AI Chat — Ask the Chief

Ask the Chief is a conversational AI interface that lets you ask natural-language questions about your meetings, people, and companies. It uses your entire meeting history as context to deliver precise, sourced answers.

Overview

Unlike traditional search that returns a list of meetings, Ask the Chief understands your question and synthesizes an answer from relevant meeting transcripts, summaries, and notes. Every answer includes citations linking back to the exact meeting and timestamp where the information was found.

Ask the Chief is available in four contexts, each scoped to a different level of your meeting data:

In-Meeting Chat: Ask questions about the current meeting in real time
Cross-Meeting Chat: Search and synthesize across your entire meeting history
People Chat: Ask about a specific person across all meetings involving them
Company Chat: Ask about a specific company across all related meetings

How it works under the hood

Ask the Chief uses retrieval-augmented generation (RAG). Your question is converted into an embedding vector, matched against your meeting transcript embeddings stored in pgvector, and the most relevant segments are passed to the AI model along with your question to generate the answer.

In-Meeting Chat

During an active capture, the chat panel is available in the meeting sidebar. Ask questions about what has been discussed so far in the current meeting.

What You Can Ask

"What did Sarah just say about the timeline?"
"Summarize the discussion so far"
"What objections has the client raised?"
"Has anyone mentioned the budget yet?"
"What action items have been assigned so far?"

In-meeting chat operates on the live transcript buffer. As new speech is captured and transcribed, the chat context grows. This means you can ask follow-up questions as the meeting progresses and get increasingly complete answers.

Real-time context

In-meeting chat updates its context every 10 seconds as new transcript segments arrive. There is a brief delay between someone speaking and that content being available for chat queries.

Cross-Meeting Chat

Cross-meeting chat is the primary Ask the Chief interface. Open it from the sidebar or by pressing Cmd+J. It searches across your entire meeting history to find and synthesize answers.

Example Questions

"What decisions have we made about the pricing model?"
"When did we last discuss the migration plan?"
"What has Mike committed to delivering this quarter?"
"Show me all the times we discussed competitor X"
"What were the key outcomes from last week's board meeting?"
"Has anyone raised concerns about the API redesign?"

Citations

Every answer includes inline citations that link back to the source. Citations show the meeting title, date, and speaker. Click a citation to jump directly to that moment in the transcript.

// Example answer with citations
Q: "What decisions have we made about the pricing model?"

A: Based on your recent meetings, three key pricing decisions have been made:

1. The team agreed to move from per-seat to usage-based pricing for
   the API tier [Product Sync, Feb 10 — Sarah Chen, 14:32]

2. Enterprise pricing will include a base platform fee plus per-user
   charges [Pricing Review, Feb 7 — Mike Torres, 22:15]

3. The free tier will be limited to 5 meetings per month instead of
   10 [Growth Meeting, Feb 3 — Lisa Park, 08:44]

Conversation memory

Cross-meeting chat maintains conversation context within a session. You can ask follow-up questions like "Tell me more about point 2" or "When was that decision revisited?" and the AI will understand the reference.

People Chat

People Chat is scoped to a single person. Open any participant profile and click the "Ask about this person" button to start a chat that searches only meetings involving that individual.

Example Questions

"What has Sarah committed to in the last 30 days?"
"What topics does Sarah bring up most frequently?"
"When was the last time Sarah mentioned hiring?"
"What concerns has Sarah raised about the project?"
"Summarize my last 5 meetings with Sarah"

People Chat is especially useful for managers preparing for one-on-ones, account executives reviewing relationship history before a client call, or anyone who needs a quick recap of their interactions with a specific person.

Prep for your next 1:1

Before a one-on-one meeting, open People Chat for that person and ask "What has this person been working on and what blockers have they mentioned?" for an instant context brief.

Company Chat

Company Chat is scoped to a specific company. Open any company profile and click "Ask about this company" to search across all meetings involving people from that organization.

Example Questions

"What is the current status of our deal with Acme Corp?"
"What pain points has Acme Corp mentioned?"
"Who are the key stakeholders at Acme Corp?"
"What products are they currently using?"
"Summarize our entire relationship history with Acme Corp"
"What has changed since our last meeting with them?"

Company Chat combines data from all meetings with anyone at that company, regardless of which team member had the meeting. This gives you a unified view of your organization's relationship with any external company. See the Companies documentation for more about how Karnyx tracks company relationships.

Tips for Getting Good Answers

Be specific about time ranges. Instead of "What did we discuss about pricing?" try "What pricing decisions were made in the last 2 weeks?" Narrower scopes produce more relevant answers.
Name people when possible. "What did Sarah say about the launch date?" is more precise than "What was said about the launch date?" and helps the AI retrieve the right transcript segments.
Ask follow-up questions. The chat maintains context within a session. After an initial answer, ask "Can you elaborate on the second point?" or "Were there any dissenting opinions?" to drill deeper.
Use the right chat context. If you are researching a person, use People Chat. If you are prepping for a client call, use Company Chat. The scoped contexts reduce noise and improve relevance.
Check citations. Click through to the source meetings to verify important details. AI answers are synthesized and may occasionally miss nuance that is visible in the original transcript.
Combine with Lenses. If a chat answer reveals something interesting, run a targeted Lens on the cited meetings for a more structured deep-dive.

Lenses Companies