WhatsApp AI agents

WhatsApp as a service channel, connected to your processes

The agent answers the usual queries with the information and limits we agree on, records the data the next step needs and alerts a person as soon as the case calls for judgement. It is a channel within the automation of your processes, not a stand-alone tool.

The problem

WhatsApp is the most direct channel and also the most demanding to sustain

It is where customers write naturally and at any hour, and where an unanswered message weighs more than in other channels. Without processes behind it, service depends on the availability of individual people.

The same enquiries, repeated every day

Opening hours, scope of services, location, required documentation. Answers that someone on the team writes out again and again, at a cost in working hours.

Messages that arrive outside service hours

Weekends, late evenings or peak workload periods. By the time a reply is sent, the prospect may already have contacted another supplier.

Conversations that leave no trace

What was discussed stays in the chat of whoever handled it, not in a shared record. If someone else takes the case over, they start without context and the customer has to repeat everything.

What we do

What a well-designed agent does

The agent works with information the client has validated, within agreed limits, and is connected to the company's processes. It has clear rules on when a case must be handed over.

Answers common enquiries using the business's real information, hours and criteria.

Responds outside the team's working hours in cases where the answer does not require professional judgement.

Collects the data the next step of the process needs and records it, rather than leaving an isolated conversation.

Classifies each inbound message: enquiry, request or case requiring immediate attention.

Qualifies the opportunity before it reaches the team, so that the first human contact starts from organised information.

Retains the context of what has already been discussed and does not ask again for information the customer has already provided.

Leaves the request documented so that the team only has to validate it, with traceability of what was collected.

Hands the case over to a member of the client's team, with everything collected, when it requires professional judgement.

How it applies

Illustrative scenarios by type of business

Real situations, told as they happen. These are not customer cases or measured results: they are examples of use.

Garage

Hypothetical scenario: many appointment requests arrive via WhatsApp. The agent collects registration number, type of work and approximate availability, and leaves the requests in order for the workshop manager to confirm.

Private clinic

Hypothetical scenario: enquiries about opening hours, preparation and documentation are answered using information validated by the clinic. Any clinical matter is handed over immediately to the appropriate person.

Renovation company

Hypothetical scenario: the same information is always needed before quoting. The agent collects it, together with photos and approximate measurements, and the request arrives complete rather than scattered across several messages.

What we need from you

That you tell us what information it can state with certainty

The information the agent may state with certainty: services, conditions, opening hours and requirements for each procedure. Anything not confirmed is not answered. We also carry out a diagnosis of how WhatsApp is set up today and of the processes it needs to connect to, before recommending any change of number or configuration.

When a person steps in

Frees up the team's working hours without removing human judgement

The agent handles the first, repetitive layer. From the outset we define which conversations must be passed to a person on the client's team (sensitive cases, complaints or any situation outside what was expected), and that handover arrives with all the context recorded. Confirming real availability would require a data source to consult; until one exists, the agent collects the request and someone on the team confirms it.

How many working hours does your team currently spend answering messages manually?

Tell us what enquiries you receive and how you handle them today. In a diagnosis we assess whether the volume and repetition justify an agent and how it would connect to your processes, and we will say so plainly even if the conclusion is that it is not yet worthwhile.

See all ways to get in touch

Talk to Nova on WhatsApp