From a message to a completed action.
A customer sends an ordinary sentence. Adjutant works out what they need, finds the answer in your systems, does the thing, and tells them it’s done. Here is what sits between those two moments.
Talk → Understand → Know → Act
Your customer only ever sees the first one.
Adjutant runs on the WhatsApp Business Platform, on a number you own. Customers message it the way they message anyone else — no app to install, no portal, no password, no training.
If you already advertise a WhatsApp number, the behaviour you need already exists. We’re improving it, not creating it.
Adjutant AI
The customer writes normally. Adjutant works out intent from the sentence and the conversation around it — what they’re asking, which of them they mean, and what still needs to be established before anything can happen.
No menu trees. No keywords. No “reply 1 for sales”. The software learns how to understand the customer rather than the other way round.
Your knowledge
Products, services, prices, policies, opening hours, documentation, procedures and customer records — whatever you decide Adjutant is allowed to use.
This is the difference between a confident guess and a correct answer. Adjutant answers from what you gave it, and says so when it doesn’t know.
Your tools
Knowing something and doing something are different capabilities. Adjutant connects to the systems that run the business — bookings, CRM, inventory, support, accounting, payments, your own APIs — as tools it can use.
You define which tools exist and what each one is permitted to do. Adjutant decides when to use them.
“Can you resend my invoice from March?”
Eight seconds of work, and about eight steps. The customer sees one reply.
Identify who is asking
Match the WhatsApp number to a customer record, so “my invoice” means something specific.
Establish what they mean
Which March, which account, and whether more than one invoice fits. If it’s ambiguous, Adjutant asks rather than guesses.
Check what’s permitted
Documents belong to an account. Adjutant only retrieves what this customer is entitled to receive.
Use the right tool
Query the accounting system for that invoice, and retrieve the document itself.
Answer, and act
Send the PDF into the conversation with a short, plain reply. No links to a portal, no “please hold”.
Record what happened
The conversation, the actions taken and their results are logged, so the whole exchange is reviewable later.
Connect the tools you already use.
You don’t replace anything. Your booking system stays your booking system.
Adjutant reaches your systems through their APIs, and gives each connection a narrow, explicit job: availability.check, booking.update, invoice.fetch. That is what makes its behaviour reviewable — every action is a named capability you agreed to, not an open door into your database.
Where a system has no usable API, we say so early rather than promising an integration that will not hold up.
- Booking & scheduling
- CRM
- Inventory & stock
- Orders & fulfilment
- Accounting & invoicing
- Help desk & ticketing
- Payment providers
- Document storage
- Internal APIs
- Custom workflows
Knowledge can also come from what you already have written down — price lists, policies, product documentation, procedures and FAQs.
What a pilot looks like.
We onboard a small number of businesses at a time. Every pilot starts narrow, on the requests you actually receive most.
Look at the messages
We go through what your customers actually ask, and where your staff go to answer it.
Pick the first jobs
Two or three high-volume requests. Enough to matter, small enough to get right.
Connect and constrain
Wire the systems those jobs need, set what requires approval, and agree the handoff rules.
Run it, then widen
Watch real conversations, correct what it gets wrong, and add capabilities once it has earned them.