Table of contents
- What are AI agents for WhatsApp customer support?
- How do AI agents decide when to reply on WhatsApp?
- How do you set up an AI agent to reply on WhatsApp?
- What can AI agents do beyond replying on a WhatsApp ticket?
- What can’t AI agents handle on WhatsApp yet?
- How is this different from a generic WhatsApp chatbot?
- How much does an AI agent reply cost on WhatsApp?
- FAQ
What are AI agents for WhatsApp customer support?
LiveAgent’s automated agents can now send real, customer-visible replies inside WhatsApp tickets, not just an opening greeting. The capability landed in early August 2026 (LiveAgent is currently on version 5.69.0), and WhatsApp is only the second channel to get it.
Live chat has had AI auto-reply since 2024. Email doesn’t have it yet. Before this release, a WhatsApp ticket sat untouched until a human agent picked it up, the same as an email does today, no matter what time a customer wrote in.
How do AI agents decide when to reply on WhatsApp?
An AI agent only replies on WhatsApp if you have set up a rule with the “Run AI agent” action. Nothing sends itself by default. That action has two fields: which AI agent runs, and which budget it runs under.
AI Agents also aren’t available out of the box. Your LiveAgent plan needs the AI Agents module enabled, and your own user role needs the right privilege, before you can even create this kind of rule.
Once the rule exists, the agent can reply from the customer’s first message onward, without a person touching the ticket. That includes messages that arrive outside business hours, since the agent doesn’t need anyone to be logged in.
Ready to grow your business?
Start your free trial today and see results within days.
How do you set up an AI agent to reply on WhatsApp?
You set this up the same way you set up any other automation rule in LiveAgent: you create a rule, then add “Run AI agent” as the action. The rule’s channel list covers every channel you’ve enabled for your account, not just WhatsApp: Chat button, Contact form, Email, Invitation (proactive chat), Call, Call widget, Facebook, Facebook message, Instagram mention, Forum, Suggestion (feedback), Telegram, Viber, X, and WhatsApp all sit in the same checklist, so you tick WhatsApp alongside whichever other channels you want the same agent to cover.
Two choices matter most when you build the rule. First, which AI agent runs. Agents can be named per purpose, for example a “WhatsApp Autoreply Agent” set up specifically for that channel, separate from an agent you use elsewhere. Second, which budget the rule runs under. Budget attaches to the rule, not the agent, so the same agent can run under different budgets depending on which rule triggers it. A budget is a monthly cap on tool calls, 300,000 by default, not a dollar limit.
What can AI agents do beyond replying on a WhatsApp ticket?
The same agent that sends the reply can also read the full conversation, search your knowledge sources for an answer, tag the ticket, transfer it to the right department, and mark it resolved once the matter is settled. Those actions work the same way regardless of which channel the ticket came in on, since tagging, transferring, and resolving aren’t tied to WhatsApp specifically.
The agent has well over 30 tools available in total, not just the ones above. Add or assign notes, reopen a resolved ticket, mark something as spam, fill in a custom field, or even create a new knowledge base article are all things the same agent can do while it works a ticket.
This matters because a WhatsApp conversation is not always a single message. A customer might ask a follow-up, and the agent can keep working the same ticket instead of only sending one canned reply.
There is a side effect worth knowing about. Because the agent replies within moments of a customer’s message, that reply lands inside WhatsApp’s 24-hour customer service window . Meta’s own documentation for this window states: “When a WhatsApp user messages you or calls you, a 24-hour timer called a customer service window starts,” and if the user writes again before it expires, “the timer resets to 24 hours.” A fast AI reply keeps that window open by design, instead of a support team accidentally missing it and being forced to fall back on a pre-approved template message.
Join our newsletter
The latest tips and deals delivered straight to your inbox.
What can’t AI agents handle on WhatsApp yet?
AI agents on WhatsApp can’t start a new conversation on their own. This is a LiveAgent limitation today, not a WhatsApp rule: LiveAgent already supports sending the first message on WhatsApp (proactive outreach with an approved template), but only a person can trigger that, not an AI agent rule.
Once the 24-hour customer service window closes, the AI agent can’t send anything at all until the customer writes in again. Sending a template message after the window closes is a manual, human-only action in LiveAgent today. LiveAgent confirmed this directly: in an internal test, a WhatsApp ticket sat for the full 24 hours, and a person on the team was able to send a template message once the window closed, but no rule or AI tool can do that step automatically.
Replies are also text only right now. An AI agent can’t send images, voice notes, or other media on WhatsApp.
It’s also worth being specific about what “automatic” means here. The AI agent only runs where a rule tells it to. A WhatsApp department with no matching rule keeps working exactly as it did before, handled by your human agents.
How is this different from a generic WhatsApp chatbot?
A generic WhatsApp chatbot usually follows a fixed decision tree that someone built in advance, and it typically only replies inside the chat window itself. LiveAgent’s AI agent works inside your existing ticketing system instead, so the same tag, transfer, and resolve actions apply no matter which channel the ticket came in on, whether that’s WhatsApp, email, or live chat. Automatic replying itself is more limited: live chat has had it since 2024, and WhatsApp is the only other channel that has it so far.
| AI agent rule (LiveAgent) | Manual agent reply | Generic WhatsApp chatbot | |
|---|---|---|---|
| Replies outside business hours | Yes, immediately | No, waits for staff | Yes, but script-only |
| Reads and searches your knowledge sources | Yes | Yes, manually | Rarely |
| Tag, transfer, or resolve the ticket, on any channel | Yes | Yes, manually | No |
| Needs a decision tree built per flow | No, uses existing rules and knowledge | Not applicable | Yes |
| Also sends automatic replies on | Live chat only, since 2024 | Every channel, manually | Nowhere else |
How much does an AI agent reply cost on WhatsApp?
Two separate things affect cost here, and they’re easy to mix up. The rule’s budget is a monthly cap on tool calls (300,000 by default), not a dollar limit, so it controls how many times the agent can act, not what each action costs. The actual dollar cost per reply depends on which AI provider and model the agent uses, or, for a FlowHunt agent, how the flow itself is built, plus how much it needs to search your knowledge sources to answer.
To put a rough number on it: in early internal testing on a small sample of two messages, a WhatsApp AI agent reply using GPT-5.4 cost about $0.26 each. That is one limited test, not a published rate, and a cheaper model would cost less per reply.
There is a second cost to plan for that has nothing to do with LiveAgent. Starting October 1, 2026, Meta begins charging businesses directly for WhatsApp service messages, the free-form replies sent inside that 24-hour customer service window, once a business phone number goes over 1,000 free messages in a month. That charge lands at what Meta calls the service rate, which mirrors the utility and authentication pricing already used in each market, according to 360dialog’s coverage of the change . An AI agent replying automatically doesn’t change how Meta counts those messages. It just means the messages get sent faster.
How LiveAgent helps
LiveAgent is our product, and this article covers a feature we built, so treat that as the disclosure it is. If you already use LiveAgent for WhatsApp, an AI agent rule is the fastest way to close the gap between “customer messages at midnight” and “customer gets a real answer,” without hiring a night shift.
If you’re evaluating help desk software for WhatsApp support, this is one of the areas where a channel-agnostic ticketing system has an edge over a WhatsApp-only chatbot tool: the same tagging, transferring, knowledge base, and reporting you’d use for email or live chat carries over to WhatsApp without a separate setup.
Conclusion
If your team already runs WhatsApp support through LiveAgent, the fastest next step is setting up one AI agent rule and watching what it catches overnight. Turn on an AI agent rule for your WhatsApp inbox and see how much of tomorrow’s queue is already answered before your team logs in.
