> ## Documentation Index > Fetch the complete documentation index at: https://docs.spinnable.ai/llms.txt > Use this file to discover all available pages before exploring further. # How Workers Learn > Understanding how AI workers improve through conversation and memory ## Overview Spinnable AI workers learn and improve through two complementary systems: 1. **Conversational Refinement** - Real-time learning during interactions 2. **Long-term Memory** - Persistent knowledge that carries across conversations This dual approach allows workers to both adapt immediately to feedback and retain important information over time. *** ## Conversational Refinement ### How It Works During a conversation, workers learn from: * **Corrections** you provide * **Clarifications** you offer * **Preferences** you express * **Context** you share This learning is active throughout the entire conversation, allowing the worker to adjust its approach in real-time. ### Example ``` You: "Create a sales report" Worker: [Creates generic report] You: "Actually, I prefer bar charts over pie charts, and always include quarter-over-quarter comparison" Worker: [Adjusts current report and remembers for this conversation] ``` The worker will apply these preferences for the rest of the conversation. ### Limitations * Learning persists only within the current conversation * Resets when you start a new conversation * Best for immediate corrections and session-specific preferences *** ## Long-term Memory ### What Gets Remembered Workers store important information in their memory knowledge base: * Communication style preferences * Favorite tools and formats * Decision-making criteria * Work schedule and availability * Team structure and roles * Project backgrounds * Company policies * Standard procedures * Industry-specific terminology * Technical specifications * Best practices * Historical decisions and rationale * How you like tasks completed * Common workflows * Quality standards * Exception handling ### How Memory Is Created During conversations, the worker recognizes potentially valuable information worth remembering. The worker asks: "Should I remember this for future conversations?" You approve or decline the memory suggestion. Approved information is stored in the worker's knowledge base. ### Memory in Action **First Conversation:** ``` You: "When I ask for reports, I want them in PowerPoint format with our company template, and always include an executive summary on the first slide" Worker: "I'll remember that preference. Should I store this for all future reports?" You: "Yes, please" Worker: ✓ Memory saved: Report format preferences ``` **Later Conversation (Days or Weeks Later):** ``` You: "Create a Q4 sales report" Worker: "I'll create that in PowerPoint format with an executive summary on the first slide, using your company template." [Automatically applies remembered preferences] ``` *** ## Memory Management ### Viewing Memories Access your worker's stored memories through: * The **Knowledge Base** section in worker settings * Direct request: "What do you remember about me?" ### Updating Memories Memories can be modified when circumstances change: ``` You: "I no longer need executive summaries on reports" Worker: "I'll update my memory. Should I remove the executive summary requirement?" You: "Yes" Worker: ✓ Memory updated: Report format preferences ``` ### Deleting Memories Remove outdated or incorrect information: * Manually through the knowledge base interface * By asking: "Forget my preference about \[topic]" *** ## Best Practices Clearly state when information should be remembered long-term vs. just for the current task. When the worker asks to save information, confirm to ensure important preferences are retained. Periodically review stored memories to keep them current and relevant. Help workers understand why something matters for better memory decisions. *** ## Privacy & Control You have complete control over what gets stored in worker memory. Workers will always ask before saving new information. ### Your Rights * **Full visibility** into all stored memories * **Edit access** to modify any memory * **Delete capability** for any stored information * **Approval required** before new memories are created ### Data Handling * Memories are specific to each worker * Information is encrypted and secured * You can export or delete all memories at any time * Memories are only used to improve your worker's assistance *** ## Common Questions Memories persist indefinitely until you remove them or the worker suggests an update due to changed circumstances. No. Workers selectively identify important, reusable information. Routine task details aren't automatically stored. You can correct any memory at any time. Simply point out the error, and the worker will update its knowledge base. No. Each worker maintains its own knowledge base. However, you can manually share relevant information with other workers. There's no practical limit. Workers efficiently organize and retrieve information regardless of quantity. *** ## Skills: Saving Learned Processes Beyond memory and conversational refinement, workers can also save complex processes as **Skills** — reusable workflows they can execute reliably. When your worker successfully handles a multi-step process (like processing invoices or generating reports), you can ask them to save it as a skill. They'll remember the exact steps and execute them consistently in the future. Understand how to create and use skills *** ## Next Steps Save complex processes as reusable workflows Learn about how workers remember information Best practices for training workers Learn how to manage and optimize your workers Understand security and permissions