AI Chatbot Conversation Flow Templates: 10 Proven Flows to Copy
A great chatbot conversation has a shape: an opener, a branch, a capture and a handoff. These 10 battle-tested AI chatbot conversation flow templates give you the structure for the most common jobs — adapt the copy to your brand.
Key takeaway
You don't need to invent conversation design from scratch. Almost every useful chatbot job — qualifying a lead, booking a meeting, deflecting a ticket, recovering a cart — follows one of a handful of repeatable shapes. Copy the structure of these AI chatbot conversation flow templates, swap in your brand voice and data fields, and ship. The structure is what makes a flow convert; the copy is what makes it yours.
The anatomy of a good conversation flow
Before the templates, internalize the shape they all share. Every effective flow moves the user through the same six beats. When a flow underperforms, it is almost always because one of these beats is missing, bloated, or in the wrong order. If you want the deeper theory behind each beat, our conversation design guide walks through the principles; this post is the applied, copy-and-paste version.
Greeting
Open with brand-voiced warmth and a clear signal of what the bot can do. One or two lines, never a wall of text.
Intent branch
Offer 2-4 quick-reply options or read free text to route the user. This is the fork that sends each visitor down the right path.
Qualify or answer
Ask the minimum questions needed, or surface the answer. Each question must earn its place — drop anything you will not act on.
Capture or action
Collect the data or trigger the action: book the slot, add to CRM, recover the cart, create the ticket.
Confirm and handoff
Confirm what happened, set expectations ("a human will reply within an hour"), and pass context to a person if needed.
Fallback
Every node needs an escape hatch. If the bot cannot understand after two tries, offer a human or a callback — never loop.
The single biggest lever is the intent branch. A flow that opens with "How can I help?" and a blank text box puts all the work on the user. A flow that opens with three quick-reply buttons routes 80%+ of visitors instantly and sets up everything that follows. Notice how every template below front-loads that branch.
10 AI chatbot conversation flow templates
Each template below is a complete, turn-by-turn flow: the goal it serves, the trigger that fires it, the step-by-step script with branch logic, the data it captures, and the success metric you should watch. Bot lines are written to copy; swap the brand name, plans, and prices for yours. For more finished dialogue you can lift directly, pair these with our library of real chatbot conversation examples.
Lead qualification
Trigger: Fires on pricing, contact, and demo pages after 5-8s.
The flow
- Bot: "Hi! Looking to see if EzyConn fits your team? I can qualify you in 30 seconds — what brings you in today?" [Buttons: Pricing · Demo · Just browsing]
- User picks Pricing → Bot: "Great. How big is the team that'll use it?" [1-10 · 11-50 · 50+]
- Bot: "Got it. What's the main job you want it to do — support, sales, or both?"
- Bot: "Perfect. Drop your work email and I'll send a tailored plan and a 14-day trial link." [Capture: email]
- Bot: "Done! Your trial is ready. Want me to book a 15-min setup call too?" → routes hot leads to booking flow (#2).
Data captured
Name, work email, company size, use case, budget tier
Success metric
Qualified leads per 100 conversations
Demo / meeting booking
Trigger: Triggered by a "Book a demo" CTA or qualified hot lead.
The flow
- Bot: "Happy to set up a live demo. What's your name and work email?" [Capture: name, email]
- Bot: "Thanks, {name}! What's the best timezone for you?" [Detect from browser, confirm]
- Bot: "Here are the next open slots — pick one that works." [Embedded calendar with 4-6 slots]
- User selects → Bot: "Booked for Thursday 2pm. I've sent a calendar invite and a join link. Anything specific you want us to cover?"
- Bot: "Noted — I'll pass that to your specialist. See you Thursday!" → writes notes to CRM, notifies the rep.
Data captured
Name, email, company, timezone, chosen slot
Success metric
Booked meetings ÷ conversations started
Support triage & deflection
Trigger: Default flow on help, docs, and account pages.
The flow
- Bot: "Hi, I'm the EzyConn assistant. What can I help with?" [Billing · Setup · A bug · Something else]
- User picks Setup → Bot reads the knowledge base and answers inline with a source link.
- Bot: "Did that solve it?" [Yes, thanks · Not quite]
- If Yes → Bot: "Glad I could help! Anything else?" (logged as deflected).
- If Not quite → Bot: "No problem — let me get a human on this. Quick: what's your account email?" → creates ticket with transcript, routes to live agent.
Data captured
Issue category, order/account ID, sentiment
Success metric
Deflection rate (resolved without an agent)
Cart abandonment recovery
Trigger: Fires when a user is idle 60s on checkout or moves to exit.
