Travel Agency AI Assistant with ChatSyncs — Implementation Guide
Step-by-step guide to building a ChatSyncs AI Agent for a travel agency — the Training Campaign, Prompt Message, and routing Talk to Human and Sales intents.
On this page
- Prerequisites
- Implementation Guide
- Step 1 — Enable the AI Agent and attach a Training Campaign
- Step 2 — Write the Prompt Message
- Step 3 — Build the AI Training Campaign
- Step 4 — Route "Talk to Human" requests with Bot Flow Trigger
- Step 5 — Route sales/complaint messages with Agent Assignment
- Step 6 — Point specific topics at a dedicated campaign with AI-Driven Responses
- Features Used
- Expected Result
- FAQs
- Related Documentation
Prerequisites#
- Read AI Agent first to understand how the whole feature works — this guide only covers how the pieces combine for a real business, not the feature itself.
- A defined business profile to train on — company info, services, destinations, pricing, and policies (see the TravelEase Holidays profile used throughout this guide, below).
- A decision on tone and boundaries for the Prompt Message — how long responses should be, what the AI should never say, and when it should hand off to a human.
- Whichever training source you'll use ready to go — a written FAQ list, a website URL, a brochure/PDF, a Google Sheet, or an API endpoint. Only one source is active per campaign at a time (see How to Train Your AI Chatbot for the mutual-exclusion rules).
- A list of intents worth detecting (e.g. "Talk to Human", "Sales Enquiry") and which flow or team each one should route to.
Implementation Guide#
Step 1 — Enable the AI Agent and attach a Training Campaign#
Objective: Turn on the component that actually talks to the customer.
Explanation: The AI Agent, trained per How to Train Your AI Chatbot, receives the message, searches the configured training data, understands the intent behind it, and generates a natural response — it does nothing until a Training Campaign is attached.
Actions to perform:
- Open Chatbot Manager → AI Agent and select (or create) the TravelEase Holidays AI Training Campaign.
Expected result: A customer message like "I want a honeymoon package" gets an AI-generated recommendation of honeymoon destinations, followed by a budget question.
Notes: Without an attached campaign, the AI Agent has no business-specific knowledge to answer from at all.
Step 2 — Write the Prompt Message#
Objective: Define how the AI behaves — role, personality, tone, response length, restrictions, and conversation style.
Explanation: The Prompt Message is what stops an AI Agent from confidently inventing prices, dates, or policies it was never told.
Actions to perform:
- For a fast-moving WhatsApp conversation, a short numbered rule set works well:
1. Never write more than 40 words in a single response.
2. Never use bullet points, numbered lists, or long explanations unless the user explicitly asks for them.
3. Answer only the customer's current question. Do not provide additional information that wasn't requested.
4. End your response with ONE relevant follow-up question whenever appropriate.
5. Use simple, conversational language.
6. Do not repeat information.
7. Do not greet the customer in every message.
- For a business that needs a strict role, tone, and full set of boundaries defined at once, a much longer template covers every dimension — role, personality, communication style, tone, response format, knowledge usage, conversation behavior, restrictions, escalation rules, and safety, ending with the AI's overall goal.
Expected result: The AI's responses match the rules — short and follow-up-driven for the concise version, or fully scoped and boundaried for the comprehensive version.
Notes:
Step 3 — Build the AI Training Campaign#
Objective: Give the AI a knowledge base to answer from — without one, it has nothing business-specific to say.
Explanation: TravelEase Holidays' campaign contains company information, services, destinations, FAQs, budget recommendations, and conversation rules — fed from exactly one of four sources (only one can be active at a time).
Actions to perform:
- Add direct question/answer pairs under Content or FAQ — e.g. "Do you provide visa assistance?", "Can I customize my trip?", "Which destinations are best for honeymoon?".
- Instead of FAQs, choose one other source depending on how the data changes:
- URL Training — the AI learns directly from the company website; refresh training whenever the website content changes.
- File Training — upload a PDF, DOCX, catalog, or brochure (e.g. TravelEase's package brochure) and the AI reads it directly.
Google Sheet Training — structured, frequently-updated data without editing the AI manually, e.g.:
Destination Price Days Bali ₹80,000 5 Dubai ₹95,000 6 - API Training — retrieves live information instead of static text: flight status, hotel availability, package availability, currency rates.
Expected result: When a message comes in, the AI Agent works through its sources in order until it has enough to answer:
Customer message
→ AI Agent receives it
→ Searches the AI Training Campaign
→ Reads the Prompt Message
→ Searches FAQs
→ Searches Files
→ Searches URL content
→ Searches Google Sheet
→ Calls API (if configured)
→ Generates response
→ Replies to customerNotes: Use FAQs/File Upload for information that rarely changes, Google Sheet for structured data someone else updates often, and API for anything true right now and false five minutes later.
Step 4 — Route "Talk to Human" requests with Bot Flow Trigger#
Objective: Hand off to a fixed, structured flow instead of letting the AI attempt an answer.
Explanation: Intent Detection reads the customer's message, matches it to a named intent, and performs one of three actions — routing the conversation instead of sending every message through the same knowledge base.
Actions to perform:
- Create an intent named Talk to Human matching phrases like "I want to talk to a human", "Connect me with an agent", "I need customer support", "Can I speak with someone?".
- Enable Bot Flow Trigger, pointed at a Human Support flow built following How to Create a Keyword-Based WhatsApp Chatbot, that asks the customer to pick Sales, Technical Support, Billing, or General Enquiry, then collects Name, Phone Number, Email, and Query before handing off.
Expected result: "I want to talk to a human" → Bot Flow Trigger starts the Human Support flow → customer selects Sales → the conversation is assigned to the Sales team with no manual intervention.
