Flag Twilio Conversations for Review Based on Custom Criteria

When your AI agent is handling conversations—or even when you're replying manually—some messages require a second look.

Maybe a customer is asking for a refund. Maybe they seem frustrated. Maybe they’re trying to escalate the issue and ask for a real human. Whatever the case, you don’t want those moments to slip through the cracks.

That’s why we built Poku’s flagging feature.

It lets you automatically flag inbound or outbound messages for human review based on your own rules—so nothing sensitive, risky, or important goes unnoticed.

🔶 What Is Poku’s Flagging Feature?

Poku's flagging system helps you identify and review high-risk or high-value moments in a conversation. When a message meets your criteria, Poku:

• Flags the conversation with an orange flag and surfaces it to the top

• Adds a “Review Required” label to the chat

• Keeps the flag active until someone manually hits “Resolve”

Whether you're using an AI agent or responding manually, this feature ensures you stay in control.

🔧 Choose Your Setup Path

(1) Are you using an SMS AI Agent with Poku?

• Yes → Continue to question (2)

• No → Skip ahead to 🚨 Using Poku Flagging Without an AI Agent

(2) Which of the following best describes you?

🧰 I'm using Make.com or n8n to build AI workflows

Follow our SMS AI Agent Setup Guide to build a scenario that connects your SMS AI agent to the Poku dashboard. Once integrated, your agent can trigger flagging automatically using rules defined in the dashboard (see section "Customize your Flagging Rules" below).

🧠 I'm experienced with using MCP Tool Calls

If you’re building your own agent logic and want to control flagging programmatically:

Use the flag_chat tool via our MCP server:

• Server URL (Streamable): https://mcp.pokulabs.com

• Server URL (SSE): https://mcp.pokulabs.com/sse

• Timeout: 60 seconds

• Headers:

  Authorization: Bearer YOUR_POKU_API_KEY

Note You must pass the to and from numbers as variables to your agent so your agent can tell Poku which unique chat thread to flag.
Once connected, simply prompt your agent with criteria for when to trigger the flag (e.g., “use flag_chat any time the user asks for a human”).

Sample prompt:

Use flag_chat tool if “customer asks for a discount”]. When using flag_chat tool and passing toPhoneNumber use {{ $('Twilio Trigger').item.json.data.to }}. When passing fromPhoneNumber use {{ $('Twilio Trigger').item.json.data.from }}

⚠️ No need to configure any rules in the Poku dashboard if you're using the MCP tool flag_chat—you control everything directly from your agent prompt.

🚨 Using Poku Flagging Without an AI Agent

Yes, you can use Poku’s flagging feature even if you’re replying to texts manually without an AI agent. Here’s how:

Step 1: Customize Your Flagging Rules

1. Go to the Human Intervention tab

2. Open the Flagging section

3. Enter your OpenAI API Key (used to evaluate message content)

4. Add your own criteria for what should trigger a flag

Examples:

• Flag inbound messages when “the customer seems angry or frustrated”

• Flag outbound messages when “we promise someone will follow up soon”

• Flag any message where “the user asks for a refund”

• Flag inbound messages when “a human is explicitly requested”

Pro Tip: You can write your rules in any language.

Step 2: Build a Scenario in n8n or Make.com

To activate flagging logic, you’ll need a basic automation that sends each incoming message to Poku for analysis.

💡 Pro Tips

• You can flag both inbound and outbound messages

• Flags are tied to message content, not sender identity

• Combine this with the chat toggle to let humans fully take over flagged conversations

• Set up real-time alerts for flagged messages using Slack, email, or internal tools

🤝 Need Help?

We’re happy to walk you through your setup or brainstorm flagging rules with your team.

📧 Email us at hello@pokulabs.com
📅 Book a setup call