The flow
- Bot: "Still thinking it over? I can help you check out faster." [Apply a code · Shipping question · Just looking]
- User picks Shipping → Bot: "Orders over $50 ship free and arrive in 2-3 days. Want me to take you back to checkout?"
- User picks Apply a code → Bot: "Here's 10% off your first order: WELCOME10. It's applied — your total dropped to $54."
- Bot: "Want me to email this cart so you can finish later?" [Capture: email]
- Bot: "Sent! Your code holds for 24 hours." → triggers an abandoned-cart email sequence.
Data captured
Cart contents, blocker reason, email for follow-up
Success metric
Recovered revenue per triggered session
Product / plan recommender
Trigger: On catalog, pricing, or "help me choose" CTAs.
The flow
- Bot: "I can match you to the right plan in 3 questions. Ready?" [Yes · Show all plans]
- Bot: "Who's it for — a solo founder, a small team, or a department?"
- Bot: "What's the one feature you can't live without?" [Live chat · Automations · Analytics · Integrations]
- Bot: "Based on that, the Growth plan fits best — automations plus analytics at $49/mo." [See Growth plan · Compare all]
- Bot: "Want a free trial of Growth, or should I answer a question first?" → routes to signup or support.
Data captured
Use case, team size, must-have features, budget
Success metric
Recommendation click-through to product page
Onboarding / getting started
Trigger: First login after signup, or on the empty dashboard.
The flow
- Bot: "Welcome aboard, {name}! Let's get you live. What do you want to do first?" [Add the widget · Connect WhatsApp · Import FAQs]
- User picks Add the widget → Bot: "Easy. Copy this snippet into your site's <head>, or send it to a dev." [Copy snippet · Email my dev]
- Bot: "Once it's in, refresh your site and I'll confirm it's live." [Detects install → checkmark]
- Bot: "You're live! Next, want to import your FAQs so I can answer customers?"
- Bot: "Nice work — you've completed setup. Want to invite a teammate?" → logs activation milestone.
Data captured
Goal for the product, connected channel, invited teammate
Success metric
Activation rate (first key action completed)
Appointment scheduling (local services)
Trigger: On a clinic, salon, or trades homepage.
The flow
- Bot: "Hi! I can book your appointment. What service do you need?" [Haircut · Color · Consultation]
- Bot: "Great choice. Which day works best this week?" [Shows real availability from the calendar]
- Bot: "We have 10am, 1pm, and 4pm on Friday. Which one?"
- Bot: "Locked in for Friday 1pm. What's the best number to text a reminder to?" [Capture: phone]
- Bot: "You're booked! We'll text a confirmation now and a reminder the day before." → writes to booking system.
Data captured
Service type, preferred date, contact number, location
Success metric
Confirmed appointments per conversation
FAQ deflection with escalation
Trigger: Any page where common questions repeat.
The flow
- Bot: "Ask me anything — returns, hours, shipping, sizing."
- User: "What's your return policy?" → Bot answers from the knowledge base with the exact policy and a link.
- Bot: "Was that helpful?" [Yes · No]
- If No twice in a row → Bot: "Let me connect you with someone who can dig in." → opens handoff with transcript.
- If after-hours → Bot: "Our team is offline, but I'll have someone reply by 9am. What's your email?" → routes to after-hours flow (#10).
Data captured
Question text, answer rating, escalation reason
Success metric
Self-serve resolution rate
Feedback / CSAT collection
Trigger: After a resolved chat or completed order.
The flow
- Bot: "Quick one — how was your experience today?" [1 to 5 star buttons]
- User picks 5 → Bot: "Love to hear it! Mind leaving a quick review?" [Sure · Maybe later]
- User picks 2 → Bot: "Sorry we missed the mark. What went wrong?" [Free text]
- Bot: "Thank you — I've flagged this to the team. Can we email you a fix?" [Capture: email]
- Bot: "Appreciate the honesty. We'll be in touch." → low scores alert a manager in real time.
Data captured
Rating (1-5), free-text comment, follow-up consent
Success metric
Response rate and average CSAT score
After-hours capture with callback
Trigger: Activates outside business hours on every page.
The flow
- Bot: "Hi! Our team is offline until 9am, but I can still help or set up a callback." [Ask a question · Request a callback]
- User picks Request a callback → Bot: "Great. What's your name and the best number to reach you?" [Capture: name, phone]
- Bot: "What's this about, so the right person calls you?" [Sales · Support · Billing]
- Bot: "When works best — morning or afternoon?"
- Bot: "Perfect. We'll call you tomorrow morning. You'll get a text confirmation now." → queues the callback and notifies the on-call team.
