A WhatsApp AI agent that works for your business, not instead of you.

A WhatsApp AI agent for business is software you talk to on WhatsApp, by text or voice note, that carries out real work for you: answering customers, following up with people who have gone quiet, replying to reviews and updating your business apps. Unlike a chatbot or an auto-reply, it understands what you ask, takes the steps itself, and with Flann waits for your approval before anything is sent or changed.

That is the short version. The longer one matters, because "AI on WhatsApp" now covers three very different things. Only one of them takes work off your plate.

What a WhatsApp AI agent actually is

Start with the word agent. An agent acts.

You give it a goal in plain language, the way you would brief a capable member of staff. It works out the steps, looks up what it needs, drafts what has to be written, and comes back to you with something ready to go. You never fill in a form or build a flow. You just say what you want.

With Flann, that conversation happens in WhatsApp, or in the Flann app if you prefer. You can type, or you can send a voice note while you are walking between the stockroom and the counter. "Reply to everyone who asked about opening hours this weekend" is a complete instruction. So is "Find out who hasn't ordered since the summer and draft them something friendly."

An agent is also not limited to one inbox. Flann answers customers on WhatsApp, Instagram and Facebook (messages, comments and posts), TikTok messages and comments, and Google Business reviews. It posts to your social accounts. And it works inside more than 1,400 business apps, from Gmail and Google Sheets to Stripe, QuickBooks, HubSpot and Odoo.

Chatbot, auto-reply or agent?

The three get confused constantly. They do different jobs.

The WhatsApp Business auto-reply

The free WhatsApp Business app lets you set a greeting message, an away message and a handful of quick replies. Useful, and very limited. It sends the same text to everybody, whatever they asked, and it cannot look anything up. A customer asking whether you have a size 10 in stock gets "Thanks for your message, we'll be in touch."

The chatbot

A chatbot follows a script that somebody built in advance. Press 1 for bookings, press 2 for prices. It handles the questions its designer predicted and stalls on the rest. It also only works one way: it talks to customers, never to you, and it cannot go and do anything in your other tools.

The AI agent

An agent reads what the customer actually wrote, in the language they wrote it in, and answers it. It can check your calendar, your sales records or your stock sheet before replying. Crucially, it also takes instructions from you, the owner, and carries them out across your channels and apps.

If all you need is an away message on a Sunday, the auto-reply is fine. If customers are asking real questions and you are the only person who can answer them, you need something that can think.

What it does on an ordinary day

The work of a small or medium business does not come in neat categories. It arrives as a stream. Here is how the main parts of that stream are handled.

Customer replies across every channel

Questions about prices, opening hours, delivery, availability and bookings land in WhatsApp, Instagram DMs, Facebook comments and TikTok all day. Flann reads each one, drafts a reply in your tone, and puts it in front of you. A café, an online shop, a physiotherapy clinic and a wholesale distributor all get the same thing: answers that sound like the business, not like a help desk.

Follow-ups with customers who have gone quiet

Every business has people who used to come in and stopped. Flann finds them and drafts a personal message to bring them back. If you use the MSTRMND loyalty programme, Flann can read the visit history directly, so it knows who your regulars were and when they last came. There is more on this in our guide to customer retention software.

Reviews

Flann drafts replies to your Google Business reviews, good and bad. The kind ones get a warm thank you. The difficult ones get a calm, specific response you can sharpen before it goes public.

Posting to your socials

Tell it what is happening this week and it drafts the post for Instagram or Facebook, ready for your approval. It can work with Canva for the visual side.

Working inside your apps

This is where an agent separates itself from everything else. Ask it to check tomorrow's diary in Google Calendar, chase an unpaid invoice in QuickBooks, look up a payment in Stripe or add a new lead to HubSpot. The full list lives on the integrations page.

Reports when you ask

Send "How are we doing?" and a report card comes back in the chat. It covers Facebook and Instagram reach, loyalty visits and returning customers, sales from Odoo, and the next action worth taking. No dashboard to open.