Notes: Use Bot Flow Trigger whenever the business wants structured data collection or a fixed path, rather than letting the AI free-answer.
Step 5 — Route sales/complaint messages with Agent Assignment#
Objective: Skip the AI entirely for messages that should go straight to a team.
Explanation: Agent Assignment sends the conversation straight to a team, with an automated acknowledgement, instead of the AI attempting to answer a sales or complaint message itself.
Actions to perform:
- Create an intent named Sales Enquiry matching "I want to buy", "I need pricing", "Send me a quotation", "Product cost", with Agent Assignment enabled, assigned to the Sales team.
Repeat the same pattern for other intents and teams:
Intent Assigned team Sales Enquiry Sales Complaint Customer Care Technical Issue Technical Support
Expected result: A message matching "Sales Enquiry" is assigned straight to the Sales team — the AI never attempts to answer it.
Notes: A typical team structure behind these assignments — Marketing (campaigns, promotions), Sales (enquiries, pricing, purchases), Customer Care (complaints, refunds), Technical Support (product issues), HR (recruitment) — decides how many intents you'll need.
Step 6 — Point specific topics at a dedicated campaign with AI-Driven Responses#
Objective: Let the AI keep answering, but from a more focused knowledge base than the account's default.
Explanation: AI-Driven Responses is how one account runs several AI "experts" side by side — each intent pointed at its own training campaign.
Actions to perform:
- Create an intent named Travel Questions matching "Holiday Packages", "Visa", "Hotels", "Flights", "Destinations", with AI-Driven Responses enabled and pointed at the Travel Assistant campaign.
The same pattern lets other departments run their own dedicated campaign:
Intent Dedicated training campaign Contains Travel Questions Travel Assistant Holiday packages, visa, hotels, flights, destinations Healthcare Hospital Assistant Doctors, departments, appointments, medical services Real Estate Property Assistant Property listings, pricing, amenities, site visits Education School Assistant Admissions, courses, fees, timings, faculty information E-Commerce Shopping Assistant Products, inventory, orders, shipping, returns
Expected result: "I want to know about honeymoon packages" → Intent Detection matches "Travel Questions" → AI-Driven Responses answers from the Travel Assistant campaign instead of the account's default one.
Notes: Without Intent Detection, every message is processed by the same AI knowledge base regardless of what's actually being asked.
Features Used#
| Feature | Purpose | Documentation |
|---|---|---|
| AI Agent | Talks to the customer using trained knowledge | AI Agent |
| AI Training Campaign | Stores the business's knowledge (FAQs/URL/File/Sheet/API) | How to Train Your AI Chatbot |
| Prompt Message | Defines the AI's tone, length, and boundaries | How to Train Your AI Chatbot |
| Intent Detection | Routes messages to a human, a team, or a different campaign | Agent Routing |
| Keyword Replies | Runs the structured Human Support flow | How to Create a Keyword-Based WhatsApp Chatbot |
Expected Result#
- Customers get instant, accurate answers about packages, visa, hotels, and flights at any hour.
- Every customer gets the same consistent tone and information, governed by the Prompt Message.
- "Talk to Human" requests are collected and handed off with no manual intervention.
- Sales enquiries and complaints skip the AI entirely and go straight to the right team.
- Multiple knowledge bases (Travel, Healthcare, Real Estate, Education, E-Commerce) can run side by side in the same account, each answering only its own topic.
FAQs#
How do multiple AI Training Campaigns work in one account?
Create one training campaign per topic or department (e.g. Travel Assistant, Hospital Assistant), then have each Intent Detection intent point its AI-Driven Responses action at the matching campaign — see Step 6 above.
When should a message trigger a Bot Flow instead of an AI answer?
When the business wants structured data collection or a fixed path — e.g. escalating to a human, or booking an appointment — use Bot Flow Trigger instead of letting the AI free- answer. See Step 4 above.
How do I stop the AI from answering sales or complaint messages itself?
Create an intent for it (e.g. Sales Enquiry, Complaint) with Agent Assignment
enabled, pointed at the right team. The message skips the AI and goes straight to that team
— see Step 5 above.
In what order does the AI Agent check its knowledge sources?
It searches the AI Training Campaign, reads the Prompt Message, then checks FAQs, Files, URL content, Google Sheet, and finally calls the API if one is configured — see the Expected Result in Step 3 above.
Can this same setup work for a business that isn't a travel agency?
Yes — the structure (AI Agent + Training Campaign + Prompt Message + Intent Detection) stays identical. Only the training content and the intents you define change — swap the business profile in Prerequisites for a hospital's, school's, or store's own details.
What stops the AI from inventing a price or policy it was never told?
The Prompt Message — an explicit restriction telling the AI what to refuse and when to hand off to a human instead of guessing. See the Warning in Step 2 above.
Which training source should I use for pricing that changes often?
Google Sheet Training for data someone updates regularly (e.g. destination prices), or API Training for anything that needs to be live right now, like flight status or hotel availability. See Step 3 above.
Does a 'Talk to Human' request lose any customer details in the handoff?
No — the Bot Flow Trigger's Human Support flow collects Name, Phone Number, Email, and Query before handing off, so the receiving team doesn't have to ask again. See Step 4 above.
Related Documentation#
- AI Agent — the full feature guide this use case builds on.
- How to Train Your AI Chatbot — the four training sources and their mutual-exclusion rules.
- How to Create a Keyword-Based WhatsApp Chatbot — the bot flow builder used for the Human Support handoff.
Start over from the beginning of this use case.
Was this page helpful?
Still stuck? Open a support ticket and our team will help.