, it refers to the ‘Marketing_Qualified_Leads’ table.”
• Exclusions: “Always exclude internal test accounts (user_id < 100) from revenue reports.”
• Calculations: “Net Profit is calculated as (Total_Revenue - Total_Tax - Shipping_Cost).”

---

The Approval Workflow

To ensure accuracy, AI Memories follow a simple governance workflow:

1. Pending: When you first add a memory, it is in ‘Pending’ status.
2. Review: An administrator reviews the memory to ensure the logic is correct.
3. Active: Once approved, the memory is instantly applied to all future AI Chat conversations and Automations.

---

How to Add a Memory

1. Navigate to the AI Memories page in your project.
2. Click Add New Memory.
3. Define the Rule: Write the rule in plain English (e.g., “Always filter by active=true for user queries”).
4. Save: The memory will enter the ‘Pending’ queue for approval.

---

Tips for Effective Memories

• Keep them Focused: One rule per memory is better than one giant paragraph.
• Use Real Column Names: If your rule refers to a specific column like trx_status, mention it by name.
• Test in Chat: After approving a memory, ask a question and see how the AI applies the logic in the “Thought” process.

Expert Mode

AnalytAI also uses the Project Context to learn basic table meanings. Use Memories for custom logic, and let the Automatic Description Generator handle the basic definitions.