Customizable AI Profiles for Every Meeting Type

A sales call is different from a technical interview. A team standup is different from a customer support session. The questions you need answered, the tone you want from your AI assistant, and the background knowledge that's relevant are completely different in each case.

That's why Raven has Modes.

A mode is a customizable AI profile that shapes how Raven's AI assistant behaves during a meeting. Each mode has its own system prompt, its own set of context documents, and its own notes template. You switch modes from the overlay before or during a recording, and the AI immediately adjusts its behavior.

What's in a mode

Every mode in Raven has the following:

Name and icon - a quick visual identifier so you can distinguish between modes in the overlay's mode selector. Each mode gets a name (like "Sales," "Interview," or "Standup") and an icon with a color.

System prompt - this is the core of a mode. It's the set of instructions that tells the AI how to behave. For example:

A sales mode might have: "You are assisting me during a sales call. Focus on the prospect's pain points and buying signals. When I ask for help, suggest responses that acknowledge their concerns and guide the conversation toward our solution. Flag when it's appropriate to discuss pricing. Be concise — I'm reading this during a live call."

An interview mode might have: "I'm conducting a technical interview. Help me evaluate the candidate's answers for correctness, depth, and communication clarity. Suggest follow-up questions that probe areas they skipped or oversimplified. Don't provide the answers yourself - help me assess."

A standup mode might have: "I'm in a team standup meeting. Track what each person says they're working on, any blockers they mention, and commitments they make. When I ask for a recap, organize by person."

Notes template - a structured set of sections that define how Raven generates session notes. Each section has a title and instructions that tell the AI what to extract. For a sales mode, you might have sections like "Pain Points Identified," "Objections Raised," "Next Steps," and "Pricing Discussion." For an interview mode: "Technical Assessment," "Communication Skills," "Red Flags," "Recommendation."

Context files - documents that get embedded locally and searched during conversations. Upload a PDF of your product specs, your company's pricing sheet, your interview rubric, a customer's previous support tickets — whatever reference material is relevant for that meeting type. During the call, when you ask the AI for help, it automatically searches these documents and includes the most relevant sections in its response.

How to create a mode

From the Raven dashboard, open the Modes section and click to create a new mode. Give it a name, pick an icon and color, write your system prompt, and optionally set up a notes template and upload context files.

That's it. The mode appears in the overlay's mode selector and is ready to use.

Quick actions work with modes

Raven's overlay has preset quick action buttons:

• Assist - general AI suggestion

• What should I say? - get a suggested response

• Follow-up - get a follow-up question

• Recap - summarize the conversation so far

• Custom - type any question

The system prompt from your active mode shapes how the AI handles each of these. The "follow-up" action in a sales mode will suggest a question about the prospect's budget or timeline. The same action in an interview mode will suggest a technical probe. The behavior is driven entirely by the system prompt - no code changes needed.

Built-in defaults

Raven comes with a default mode out of the box that provides general-purpose meeting assistance. You can use it as-is, modify it, or create entirely new modes from scratch.

Modes are local - they're stored in your SQLite database alongside your sessions and preferences. If you're on Raven Pro, they sync across devices.

Why this matters

The difference between a generic AI assistant and a useful one often comes down to context and instructions. A generic prompt like "help me in this meeting" produces generic suggestions. A specific prompt like "I'm interviewing a senior backend engineer - evaluate their system design answers for scalability thinking and production experience" produces dramatically more useful output.

Modes let you do that setup once and reuse it every time you have that type of meeting.