Three worked examples

Abstract descriptions only go so far. These are the kind of exchanges owners actually have.

A clinic catching up on enquiries

The owner sends a voice note at seven in the evening: "Can you answer everyone who messaged about appointments today? Offer them the free slots on Thursday."

Flann reads the day's WhatsApp and Instagram messages, checks Thursday in Google Calendar, and drafts a reply to each person with the real available times. One patient wrote in Polish, so that reply is in Polish. Another wrote in Gulf Arabic and gets an answer in the same dialect.

The owner sees an approval card for each reply. She approves most as written, edits one to mention parking, and rejects one because she knows that patient prefers a phone call.

An online shop winning back quiet customers

The owner types: "Who bought from us last spring but hasn't ordered since? Draft them something short about the new range."

Flann looks up past payments in Stripe and the order sheet in Google Sheets, builds the list, and leaves out anyone who has opted out. It drafts a short, personal message for each customer that mentions what they bought before.

The owner reviews the cards, edits the opening line on two of them, and approves the rest. Because some of these customers have not messaged the business recently, Flann paces the approved messages over several days to keep the WhatsApp number in good standing.

A restaurant dealing with a bad review

A one-star review appears on Google overnight. The owner writes: "Draft a reply to that review. Apologise for the wait, don't make excuses, invite them back."

Flann drafts a reply that does exactly that, in the measured tone it has learned from the owner's past approvals. The owner reads it, softens one sentence, and approves. Only then does it appear publicly.

How it works

Getting started is deliberately short. Flann is live within 24 hours.

1. Link your number. Scan a QR code to connect the WhatsApp Business number you already use. Your customers notice nothing change.
2. Connect your channels and apps. Add Instagram, Facebook, TikTok, Google Business and whichever business apps you rely on.
3. Talk to it. Send instructions by text or voice note, as you would to a member of staff.
4. Approve, edit or reject. Every message, post and change arrives as a card. Nothing goes until you say so.
5. Let it learn. Your approvals and corrections teach Flann your tone. Once you trust a particular job, you can let it run on its own, and switch it back whenever you like.

Approval and safety

Handing your customer conversations to software is a reasonable thing to be nervous about. So Flann starts from the cautious end.

By default, every outgoing message, every social post and every change to an app waits for you as an approval card. You approve it, edit it or reject it. The rule is simple: looking something up in an app happens straight away, because reading a calendar harms nobody. Anything that creates, changes, sends or deletes is a card first.

Trust is earned job by job. You might let review thank-yous run on their own after a month of approving them unchanged, while keeping win-back messages under your eye. Each job can be switched back to approval at any time.

The number itself is protected too. Opt-outs are always skipped. New contacts are paced rather than messaged all at once, which keeps the WhatsApp number you have spent years building safe.

And Flann replies to each customer in the language they wrote in. English, Polish, and Arabic dialects included. That matters for a business with a mixed customer base, and it matters for the customer, who should never get a reply they cannot read.

Do you need a WhatsApp AI agent?

Not every business does. A few honest questions will tell you.

• Do customer messages regularly wait hours, or until the next morning, because you are busy serving the people in front of you?
• Are your messages spread across WhatsApp, Instagram, Facebook and reviews, with no single place to deal with them?
• Do you know there are customers who have drifted away, but never find the time to write to them?
• Do you switch between four or five apps to answer a simple question about your own business?
• Would you rather send a voice note than open a laptop?

If you answered yes to two or more, an agent will probably earn its place. If you get a handful of messages a week and enjoy answering every one by hand, keep doing that.

Put side by side, the two ways of handling the day's messages look like this.

The sensible test is to try it on the work you dislike most. Start with review replies or a single follow-up. Watch what it drafts. Approve, edit, reject. You will soon know whether it sounds like you.

Every plan includes every channel and every app, so there is no feature list to compare. See the plans and pick the one that fits the size of your business.