Answers the questions you keep answering
Hours, coverage, meet-and-greet, child seats, cancellation terms. You write the answers in your own words and it uses your wording, not a paraphrase of it.
An AI assistant that answers your website, WhatsApp and SMS conversations the moment they arrive — prices the trip from your own rates, takes the booking, and hands anything it should not touch to your team.
It answers from what you have taught it, and it reaches for your own systems when an answer needs one — the rate engine behind your booking site, and the booking record behind a verified customer.
Hours, coverage, meet-and-greet, child seats, cancellation terms. You write the answers in your own words and it uses your wording, not a paraphrase of it.
Every active customer-facing vehicle type in one reply, with your zone rules, stop charges and minimum hours — the same fares your booking website would quote. Vehicle types you keep internal never appear.
Pickup, drop-off, date and time, passengers, and the hours on an hourly booking. It asks for whatever is missing instead of guessing, and reads the matched addresses back.
A customer signed in on your website is recognised straight away. Anyone else is offered a one-time code first — and the code goes from the generator into the message without the model ever seeing it.
Draft-only mode writes the reply as a private note on the conversation. Your team reads it, edits it, sends it. Nothing reaches the customer on its own.
An after-hours assistant answers what it safely can, takes the details for everything else, tells the customer when someone will be back — and closes idle threads on your timer.
Colour, corner, greeting, the label beside the bubble and the assistant's own name and tone — all set in your console, all live on the next page load. Try it.
Nothing on your page says LimoSphere. Restrict the widget to your own domains and it refuses to boot anywhere else.
Pick the channels it answers on. Same instructions, same written answers, same guardrails — and every conversation lands in the same shared inbox as the ones your team handles.
On website chat it is also told whether the visitor is on a phone or a computer, their timezone and their browser language — enough to judge urgency, and never mentioned back.
A brand-new assistant gets three questions rather than a blank form — what it is for, where it works, and how it sounds. Everything it writes is an ordinary profile you can edit the moment set-up ends.
Its name, its business context, the handoff and resolution messages your customers read, and up to twenty guardrails in your own words.
Your questions and your answers. Every active entry goes into the assistant's instructions word for word — a meter shows how much room is left.
Which inboxes it answers on, and the one script tag your website needs. Add a channel without leaving the page.
Your real widget on a mock page, answered by your real assistant — live rate quotes and all. Nobody is notified, nothing counts against your caps.
Every run it made: replied, resolved, handed off, failed or budget-capped, with the tools it used and how long it took.
Rate figures are estimates until a person confirms them, and the assistant is instructed to say so.
Channels → Connect a channel → Website chat. Name it, pick the colour, the corner and the welcome copy.
Copy it from the inbox settings and paste it before the closing body tag. It never changes again — later appearance edits reach your site on the next load.
<script>
window.$sphereChat = { websiteToken: 'your-token' };
</script>
<script async src="https://crm.limosphere.com/widget.js"></script>Chatbot → Channels & install, tick the new inbox, and every website conversation opens with the assistant answering.
Whatever you teach it, plus what it can look up. You write the business context and a list of your own questions and answers — hours, service areas, meet-and-greet policy, cancellation terms — and the assistant prefers your wording over its own. On top of that it can price a trip against your live rate engine and, for a customer whose identity it has verified, answer questions about their bookings.
Yes, if you switch rate estimates on. It prices the trip the way your booking website does — every active customer-facing vehicle type in one answer, with your zone rules, stop charges, minimum hours and, for a verified customer, that customer's own fare terms. It reads the addresses back so the passenger can correct a mis-matched one, and it says the figures are estimates.
The harness forbids inventing prices or availability before you add a single rule, and your guardrails sit on top of it — up to twenty hard rules in your own words. When it cannot answer, it hands the conversation to your team rather than guessing.
Three ways, all deliberate: it decides it cannot answer, one of your agents replies publicly, or it fails or hits a budget cap. Every one of them sends the customer your handoff message, moves the conversation into your team's queue and notifies them. A conversation a person has taken over stays theirs.
Yes. In draft-only mode the same assistant writes its answer as a private note on the conversation and one of your team reads, edits and sends it. Nothing reaches the customer until a person presses send.
The bubble colour, which corner it sits in, the welcome title and tagline, the label beside the bubble and the assistant's own name and tone are all yours. The widget carries no LimoSphere branding on your site.
Yes. The assistant answers on whichever inboxes you pick — website chat, WhatsApp, SMS, email or Telegram. It is the same assistant with the same instructions and the same knowledge everywhere.
Connect the website-chat channel in your console, then paste one script tag before the closing body tag on your site. The snippet never changes, so colour, greeting and position edits reach your pages on the next load without touching the site again. You can also restrict it to your own domains so nobody can host your chat elsewhere.
It is created switched off, every time. The Playground runs your real assistant in your real widget on a mock page — same pipeline, same tools, live rate quotes — with nobody notified and nothing counted against your daily limits. Turn it on when you like what it says.
Every run is logged: the outcome (replied, resolved, handed off, failed, superseded or budget-capped), the tools it used, tokens and duration, filterable by conversation. And 'Handled by chatbot' in Communications lists every conversation it has ever answered, however that conversation ended.
We'll set the assistant up on your own rates and your own answers, and let you try to catch it out.
Live on your site the same week.