# Assistant memory and writing style Review remembered rules and personal facts, dismiss learned conclusions and manage the writing preferences used in your workspace. Canonical: https://docs.lead-knight.com/en/assistant/memory-and-writing-style/ ## Three different kinds of context Open **Settings** → **Assistant**. The page has separate controls for explicit memories, learned information and writing preferences. | Card | What it contains | Who it affects | | --- | --- | --- | | **What the assistant remembers** | Explicit remembered **Rule** and **Fact** items. | Rules apply across the workspace. Facts belong to you alone. | | **What the assistant has learned** | Conclusions learned from conversations and observed outcomes, with scope and supporting evidence. | A workspace-wide item affects workspace members; a personal item affects its owner. | | **Writing style** | **Prefer** and **Avoid** preferences for AI writing. | Only your writing in the current workspace. | Business facts belong in **Business identity** in [Settings](https://docs.lead-knight.com/en/settings/overview/). Archiving a conversation is another separate action: it hides the chat but does not erase the assistant's memory. ## Before you start Check the workspace and the scope of the item. Removing a shared rule or dismissing a workspace-wide conclusion can affect other people. Personal writing preferences do not change everyone else's writing. Use clear, factual instructions. For example, a preference for short sentences belongs in writing style; an instruction that should apply to everyone in the workspace belongs to shared context. Do not store passwords or API keys as assistant memories. ## Ask the assistant to remember something 1. Open **Assistant** and state what should be remembered. 2. Be explicit about scope: is it your own fact, or a standing rule for this workspace? 3. Read the response rather than assuming every sentence in a conversation was stored as a memory. 4. Open **Settings** → **What the assistant remembers** to check the resulting item and its **Rule** or **Fact** label. If the card says **Nothing remembered yet. Ask the assistant to remember something.**, there are no items to display in that list. This is not a statement that the chat archive is empty. ## Remove remembered information 1. In **What the assistant remembers**, read the item and its scope. 2. Select **Forget** beside one item to remove it from the assistant. 3. To remove the workspace's shared rules and your own facts together, select **Forget everything**. 4. Read the confirmation carefully. Confirm with **Yes, forget everything**, or select **Cancel** to keep the memories. Bulk forgetting cannot be undone. It does not mean that another member's private facts are being shown or deleted. > **Important:** Forgetting removes a memory from the assistant immediately. Activity records that may include that memory's text can be kept for up to 180 days for security auditing, then deleted. Forgetting is not a promise of instant removal from every audit record. ## Review learned conclusions 1. Find **What the assistant has learned**. 2. Read the conclusion, scope, subject, supporting-source count and when it was last supported. Labels such as **User stated**, **Observed** or **Contested** help you interpret it; they do not make an incorrect conclusion true. 3. If the conclusion should no longer be used, select **Dismiss**. 4. Read the confirmation, especially for a workspace-wide item. Select **Dismiss belief** to confirm. Dismissing suppresses that exact value. The assistant can still learn a corrected value later. If the card says that only recent beliefs are shown, older ones are not available there to review or dismiss. If your role can view but cannot dismiss an item, the app explains the restriction. Ask someone with the appropriate workspace role rather than repeatedly submitting the dismissal. ## Add or edit writing preferences 1. Find **Writing style**. The counter shows the number of saved preferences out of 30. 2. Select **Add preference**. An editor appears. 3. Choose **Prefer** for something you want more of, or **Avoid** for something to leave out. 4. Enter a concise preference, up to 200 characters. For example, “short sentences” or “avoid sales jargon” describes a style, not a business fact. 5. Select **Save**. Check that the saved preference appears in the list. 6. Use **Edit** on an existing row to change its text or type, then **Save**. Use **Cancel** to leave the editor without saving. The card allows up to 30 preferences. Preferences can also be learned when you close **Discuss** in the AI writing interface. Review the resulting list if a writing habit is no longer useful. ## Delete writing preferences - **Delete** removes the selected preference. - **Clear all** opens a confirmation for all of your writing preferences in this workspace. Confirm with **Clear all**, or choose **Cancel**. Clearing writing preferences is separate from **Forget everything** and from dismissing learned conclusions. Use the control for the kind of context you actually want to change. ## If something goes wrong - You removed a chat but the assistant still uses a preference: archive and memory are different. Find the relevant memory, learned conclusion or writing preference and remove it there. - **Add preference** is disabled: wait for loading or a save to finish, or remove an unneeded preference if the count is 30. - A style preference cannot be saved: enter non-empty text within 200 characters and check the displayed error. - A load or delete operation fails: use **Retry** where offered and wait for confirmation. Do not treat a failed request as successful forgetting. - Dismissal is forbidden: ask an appropriately authorized workspace member. Changing the wording in the chat does not bypass the permission. ## Related pages - [Assistant overview](https://docs.lead-knight.com/en/assistant/overview/). - [Past chats](https://docs.lead-knight.com/en/assistant/past-chats/). - [Settings and business information](https://docs.lead-knight.com/en/settings/overview/). ## Assistant settings: memory, writing style and approvals [Assistant settings: memory, writing style and approvals](https://kwqpaiaguxkmadaipfrf.supabase.co/storage/v1/object/public/guide-videos/guides/videos/M4/nadlanite-en-348904e0695c.mp4) 1. In Settings, the Assistant section shows what the assistant has learned and remembers about you. 2. To shape how it writes for you, click Add preference. 3. Describe your style in a few words, and save. 4. It now writes in your style, for you only. 5. Approvals decide whether it publishes on its own. Off means every publish waits for you. 6. Once WhatsApp is connected, it can also send reminders and a morning briefing.