WhatsApp is where the work already happens.

An AI WhatsApp assistant for business can run your inbox the way a steady office manager would, minus the coffee breaks. It reads every incoming thread, works out what each customer or supplier actually wants, drafts a reply in your voice, and then waits. Nothing goes out until you approve it. When a message hides a task inside it, the assistant carries that task into the tools you already use. That is the job Clark does inside WhatsApp. The rest of this piece explains how it works, what stays firmly under your thumb, and where the honest limits sit.

For many businesses, WhatsApp is the back office

For a lot of small businesses, WhatsApp is where the actual work happens. Orders land there. Questions arrive there. Suppliers send updates there. Bookings, quotes, complaints, and the blessed "payment sent" screenshot all pass through the same app the owner checks in the gaps between doing everything else. In many markets, WhatsApp is a normal way for customers to reach a business, which means one green inbox is quietly doing the work of a CRM, a calendar, and a front desk, without the salary of any of them.

That convenience comes with a bill attached. Threads stack up through the day. A booking mentioned at nine in the morning is buried by noon and forgotten by two. A supplier confirms a delivery and the note never reaches the right order. A quote gets promised, then vanishes under forty newer messages about something else entirely. The information is all there. It is just filed under chaos, in a format no human can keep up with by hand, especially when the human reading the messages is also the human doing the work.

An assistant that lives inside WhatsApp and never loses the thread earns its keep in that setting. That is the gap Clark was built for.

What does an AI WhatsApp assistant for business actually do?

An AI WhatsApp assistant for business reads your threads, decides what each one needs, drafts a response for you to approve, and then does the task the message implies. Clark does these four things in order, and each one deserves a proper look.

It reads every thread

Clark reads every conversation in the connected WhatsApp account. The ones you have opened, and the ones still sitting there unread and faintly accusing. It follows the history of each thread, so a message that only makes sense in light of last Tuesday's exchange gets read in that light. Nothing slips through because the day got loud or the phone got left face-down on a counter.

It understands what each thread needs

Reading is the easy part. The trick is working out what each thread is actually asking for. A price question needs an answer. A "can you fit me in Thursday" needs a booking. A photo of a purchase order needs an invoice. A supplier writing "shipment delayed to Friday" needs that fact pinned to the right job. And a customer who is clearly, thoroughly annoyed needs a real person, so Clark flags that one for you rather than papering over it with a cheerful stock reply.

It drafts a reply in your voice

Clark writes the reply for you, in the tone you already use with your customers. It takes its cues from how you have answered similar messages before, so your quotes still sound like your quotes and your thank-you still sounds like yours, rather than like a form letter that wandered in from a call centre. Then the draft sits and waits. You read it, tweak a word if the mood takes you, and send. Or you tell Clark to have another go.

It turns messages into actions

This is the part a plain chatbot cheerfully skips. When a thread contains a task, Clark carries that task into the other tools you run:

• A booking becomes an event on your calendar, with the name, time, and details already filled in.
• An order becomes an invoice, ready for you to check and send.
• A supplier update becomes a note on the right deal, so the status is current without anyone retyping a thing.

The message and the action stay linked, so you can always trace why something exists. No mystery invoices. No calendar events that appear out of nowhere like uninvited guests.

How is this different from a chatbot?

A chatbot answers messages. An operator answers messages and then does the work those messages create. That gap is the whole story. Clark is an operator. It runs on its own hosted machine, keeps working through the day whether or not you are watching, and treats each message as the start of a small job rather than a line to reply to and forget. The reply is only the first half. The calendar event, the invoice, and the updated deal are the half that usually gets dropped, and the half Clark picks up.

Who stays in control?

You do, at every step. Clark is built to draft and prepare, and to stop and wait for a human on anything that leaves a mark. That principle runs through the entire product, top to bottom.

Nothing is sent without your approval

Replies stay drafts until you approve them. Clark does not message a customer on a whim of its own. You see what it wrote, and you decide whether it goes out the door. Over time you will trust some kinds of replies more than others, and you can loosen the reins at your own pace as that trust builds. No pace is the wrong pace.