Data captured
Name, phone/email, reason, preferred callback window
Success metric
Callbacks booked ÷ after-hours visitors
Build any of these in a visual editor
EzyConn turns each step above into a node on a canvas — message, question, branch, capture, handoff. No code, no dev tickets. Build your flows with no code and go live in an afternoon.
Start building freePersonalization, fallback, and human handoff inside flows
A template is the skeleton. Three layers turn it into something that feels alive and never traps the user.
Personalization
Inject context you already have: the visitor's name, plan, last conversation, the page they're on, or items in their cart. A returning user should never be greeted like a stranger. Done well, personalization lifts flow completion 20-35%. Done badly — "Hi friend" with no real data — it reads as fake. Use the variable if you have it; fall back to a clean, brand-voiced default if you don't. The same discipline applies to your opener, which is why it pays to study writing chatbot greetings before you finalize step one of any flow.
Fallback
Every question node needs an answer for "the user said something I didn't expect." The rule of thumb: re-prompt once with a clarifying nudge, and on the second miss, escalate — offer a human or a callback rather than looping. A flow that asks the same question three times feels broken and is the number-one cause of abandonment. Always include a persistent escape option ("Talk to a human") that works from any step.
Human handoff
Hand off on three triggers: the user asks for a person, the bot fails the same intent twice, or the intent is high-stakes (billing dispute, cancellation, complaint). When you hand off, pass the full transcript and every captured field so the customer never repeats themselves. If no agent is online, the handoff should gracefully become an after-hours capture — template #10 — not a dead end.
Test before you scale
Ship one flow, watch the step-by-step drop-off chart, and fix the cliff. The most common failure is a single question asked too early — like requesting an email before delivering any value. Move that capture later, A/B test the order, and completion usually jumps double digits. Treat every flow as a living funnel, not a finished script: review the drop-off data weekly for the first month, then monthly.
How to choose your first flow
You do not need all ten on day one. Pick the single flow tied to your biggest, most measurable goal and prove it before adding the next. A simple way to decide:
- Leads are the goal → start with lead qualification (#1) and demo booking (#2).
- Ticket volume is the pain → start with support triage (#3) and FAQ deflection (#8).
- Revenue leaks at checkout → start with cart recovery (#4) and the plan recommender (#5).
- Churn is early-life → start with onboarding (#6).
- You miss leads after hours → start with after-hours capture (#10).
Prove one, then layer the next. Three well-tuned flows beat ten neglected ones every time.
Frequently Asked Questions
How long should a chatbot conversation flow be?
Aim for the shortest path that completes the job. Most high-performing flows reach their goal in three to six turns. Every extra question costs roughly 5-8% of completions, so only ask for data you will actually use. If a flow needs more than six steps, split it: capture the essentials first, then continue async by email.
What is the difference between a linear and a branching chatbot flow?
A linear flow asks the same questions in the same order for everyone — ideal for simple, single-intent jobs like newsletter signup or CSAT collection. A branching flow forks based on the user's answer or detected intent. Start linear for narrow tasks and add branches only where user paths genuinely diverge; over-branching creates dead ends and is far harder to test and maintain.
When should a chatbot flow hand off to a human?
Hand off on three signals: explicit request ("talk to a human"), repeated failure (the bot misses the same intent twice), and high-stakes intent (billing disputes, cancellations, complaints, legal or medical questions). Always pass the full transcript and captured fields so the customer never repeats themselves. If no agent is available, switch to after-hours capture rather than leaving the user stranded.
Can I build these conversation flows without code?
Yes. Modern platforms like EzyConn use a visual, no-code flow builder where each step is a node — a message, a question, a branch, a capture field, or a handoff — connected on a canvas. You drag the nodes, type the bot copy, and set branch conditions. An AI layer handles free-text understanding so you do not script every phrasing. Non-technical marketers and support leads ship production flows in an afternoon.
How do I measure whether a conversation flow is working?
Track one primary success metric per flow — qualified leads, booked meetings, deflection rate, recovered carts — plus completion rate and drop-off by step. The step-level drop-off chart is the most useful: a sudden cliff at one question usually means it is confusing, too early, or asking for data the user is not ready to share. A/B test that step, redeploy, and watch the funnel re-balance.
How many conversation flow templates does a typical business need?
Most teams run three to six flows, not ten. Start with the one tied to your biggest goal — usually lead qualification or support triage — prove it, then add the next. A small e-commerce store might run cart recovery, support triage, and a product recommender; a B2B SaaS team typically runs lead qualification, demo booking, and onboarding. Add new flows only when an existing one is measurably winning.
Ship your first flow today
EzyConn includes a visual no-code flow builder, AI intent detection, human handoff, and step-by-step analytics. Start free, or watch it in a live demo.
Last updated . View more guides.