AI Chat conversations
AI Chat is Ultrathink's conversational interface for interacting with AI about your knowledge base. Unlike AI Canvas, which is built around a single prompt, AI Chat lets you have a back-and-forth dialogue where each message builds on the previous ones.
What AI Chat is
AI Chat gives you a chat-style interface where you can ask questions, get answers, and then follow up naturally. The AI has access to your knowledge base, so it can draw on your saved entries, summaries, and metadata to provide relevant, personalised responses.
It works like having a conversation with someone who has read everything you have saved.
How AI Chat differs from AI Canvas
Both tools use AI to work with your knowledge, but they serve different purposes:
| Aspect | AI Chat | AI Canvas |
|---|---|---|
| Interaction | Multi-turn conversation | Single prompt with attachments |
| Context source | Draws from your entire knowledge base automatically | You manually attach specific marked bits |
| Best for | Exploration, follow-up questions, iterative thinking | Focused tasks with defined inputs and outputs |
| Memory | Remembers the full conversation thread | Each composition is independent |
| Workflow | Open-ended, conversational | Task-oriented, structured |
Choose AI Chat when you:
- Want to explore a topic across your knowledge base
- Need to ask follow-up questions based on what you learn
- Are not sure exactly what you are looking for
- Want to have a dialogue that evolves naturally
Choose AI Canvas when you:
- Have specific entries to use as context
- Know exactly what output you want
- Need fine-grained control over reasoning level and web search
- Want to produce a single, polished piece of content
Starting a conversation
- Navigate to AI Chat in the navigation
- Type your question or prompt in the message input
- Press Enter or click Send
- The AI responds with an answer drawn from your knowledge base
The first message sets the context for the conversation. You can start with a broad question and then narrow down, or begin with something specific.
Knowledge base context
AI Chat automatically searches your knowledge base to find relevant entries when answering your questions. You do not need to manually attach or reference specific entries.
How context works
When you send a message, the AI:
- Analyses your question to understand what you are asking
- Searches your entries for relevant content
- Uses matching entries as context for its response
- Cites which entries it drew from
What the AI can access
| Content | Available to AI Chat |
|---|---|
| Entry titles and summaries | Yes |
| Your notes and annotations | Yes |
| Extracted article content | Yes |
| AI-generated classifications | Yes |
| Conversation transcripts | Yes |
| Audio transcriptions | Yes |
| Screenshot descriptions | Yes |
| PDF content | Yes |
The AI works with text-based representations of your content. For screenshots, it uses the AI-generated descriptions and any extracted text (OCR) rather than the images themselves.
Follow-up questions
One of AI Chat's strengths is the ability to ask follow-up questions. The AI remembers the entire conversation, so each new message builds on what came before.
Effective follow-up patterns
- Drill down: "Tell me more about the third point you mentioned."
- Redirect: "What about the same topic but from a cost perspective?"
- Clarify: "When you mentioned the API integration, which entries were you referencing?"
- Expand: "Can you find any other entries I have saved that relate to this?"
- Summarise: "Summarise everything we have discussed so far in bullet points."
Example conversation flow
- You: "What have I saved about React performance optimisation?"
- AI: Responds with a summary of relevant entries, citing specific articles and notes
- You: "Which of those techniques would apply to our dashboard project?"
- AI: Cross-references the performance entries with your dashboard project entries
- You: "Draft a list of action items based on this analysis."
- AI: Produces an actionable list grounded in your knowledge base
Managing conversations
Viewing conversation history
Your AI Chat conversations are saved and accessible from the AI Chat page. You can revisit previous conversations to review what was discussed.
Starting a new conversation
Click New conversation (or the equivalent button) to begin a fresh chat. Previous conversations remain accessible in your history.
Starting a new conversation clears the current context, so the AI will not carry over information from the previous thread unless you reference it explicitly.
Tips for better conversations
Be specific with your questions
Instead of "What do I know about marketing?", try "What articles have I saved about email marketing conversion rates in the last three months?"
Reference your knowledge
Mention specific projects, topics, or people to help the AI find the most relevant entries. For example: "What notes do I have from meetings with Sarah about the redesign project?"
Use the conversation to build understanding
Start broad and narrow down. AI Chat is at its best when you use the back-and-forth nature to progressively refine your understanding.
Ask for different formats
Request tables, bullet lists, numbered steps, or narrative summaries depending on what you need. The AI can restructure information from your entries into whatever format is most useful.
Common use cases
| Use case | Example starting question |
|---|---|
| Knowledge audit | "What have I captured about machine learning this quarter?" |
| Project briefing | "Summarise everything related to the website migration project." |
| People context | "What interactions and notes do I have involving Alex?" |
| Research exploration | "What do my saved articles say about the future of remote work?" |
| Decision support | "Based on my notes, what are the pros and cons of switching to TypeScript?" |
| Content discovery | "Are there any entries I have saved that connect sustainable design with our product roadmap?" |
Troubleshooting
| Problem | Solution |
|---|---|
| AI does not reference your entries | Try rephrasing your question with more specific keywords or topic names |
| Responses feel generic | Ask the AI to cite specific entries from your knowledge base |
| Conversation loses context | Start a new conversation if the thread has become too long or unfocused |
| Relevant entries are missed | Check that those entries have been fully processed (view AI Progress in the detail panel) |