High-stakes moves get an explicit ask

Some actions carry more weight than a friendly message. Issuing an invoice. Changing a booking. Anything with money or a firm promise attached. Clark treats those as high-stakes and asks you first, in plain language, before it moves. There is no quiet automation shuffling money around in the background while you sleep.

Every action is logged

Clark keeps a record of what it did and when. Every draft, every action, every request for approval, all of it written down. If you ever want to know why an invoice exists or where a calendar event came from, the log answers without a shrug. That paper trail is what turns "an assistant that acts on your behalf" from a leap of faith into something you can actually check. The FAQ goes into how approvals and logging work in more detail.

Do your conversations stay yours?

Yes. Clark works inside your business's WhatsApp to do your work, and the content of those threads belongs to your business. Clark reads what it needs in order to draft and act, and the entire point of the logs is that you can see exactly what it touched and when. Clark works through WhatsApp as a tool it operates on your behalf. It is not a WhatsApp product and not an official WhatsApp partner. It simply sits alongside the app you already run and helps you keep up with the flood.

What does this look like day to day?

Day to day, it looks like opening WhatsApp and finding the reading and the busywork already handled. A few plain examples show the shape of it.

Take a Tuesday morning at a hair salon. Overnight, six people messaged about appointments, because people book their haircuts at eleven at night for reasons known only to them. By the time the owner unlocks the phone with one hand and a coffee in the other, Clark has drafted replies to all six, turned three firm bookings into calendar events, and flagged a message from a supplier about a colour line that is out of stock. The owner reads down the list, sends five drafts as written, edits one, and is ready before the first client is out of their coat.

Or picture a wholesaler who takes orders by message all day. A regular sends through a list of items with the punctuation of someone typing next to a forklift. Clark drafts a confirmation in the owner's usual tone and prepares the invoice with the quantities already filled in. The owner checks the numbers, approves, done. The customer gets a fast, correct reply, and nobody had to retype a single line into a separate system that hates them.

Or a contractor who lives between sites and rarely near a desk. A supplier writes that the materials are delayed to Friday. Clark records that against the right job, so when the contractor opens that project later in the week, the delay is already noted, in writing, where it belongs. No sticky note lost in the footwell of the van. No detail that quietly dies somewhere between the site and the office.

If you want a fuller sense of the rhythm, we wrote up a week with your Clark as a longer walk-through.

What are the honest limits?

Clark has real limits, and pretending otherwise would be a poor start to a working relationship. It is an operator that reads, drafts, and prepares. It does not replace your judgment on the messages that matter.

• It can misread a thread. Conversations get ambiguous, and Clark can read the room wrong. The approval step exists for exactly this reason. You are the check on its work, which is rather the point.
• It does not send on its own. If what you want is fully automatic replies with no human anywhere in the loop, Clark is the wrong tool. Drafts wait for you, every time.
• It is not a lawyer or an accountant. Clark prepares an invoice or a reply. It does not give legal or financial advice, and it will not sign off on anything for you.
• It works through WhatsApp. Clark helps you run your WhatsApp inbox. There is no official WhatsApp partnership involved, and Clark does not bend WhatsApp's own rules or limits.
• It is early. Clark is in beta. It is being used and sharpened week by week, and some corners will get smoother over the coming months.

None of these should stop you if the daily message load is the thing quietly eating your week. They are the reason Clark keeps you in the loop, approving the replies and the actions, while it does the reading and the drafting underneath.

Where to start

The simplest way to picture Clark: your own WhatsApp, with the reading, the sorting, and the first drafts already done, every action logged, and nothing sent until you approve it. That is the whole idea in one breath.

Clark is in beta now at 250 EUR per month. From the 20th of August it moves to 488 EUR per month at launch. You can see the current terms on the pricing page.

Clark is made by Weblyfe. It is one operator among the tools a small business already runs, built to work quietly inside the app where the work already lands.