How to use AI chat with transcripts

After Mediata generates your transcript, the AI chat becomes your main tool for extracting value from it. Here's how to use it effectively.

How AI chat works

The AI chat in Mediata is grounded in your transcript. When you ask a question, the AI:

1. Searches the transcript for relevant segments
2. Constructs an answer based on what was actually said
3. References specific parts of the conversation

This means the AI won't make things up or add information that isn't in the recording. If something wasn't discussed, it will tell you.

Basic prompts

Start with simple, direct questions:

• "Summarize this recording"
• "What were the main topics discussed?"
• "Who participated in this conversation?"
• "How long did the discussion about [topic] last?"

These give you a quick overview before diving deeper.

Extracting specific information

Decisions and action items

• "List all decisions made in this meeting"
• "What action items were assigned? Include who is responsible"
• "What deadlines were mentioned?"
• "Were there any unresolved disagreements?"

Quotes and statements

• "What did [speaker name] say about [topic]?"
• "Find the exact quote where someone mentioned [keyword]"
• "List all statements about the budget"
• "Who first brought up [subject]?"

Numbers and data

• "List all numbers, dates, and percentages mentioned"
• "What financial figures were discussed?"
• "What timeline was proposed for the project?"

Advanced techniques

Structured output

Ask the AI to format its response in a specific way:

• "Create meeting minutes with sections: Attendees, Topics, Decisions, Action Items, Next Steps"
• "Make a table with columns: Task, Owner, Deadline"
• "Write a brief executive summary in 3 sentences"

Comparative analysis

When you have context from the conversation:

• "Compare what Speaker A and Speaker B said about the timeline"
• "Were there any contradictions in what was discussed?"
• "What changed between the early and late parts of the meeting?"

Topic extraction

• "What are the 5 main themes in this conversation?"
• "Create a topic outline of this recording"
• "Which topics got the most discussion time?"

Follow-up questions

The AI remembers your conversation. Ask follow-up questions to go deeper:

1. "Summarize the meeting" → read the summary
2. "Tell me more about the third point" → get details
3. "What exactly did the team agree to do about that?" → get specifics

Prompts by use case

For meetings

• "Create a meeting recap I can send to the team"
• "List every commitment made, with the person who made it"
• "What questions were raised but not answered?"

For interviews

• "Pull the most quotable statements from this interview"
• "Summarize the interviewee's main arguments"
• "What topics did the interviewee seem hesitant about?"

For lectures

• "Create study notes from this lecture"
• "Explain [concept] as discussed in this lecture"
• "What examples did the professor use?"

For legal recordings

• "Find every mention of [clause/contract/term]"
• "List all dates and events referenced"
• "Summarize the witness's key statements"

Tips for better answers

Be specific. "What did we decide about pricing?" works better than "Tell me about the meeting."

Name the speaker. If you want information from a specific person, include their name: "What did Anna say about the deadline?"

Ask one thing at a time. Complex multi-part questions sometimes get incomplete answers. Split them up.

Rephrase if needed. If the first answer isn't what you want, try asking differently. "List the action items" vs. "What does each person need to do next?" might give different perspectives.

Use the transcript as a reference. If the AI's answer seems incomplete, check the transcript panel to see if relevant content was missed.