levelpartner.ai
AI agents and AI for business

AI chatting bot: how it works and how to set it up

An AI chatting bot that only lives on your website misses the conversations your clients start elsewhere: a text reply to a reminder, a WhatsApp message on a Saturday, a direct message on Instagram after seeing a post. On the levelpartner.ai platform one AI agent can answer across web chat, SMS, WhatsApp, Facebook and Instagram messages, with every exchange on the same contact record. This article answers how to run one bot across several channels without it sounding wrong on any of them.

One brain, several channels

The advantage of a single multi-channel bot is continuity. A lead who chats on your site on Tuesday and texts on Thursday is recognised as the same person, and the bot picks up where the conversation left off. Your team sees one thread per contact in Conversations, not five apps to check.

The challenge is that each channel has its own etiquette. A three-sentence reply that suits web chat is long for SMS. A formal tone that suits email sounds odd on Instagram. The bot's knowledge stays the same; its style must flex.

How each channel differs

Web chat visitors are on your site, often comparing. They expect quick, informative replies and a clear next step.

SMS is personal and short. Replies should be one or two sentences, and messaging outside sensible hours is unwelcome. SMS also carries consent and opt-out obligations in the US, the UK and most other markets.

WhatsApp is conversational, popular with UK clients and international buyers. Business messaging there has its own rules about when you can start a conversation versus reply to one.

Facebook and Instagram messages come from people who know you from social content. They are often casual and early in their decision.

Before you start

  • Each channel connected to your account: chat widget, an SMS number, WhatsApp and your social pages.
  • An AI agent with a knowledge base that already works well on one channel.
  • Consent and opt-out handling in place for SMS and WhatsApp.
  • A rule for quiet hours: when the bot should not start or continue a conversation by text. The agent's Working Hours setting on the Deploy tab lets you tick which channels follow a schedule.
  • A channel section in the agent's prompt, with a reply length and tone note for each channel.

How to set it up

  1. Go to AI Agents > Conversation AI and open the Agents List tab.
  2. Select the bot, then click Deploy.
  3. Turn on SMS first and test from your own phone.
  4. Open the channel's edit icon if it has extra settings.
  5. Turn on Assign channel and choose the number or page.
  6. Add any Has tags and Doesn't have tags conditions.
  7. Click Update, then click Save for the bot.
  8. Repeat for WhatsApp, Facebook and Instagram, testing each one.

Start with one, add one at a time

Do not switch on five channels in one afternoon. Run the bot on the channel with the most volume, fix its answers, then add the next. Each new channel brings new kinds of messages. Social messages bring more casual and off-topic questions. SMS brings replies to your own reminders, which need the bot to understand what it sent earlier.

Illustration: a 60-person real estate firm adding channels in stages

Suppose the firm runs the bot on web chat for a month, then adds SMS replies to viewing reminders, then Instagram messages, with a week of review before each new stage. On that plan, by the third month the bot would answer on all three, viewing changes by text would be handled automatically and Instagram questions about listings would get a reply within minutes.

What it replaces

A multi-channel bot replaces several separate auto-replies, a shared phone for texts, and someone checking social inboxes twice a day. For a firm of 10 to 250 people, those inboxes are often where leads are lost, because nobody owns them. Putting one agent and one team behind all of them gives every message an owner.

The common mistake

The common mistake is copying the web chat style everywhere. The bot then sends three-paragraph replies by SMS and stiff, formal messages on Instagram. Give each channel a one-line style rule and test it on a phone before going live.

Weekly running time

Twenty to thirty minutes a week once all channels are live. Filter Conversations by channel and read a few from each, especially the newest. Messaging usage on SMS, WhatsApp and calls is billed as used, so check volumes monthly if the bot sends follow ups.

When not to put the bot on a channel

Do not attach the bot to a channel your clients use for sensitive matters with a named person, such as a mortgage adviser's direct WhatsApp with clients mid-application. Do not let it start outbound conversations on channels where you lack consent. And if a channel gets only a few messages a month, a notification to a person may be simpler than a bot.

Troubleshooting

The bot does not recognise a returning contact. The contact used a different number or social account. Ask for an email on first contact so records can be matched, and merge duplicates.

SMS replies are too long. The channel style rule is missing or too soft. State a hard limit of two sentences for SMS.

Social messages go unanswered. The page connection has expired. Reconnect the social accounts in Settings and test.

The bot replies to opt-out messages. Check opt-out handling for the SMS number and make sure the bot stops on those keywords.

Frequently asked questions

Can one AI chatting bot handle all channels at once?

Yes. One agent can answer on web chat, SMS, WhatsApp and social messages, with channel-specific style rules in its instructions. Many firms still use a second agent for existing clients, because their questions differ from new leads.

Do customers know which channel they are on?

They do, and they expect replies to match. That is why the style rules matter. The content of the answer can be the same; the length and tone change.

What about voice calls?

Calls need a voice agent, which can share the same knowledge. See AI phone answering. For channel strategy in depth, read omnichannel conversational AI and AI SMS marketing.