AI Intent Flows

Flows let you define structured conversational processes that run when a user message matches a specific Intent with enough confidence. They are used to collect data, apply tags, or escalate

What is a Flow?

A Flow is a guided interaction executed by the AI when certain conditions are met. Unlike free conversation, a Flow:

• Follows a predefined purpose

• Asks specific questions

• Collects structured fields

• Executes actions (email, tagging, escalation)

• Ends explicitly (completion or exit)

Common use cases:

• Demo / trial requests

• Lead qualification

• Contact data capture

• Human handoff

• Support escalation

---

Enable AI Intent Flows

1. Open AI Intent Flows

2. Toggle Enable AI Intent Flows

3. Click + Add intent

Once enabled, the agent can automatically route conversations into flows.

---

Intent configuration

Each Flow is linked to an Intent. The intent defines when the flow should trigger.

Intent name

A descriptive internal label.

Example:

• Test promo

• Pricing request

• Escalate

Intent confidence (0–1)

Controls how strict the trigger is.

• Higher values (e.g. 0.8) → triggers less often (more precise)

• Lower values (e.g. 0.5) → triggers more often (more recall)

Recommended starting points:

• Pricing / sales: 0.7 – 0.85

• Escalation: 0.6 – 0.75

Intent type

• Flow Collects data and can send email; applies tags on completion.

• Escalate Tags the conversation and optionally sends an escalation message.

---

Flow prompts

Description

The Description tells the AI when to trigger this intent and when not to.

Best practice:

• Be explicit and restrictive

• Include “ONLY trigger” conditions

• Include “do not trigger” conditions

Example:

ONLY trigger this intent if the user asks about pricing, costs, or wants a demo/trial. Otherwise do not use it.

Flow prompt

Defines what the AI does during the flow (questions, logic, tone). A good Flow prompt:

• Thanks the user and explains the purpose

• Handles privacy/consent before collecting data

• Asks for required fields in a clear order

• Stays on-topic until completion

Example structure:

1. Confirm consent (if collecting personal data)

2. Ask for the required fields (name, email, company, use case)

3. Confirm next steps

---

Exit handling

Exit prompt (flow stop)

Defines when the flow should stop early.

Use this when:

• The user refuses to provide data

• The user goes off-topic repeatedly

• The user already provided the required info

Exit message prompt

Message sent when the flow stops.

Best practice:

• Apologize briefly

• Confirm you can still help

• Offer alternative support

---

Completion message

The Completion message prompt is used after the flow ends successfully.

Best practice:

• Thank the user

• Confirm receipt of details

• Explain what happens next (e.g., “we will contact you soon”)

• Offer additional help

---

Tags

Tags are applied when the flow completes.

Use tags for:

• Conversation routing

• Analytics

• CRM sync

• Filtering conversations

Example:

• test_request

• agent

• pricing_lead

---

Intent examples

Intent examples help the classifier match user messages to this intent.

Use:

• Real user phrases

• Short and varied wording

• One per line

Examples (Pricing / Trial):

• How much does it cost?

• Pricing?

• I want a demo

• Can I try it?

• I’d like to test the service

Examples (Escalation):

• I want to speak with a person

• Can I talk to a human?

• Please connect me to support

---

Email actions (optional)

Flows can send an email when completed.

Email settings

• Email to: destination inbox

• Email from: sender address

• Email subject: supports variables like {name}, {email}, {company}

• Email fields: list of collected fields to include

Example:

• Subject: Test Request: {name}

• Fields: name, email, company, business_case

---

Escalation intent

Use an Escalate intent when the user requests a human agent.

When triggered:

• The conversation is tagged (e.g. agent)

• The AI can send a short escalation message

• A human can take over

Recommended escalation message:

Thanks — I’m forwarding your request to our support team now.