This guide is for dashboard users. Everything below happens inside the Kakiyo dashboard. No code or API calls required.
The Inbox is where you monitor every conversation your AI agent is having with prospects. You can watch conversations in real time, take manual control when needed, and switch between full automation and human oversight.
Accessing the Inbox
- In the sidebar, click Inbox.
- The Inbox shows all conversations across all campaigns.
- Each conversation shows the prospect name, status, last message, and timestamp.
Operation Modes
Kakiyo offers two modes for managing conversations. You set the mode per campaign, but can override it per conversation.
Autopilot Mode
The AI handles everything — from first message to meeting booking — without human intervention.
| What happens | Details |
|---|
| AI sends first message | 20 minutes to several hours after connection acceptance |
| AI handles replies | Responds to prospect messages within 1-8 minutes |
| AI qualifies prospects | Asks questions based on your prompt’s qualification criteria |
| AI books meetings | Proposes next steps when the prospect shows interest |
| AI handles objections | Uses your offering knowledge to address pushback |
Best for: Standard outreach at scale, when your prompts and offerings are well-tested.
Copilot Mode
The AI handles conversations, but you can step in at any time to take manual control.
| What happens | Details |
|---|
| AI manages conversation normally | Same behavior as Autopilot |
| You can pause any conversation | Click Pause to stop AI responses |
| You reply manually | Write messages yourself for that specific conversation |
| You resume AI control | Click Resume to hand the conversation back to the AI |
Best for: High-value prospects, complex sales cycles, or when you’re still refining your prompts.
Managing Individual Conversations
Viewing a Conversation
- In the Inbox, click on any conversation to open it.
- You see the full message history: AI messages, prospect replies, and any manual messages.
- The conversation status is displayed at the top.
Conversation Statuses
| Status | Meaning |
|---|
| Active | AI is managing the conversation normally |
| Paused | AI is stopped — you have manual control |
| Completed | Conversation reached a terminal state (qualified, meeting booked) |
| Inactive | Prospect stopped responding after all follow-ups |
Taking Manual Control of a Conversation
- In the Inbox, click on the conversation you want to control.
- Click the Pause button in the conversation view.
- The AI immediately stops responding to this conversation.
- You can now type and send messages manually.
When to take manual control:
- The prospect asks a highly technical or specific question the AI can’t handle well
- A high-value prospect needs personalized attention
- The conversation requires strategic decision-making (pricing negotiations, custom deals)
- The AI’s response seems off-track and you want to correct course
Resuming AI Control
- Open the paused conversation in the Inbox.
- Click the Resume button.
- The AI picks up the conversation from where it left off, with full context of the conversation history (including your manual messages).
AI Response Behavior
Understanding how the AI times and structures its responses:
| Behavior | Details |
|---|
| Response time | 1-8 minutes after a prospect replies (randomized to appear human) |
| First message timing | 20 minutes to several hours after connection acceptance |
| Context memory | The AI remembers the full conversation history |
| Follow-ups | Sent automatically if the prospect doesn’t reply (see Follow-up Messages) |
| Qualification | The AI assesses prospect fit based on the Context Prompt’s criteria |
| Meeting booking | The AI proposes a call/meeting when genuine interest is detected |
Monitoring Best Practices
| Practice | Frequency | What to look for |
|---|
| Review new conversations | Daily | Are first messages personalized and relevant? |
| Check qualified leads | Daily | Are prospects being correctly qualified? |
| Read flagged conversations | As needed | Conversations where the AI might need help |
| Review completed conversations | Weekly | What worked well? What could improve? |
| Compare across campaigns | Weekly | Which campaigns have the best conversation quality? |
When to Intervene vs. When to Let the AI Work
Intervene when:
- Prospect asks about something not in your offering
- Conversation involves pricing negotiation beyond your rules
- Prospect is a key account or known contact
- AI response seems incorrect or tone-deaf
Let the AI work when:
- Standard qualification conversations
- Common objection handling (competitor mentions, timing, budget)
- Routine follow-up sequences
- Conversations where the AI is performing well
Optimizing Conversation Quality
If you notice recurring issues in conversations:
| Issue | Fix |
|---|
| AI gives wrong product information | Update your Offering (sidebar > Offerings) |
| AI tone doesn’t match your brand | Edit the Context Prompt (sidebar > Prompts) |
| AI qualifies too aggressively or too loosely | Adjust qualification criteria in the Context Prompt |
| First messages are too generic | Edit the First Message Prompt and add more personalization variables |
| AI can’t handle a common objection | Add objection handling notes to your Offering |
| Follow-ups are too frequent or too slow | Adjust delays in the campaign’s Follow-ups tab |
After making changes, always test in the Sandbox before applying to live campaigns.
Last modified on March 11, 2026