WhatsApp AI customer support that knows its limits.
WhatsApp AI customer support is an AI agent that answers on your WhatsApp Business number from your own knowledge base, takes actions through the tools you connect, and hands the conversation to a person when it should. Cloodot runs that agent inside the inbox your team already works in, so a handover is a change of owner, not a change of app.
No credit card required · 7-day free trial
My parcel says delivered, but it isn't here.
It was marked delivered this afternoon. Carriers sometimes scan a parcel a little early. If it hasn't arrived by tomorrow, tap below and a teammate will open a claim for you.
What WhatsApp support actually asks of you
Customers treat WhatsApp like a chat with a person. They send a photo instead of describing the damage, record a voice note from the car, and expect an answer while the app is still open. Most of what they ask repeats, and a lot of it arrives outside the hours a support team can cover.
WhatsApp also has rules that email and webchat don't. When a customer messages you, a 24-hour customer service window opens. Inside it you can reply freely, and Meta doesn't charge for those replies. Once it closes, the only way to reach that customer is an approved message template, and any template you send first needs the customer's opt-in.
Templates come in three categories, marketing, utility and authentication, and Meta bills them per delivered message at rates that depend on the category and the customer's country. A bot that lets threads drift past the window ends up paying to restart conversations it could have finished for free.
How WhatsApp AI customer support works in Cloodot
A message arrives on your number and opens a conversation in the inbox. Cloodot assigns your default AI persona, and the agent reads what the customer actually sent: text, a photo, a voice note or a PDF. It searches your Knowledge Base and FAQs by meaning rather than exact words, so "can I send this back?" finds the returns policy you titled "Refunds".
When the answer depends on live data, the agent calls a skill. With the Shopify skill set it can look up an order, and on WhatsApp the number the message came from is what identifies the customer. For recurring scenarios you write procedures: ordered steps, skills and quick reply buttons such as "Track order" or "Talk to someone".
When it can't answer, it says so instead of guessing: the question lands in the FAQ Unanswered tab, and the agent sends your fallback message or hands over. Handover conditions are written in plain language, such as "customer asks for a refund above the policy limit", and sit alongside built-in triggers for an explicit request for a human, frustration and sensitive issues. Every handover carries a priority and a one-line reason, so the teammate who picks it up starts where the agent stopped.
If the customer goes quiet, the agent can check in after a delay you choose, from one minute up to 23 hours, which keeps the follow-up inside the 24-hour window.
What Cloodot does here
Knowledge Base and FAQs
Add text, PDFs, spreadsheets, web pages or a crawl of up to 200 pages, with auto-crawl for pages that change. Questions the agent couldn't answer collect in one list, and each answer you write becomes an FAQ.
Explore AI agentProcedures
One scenario handled end to end, a missing parcel or an address change, as an ordered list of steps, decisions, skills and quick replies, so the agent handles it the same way every time.
Skill Sets
Installed skills let the agent act, not only answer: order status from Shopify, shipment tracking from Shiprocket or AfterShip, invoices and stock from ERPNext. You enable them per persona.
Explore Skills & workflowsHandover with priority and team routing
The agent rates urgency on a six-level scale, records why, and moves the conversation into the team inbox whose description fits, or to Queued when none does.
Set up WhatsApp AI customer support in Cloodot
5 steps
Connect your WhatsApp number
In Settings → Organization → Channels, add WhatsApp and sign in with the Facebook account that administers your Meta Business account. Add a payment method in Meta Business Suite, which the API requires. A number already on the WhatsApp Business App can stay there with Coexistence Mode.
Fill the Knowledge Base
Under Intelligence → Knowledge Base, crawl your help pages or add your policies as text and files. Turn on auto-crawl for pages that change, such as delivery times.
Configure the persona
In Intelligence → Agent, set the name, greeting, reply languages, response length and personality. Write your handover conditions, and a fallback message that offers a person.
Give your top requests a procedure
Take the three or four things customers ask most and write a procedure for each, with the steps, skills and quick replies it needs. Install a skill set such as Shopify if the answers depend on order data.
Test, deploy, then review weekly
Run real customer questions through the playground, then Save & Deploy; each deploy is a version you can roll back. Once a week, answer what has collected in the FAQ Unanswered tab.
Who it suits, and what to check before you choose
It fits any business whose WhatsApp number gets the same twenty questions every day and a handful of hard ones. If most of your inbound is genuinely one-off, such as custom quotes, use the agent as a first responder that collects the details, not as the one that resolves them.
Whatever tool you evaluate, test four things. Does it answer only from your content, and admit when it can't? Can it act on your systems with an identity check? Does the handover land in the same thread with the reason attached? And does it respect the 24-hour window and template rules, or leave you to find out from a failed send?
The usual pitfalls: a knowledge base written as marketing copy instead of answers, handover rules so broad the agent escalates everything, and nobody reading the unanswered list.
Questions teams ask
How much does WhatsApp AI customer support cost?
There are two bills. Meta charges for template messages per delivered message, by category and country, through the payment method in your Meta Business Suite; replies inside the 24-hour customer service window are free. Cloodot's AI agent uses message credits from your plan for its replies. If credits run out, the agent steps back and conversations wait for your team.
Can the AI reply after the 24-hour window closes?
Not with a free-form message. WhatsApp doesn't allow that for anyone. Once the window closes you reach the customer with an approved template, and their reply opens a new window in which the agent carries on. In Cloodot workflows, a 24h Reply Window condition lets you branch between a normal message and a template automatically.
Do I need a new phone number?
No. If your number runs on the WhatsApp Business App, Coexistence Mode connects it to the API while it keeps working in the app, provided it has been active there for at least seven days. If it's with another WhatsApp API provider, you can migrate it to Cloodot and keep the same number.
What happens when the AI doesn't know the answer?
It doesn't improvise. The question is logged in the FAQ Unanswered tab, and the agent either sends your fallback message or hands the conversation to a person, depending on your handover conditions. Answer the logged question once and it becomes an FAQ the agent uses from then on.
Can it check orders and account details?
Yes, through skills. With the Shopify skill set the agent can report an order's status and tracking. On WhatsApp the phone number the message came from identifies the customer. On channels where anyone can type any email, it asks for the order number plus the email or phone used at checkout before it shows an order.
Which languages can it reply in?
It understands messages in any language and replies in the ones you choose on the persona: English, Arabic, Hindi, Urdu, Spanish, French, German and Malayalam. It matches the customer's language among those, and uses your first choice when it can't tell.
Bring us your Monday.
Start free. Train a Cloodot agent on your own docs, your own tone, your own tools, and you'll walk away with a working deployment in your inbox the same afternoon.