Draw the conversation once, answer in under a second
A canvas where you design automated conversations: ask, route by answer, check a condition, call your systems, wait a day, run an AI agent or hand off to a person. Every reply is instant, every path is consistent, and every hand-off carries the full context.
24 step types ✦ AI agents inside the flow ✦ Test chat before every deploy
The most guest-obsessed brands trust Lodgestory to power 50M+ conversations a year
BBingo Hotels
Instant replies, consistent paths, nothing lost at hand-off
When a message arrives on a channel with a live journey, Lodgestory starts a session at the Start step and walks the graph. Message steps send and continue. Input steps wait for the guest. Logic steps route or pause. Action steps call an API, run an AI turn, file a ticket or hand the chat to a person. Every run has its own variables, its own state and its own trail, visible in the inbox when a human takes over.
- Built for real messaging. Button counts, list limits, headers and footers model what the channel actually sends.
- One journey, every channel. Buttons on WhatsApp, clickable choices on web chat, HTML on email, a numbered list on Instagram.
- Follow-up nudges. Any waiting step can message again when the guest goes quiet, once or as a ladder, and stops the moment they reply.
Session trail
Built for the tyranny of small questions
Reply in under a second, all night
Journeys execute the moment a message arrives. The easy half of inbound is answered before an agent would have seen it.
Learn moreQualify, route and book
Twenty-four step types cover the patterns you actually need: buttons, lists, WhatsApp Flows, prompts with validation, conditions, delays, API calls, AI, tickets, hand-offs.
Learn morePut the AI where it matters
Validate a free-text answer with AI instead of a regex, run a conversational or knowledge agent as a step, or pass an API response through AI before using it.
Learn moreTest before you ship
A built-in test chat walks every path in a sandbox. Deploy runs validation and flips the journey from Draft to Live, deliberately.
Learn moreWhat you draw is what the guest receives
Each step models the channel's real message. Reply buttons come in ones, twos and threes, quick replies up to ten, lists in up to ten sections of ten rows. A prompt validates free text by regex, phone, email, number or a plain-language AI rule, retries up to five times and can time out to a fallback branch. A WhatsApp Flow step collects a whole form in one screen and maps every answer into variables.
- Date prompts with format, range and excluded weekdays, so "no Sundays" is a setting, not a check.
- Media and locations. Image, video, audio, document with captions, a map pin, a call-to-action link.
- Continue in browser chat. Move the conversation to a private chat page at a quarter of WhatsApp's per-reply cost.
Steps
Route by answer, by condition, by the clock
A Condition evaluates AND/OR rules over variables with equals, contains, greater-than, regex and exists. A Time Condition routes by day and time in a zone you choose, with named slots for open hours, weekends and public holidays, checked top to bottom so exceptions win. Route now, or hold the conversation until a slot opens and let Saturday night's enquiry walk into the front-desk branch on Monday at nine.
- Delays and waits. Send the welcome now and the restaurant list tomorrow morning; wait for the next message with a timeout branch.
- Holiday lists for your country, or your own upload, added per year.
- Variables it sets. The matched slot, local time and the next opening, ready for the message that says "we will pick this up at 9:00 AM Monday".
An AI turn inside the flow, with your tools and knowledge
Conversational mode is a fast single pass for greetings, qualifying questions and one-shot tool calls. Knowledge Agent mode searches your knowledge bases, pulls live facts from connected data sources and takes several steps before it answers. Either way you choose the model, write the persona, attach the tools and set the loop cap. Switch on chat actions and the agent can add a note, label the chat, move its stage, file a ticket or hand over, using only what you allow.
- Quick replies, voice notes, images. Tap-to-answer options, transcribed voice notes, pictures sent to a vision model.
- Approvals. A change marked Needs approval is proposed to your team, never run on the agent's say-so.
- Earlier messages. Let the agent read the last 5 to 50 messages so it picks up a conversation already under way.
AI Agent stepKnowledge AgentEdit safely in draft, go live deliberately
A journey is Draft or Live. In Draft you iterate and walk every path in the test chat without touching a customer; ticket and stage steps show a one-line preview of what they would do. Deploy runs validation, needs a Start step plus one more and every link pointing at a real step, then flips the journey to Live. Assign it to a channel in Connections and the next inbound message starts a session. Goals tag the milestones you care about and count how often they are reached.
- One live journey per channel. Moving back to Draft ends running sessions on purpose, so old flows never linger.
- Bot identity. Replies are attributed to a team member with the Bot role, so the inbox tells bot from human.
- API for everything. Journeys, steps, variables and sessions through the partner API.
Before you deploy
Channels
Goals · this week
Go deeper
Everything a journey is made of
The visual builder
Palette, canvas, property panel, variables, test chat and deploy validation, in one screen.
Learn moreStep types
All twenty-four steps in five groups, with the limits each one honours on every channel.
Learn moreKnowledge bases
The documents and live data sources a Knowledge Agent step answers from, with references.
Learn moreIn practice
Two hundred messages between 11 PM and 7 AM
A villa operator's single journey qualifies each after-hours enquiry: dates, party size, budget. An API step checks availability against the property system, a text step quotes a price, a date prompt books a callback. Anything ambiguous goes to an AI Agent step. Come morning, agents see qualified enquiries, not two hundred "hi, is the villa free?".
Automation you can audit
Every run is a session tied to the conversation, with its own variables, state and message trail. Test runs never touch customers. Bot replies carry a bot identity, and every ticket, stage change and hand-off a journey makes shows up in the same history, webhooks and reports as a team member's.
Visit the trust centerGuides for the people who build journeys
Create, add a greeting, connect, save, test and deploy, then assign a channel.
Post-chat CSAT, missed-call catch, no-reply escalation and the other recipes.
How a transfer from a journey lands on an advisor with capacity, or waits in the team queue.
Ready to answer the easy half automatically?
Draw a greeting, a question and a hand-off, test it in the sandbox and put it live on a channel today.
Questions, answered
The things teams ask before they automate a conversation.
A conversation you draw on a canvas instead of writing code. Blocks ask questions, route by answer, check a condition, call an external system, wait a day, run an AI agent or hand off to a person, and you wire them together. When a message arrives on a channel with a live journey, Lodgestory starts a session from the Start step and walks the graph, replying in under a second.
WhatsApp (official and unofficial), email and web chat are fully supported; Instagram and Facebook Messenger run with graceful fallbacks for interactive elements. One journey can be assigned to several channels: buttons are real buttons on WhatsApp, clickable choices on web chat, styled HTML on email and a numbered list on Instagram.
No. Every step is a block with a property panel: the text, the buttons, the validation rule, the API URL and its response mapping, the time slots. Variables are typed as {{scope.name}} inside any message. The partner API covers every part of a journey if you prefer to build programmatically.
That is a design decision you draw. A free-text prompt can retry up to five times or follow a fallback branch; an AI Agent step answers from your instructions, tools and knowledge bases and hands over when it reaches a hand-off signal; the Agent Transfer step gives the conversation to a team or person. The whole bot trail stays visible in the inbox, so the human picks up with full context.
Yes. The API call step makes GET, POST, PUT or DELETE requests with variables in the URL, headers and body, extracts fields from the response into variables and can run the raw response through AI first. The AI Agent step can also call the tools you define in AI Agents.
Every journey has a built-in test chat that walks the flow in a sandbox without touching a real customer; ticket and stage steps show a one-line preview of what they would do, and a Time Condition can be tested as if it were any day and time. Journeys are Draft or Live; deploying runs validation and flips the state, and moving back to Draft ends running sessions deliberately.