What This Template Is For
Most support chatbots fail because they are built around technology capabilities rather than customer needs. The bot can answer 200 questions, but customers cannot find the right one. The conversation feels robotic. Fallbacks dump users into a queue with no context. The result is a chatbot that frustrates customers and creates more work for agents, not less.
This template provides a structured approach to designing a support chatbot that handles common requests well and hands off gracefully when it cannot. It covers persona definition, conversation flow mapping, fallback strategy, and success metrics. If you are evaluating whether a chatbot fits into your broader support strategy, review the self-service strategy template first. For understanding how AI tools fit into your product workflow, the AI PM Handbook covers the fundamentals of shipping AI features responsibly.
How to Use This Template
- Copy the template into your design tool or documentation system.
- Start with the chatbot persona. Decide the tone and boundaries before writing any conversation flows.
- Map your top 10 support requests to conversation flows. These should cover 60-80% of inbound volume.
- Design the fallback strategy next. Every conversation will eventually hit a point where the bot cannot help.
- Define success metrics before launch. If you do not know what "good" looks like, you cannot iterate.
- Test flows with real customers (or at minimum, with support agents roleplaying as customers) before going live.
The Template
Chatbot Persona
- ☐ Define the bot's name
- ☐ Define the tone (formal, casual, friendly, neutral)
- ☐ Set boundaries on what the bot will and will not do
- ☐ Specify when the bot should identify itself as non-human
| Attribute | Definition |
|---|---|
| Name | [Bot name] |
| Tone | [e.g., Friendly and concise. No slang, no emojis.] |
| Scope | [What the bot handles: billing questions, how-to guidance, status checks] |
| Out of scope | [What the bot will never handle: refunds over $X, account deletion, legal requests] |
| Disclosure | [When does the bot say "I'm a bot"? On first message? Only if asked?] |
Conversation Flows
Map each high-volume support topic to a flow. Each flow needs an entry point, a resolution path, and an exit.
- ☐ Identify the top 10 support topics by ticket volume
- ☐ Design a conversation flow for each topic
- ☐ Include at least 2 clarifying questions per flow to narrow intent
- ☐ Map every dead end to a fallback action
Flow Template:
| Step | Bot Says | User Options | Next Step |
|---|---|---|---|
| 1 - Greeting | [Welcome message + intent options] | [Button 1] / [Button 2] / [Free text] | Route to flow |
| 2 - Clarify | [Clarifying question] | [Option A] / [Option B] | Step 3 or fallback |
| 3 - Resolve | [Answer or action] | [Helpful?] Yes / No | End or escalate |
| 4 - Close | [Closing message] | [Rate experience] | End |
Conversation Flow Inventory
| Flow ID | Topic | Entry Trigger | Resolution Type | Fallback |
|---|---|---|---|---|
| F-01 | [Topic] | [Keyword or button] | [Self-serve / Link / API action] | [Escalation type] |
| F-02 | [Topic] | [Keyword or button] | [Self-serve / Link / API action] | [Escalation type] |
| F-03 | [Topic] | [Keyword or button] | [Self-serve / Link / API action] | [Escalation type] |
Fallback Strategy
Every bot conversation eventually reaches a point where the bot cannot help. Design for this.
- ☐ Define a "confused" response for when the bot does not understand the input
- ☐ Set a maximum number of failed attempts before auto-escalation (recommended: 2)
- ☐ Pass full conversation context to the agent on handoff
- ☐ Let the user request a human at any point in the conversation
- ☐ Track fallback reasons to identify gaps in conversation flows
| Fallback Trigger | Bot Response | Action |
|---|---|---|
| Intent not recognized (1st time) | "I didn't quite get that. Could you rephrase or pick from these options?" | Show top-level menu |
| Intent not recognized (2nd time) | "Let me connect you with a team member who can help." | Create ticket with context |
| User requests human | "Connecting you now. A team member will respond within [X]." | Route to agent queue |
| Out-of-scope request | "I'm not able to help with [topic], but our team can. Let me connect you." | Create ticket, tag as out-of-scope |
Success Metrics
- ☐ Define containment rate target (conversations resolved without human handoff)
- ☐ Define customer satisfaction target for bot interactions
- ☐ Define fallback rate threshold that triggers flow redesign
- ☐ Track time-to-resolution for bot-handled versus agent-handled conversations
| Metric | Baseline | Target | Measurement |
|---|---|---|---|
| Containment rate | [Current] | [Target %] | Bot platform analytics |
| Bot CSAT | N/A | [Target] | Post-conversation survey |
| Fallback rate | N/A | [Below X%] | Bot platform analytics |
| Avg. bot resolution time | N/A | [Target] | Bot platform analytics |
Filled Example: SaaS Support Bot
Persona
| Attribute | Definition |
|---|---|
| Name | "Support Assistant" (no anthropomorphic name) |
| Tone | Friendly, concise, helpful. Uses complete sentences. No emojis or exclamation marks. |
| Scope | Password resets, billing questions, feature how-tos, status page checks, plan comparisons |
| Out of scope | Refunds over $100, account deletion, security incidents, contract negotiations |
| Disclosure | First message: "Hi, I'm the Support Assistant, an automated helper." |
Sample Flow: Password Reset (F-01)
| Step | Bot Says | User Options | Next Step |
|---|---|---|---|
| 1 | "What can I help with today?" | [Reset password] / [Billing] / [Something else] | Route to F-01 |
| 2 | "I can send a reset link to the email on your account. What email address did you sign up with?" | [User types email] | Step 3 |
| 3 | "I've sent a reset link to j*@company.com. Check your inbox and spam folder. It expires in 30 minutes." | [That worked] / [I didn't get it] | End or Step 4 |
| 4 | "Let me connect you with a team member to verify your account and reset manually." | - | Create ticket |
Conversation Flow Inventory
| Flow ID | Topic | Entry Trigger | Resolution Type | Fallback |
|---|---|---|---|---|
| F-01 | Password reset | "password", "can't log in", "reset" | Send reset email via API | Agent handoff |
| F-02 | Billing inquiry | "invoice", "charge", "billing" | Link to billing portal | Agent handoff |
| F-03 | Feature how-to | "how do I", "where is", "set up" | Link to knowledge base article | Agent handoff |
| F-04 | Status check | "down", "outage", "not working" | Pull from status page API | Agent handoff if incident active |
The chatbot works best alongside a well-structured knowledge base that the bot can link to for detailed answers. Use the AI ROI Calculator to estimate the cost savings from chatbot deflection before committing to a build. Track how chatbot performance affects your support SLA targets over time.
Key Takeaways
- Design the persona and boundaries before writing conversation flows
- Start with 5-8 flows covering your highest-volume ticket types
- Every flow needs a clear fallback path that hands context to a human agent
- Measure containment rate, CSAT, and fallback rate from day one
- Review flows monthly at first, then quarterly once performance stabilizes
About This Template
Created by: Tim Adair
Last Updated: 3/4/2026
Version: 1.0.0
License: Free for personal and commercial use