AI assistants for support and sales, connected to your systems

Assistants that answer at any hour with each customer real context. They check orders and history in your systems, resolve what is simple and hand over to a person what needs a decision.

Why this is not a chatbot

A chatbot answers from a script or an FAQ page. When the customer asks where their order is, it does not know, because it has access to nothing. That is why almost everyone quickly learns to type "talk to a human".

An assistant connected to your systems checks the actual order, that customer history and the state of the process before answering. It resolves what it can and, when it cannot, routes to the right person with the case already summarised, instead of making the customer repeat everything.

What makes it useful

Customer context: it knows who is on the other side, what they bought and what was left unresolved last time.
System access: it checks orders, stock, invoicing and process status at the moment it answers.
Routing with a summary: what it cannot resolve reaches the team with history and context, not as a blank request.

The problem: requests do not stop when the office closes

A good share of contacts arrive out of hours or in peaks the team cannot keep up with. Whoever answers first wins the deal; the rest sit in the inbox until someone has time. And the time the team does have goes mostly to answering the same questions as always.

Signs you are at this point

The team answers the same five questions every day, and each takes longer than it should because someone has to go check a system.
Requests arriving at the end of the day or at the weekend only get an answer on Monday.
Promising contacts get mixed in with support requests and spam, and someone has to separate them by hand.
You already tried a chatbot and it annoyed more customers than it helped.

Where it pays off

The cases where the return shows up first are high volume with predictable answers, not edge cases.

Sales

Qualifying incoming contacts

The assistant works out what the person is looking for, cross-checks the CRM history and scores them for fit. Promising contacts reach the sales person with context; the rest still get an answer.

Customer support

Order and process status

Where the order is, when it arrives, what is missing in the process. These are the most repeated questions and the ones the assistant handles best, because the answer sits in a system rather than in an opinion.

After sales

Returns and service requests

Requests opened with the fields already filled from customer history, routed to the right team depending on product type or contract.

Internal

Support for your own sales team

What price we quoted this client, what the contract terms are, what was agreed. The assistant answers during the call, instead of forcing someone to interrupt and go looking.

In the channels and systems you already use

The assistant lives where customers already talk to you, and reads where the information already is. We do not ask you to change platform or migrate history.

WhatsApp Business, website chat, Messenger and email
CRM: HubSpot, Salesforce, Pipedrive or whatever you have custom built
Portuguese business software: Primavera, PHC, Sage, Moloni and InvoiceXpress
E-commerce platforms and in-house order systems
Customer support tools and shared mailboxes

Control over what it says and does

The legitimate concern is the assistant telling a customer something wrong. That is handled with explicit rules, not with hope.

It answers from your data and documents, not from what exists on the internet.
When it does not know, it says so and hands over to a person, instead of inventing.
Actions with real impact, such as accepting a return or applying a discount, only run unattended if you allow it.
Every conversation and the information consulted are logged.
Data hosted in the European Union, not used to train public models.

How it is implemented

Start with one channel and a small set of intents, with people reviewing before the assistant talks to customers on its own.

01
Diagnosis
30 to 45 minute session, free
We look at what questions come in, through which channels and at what volume. You leave with a list of what is worth automating first.
02
Design and rules
2 to 3 weeks
Definition of intents, tone, what requires human approval and the routing criteria. Scope, timeline and investment fixed.
03
Supervised pilot
4 to 6 weeks
The assistant goes into one channel with human review of answers before they go out. It is tuned on real cases until it is ready to answer alone.
04
Rollout
Depends on scope
More channels, more intents and more integrations, based on what the pilot data shows is worth it.

Investment

Reference bands upfront. The exact figure is fixed in the proposal, after the diagnosis, and does not change midway.

One channel

€7,500 to €15,000

One assistant in production, in one channel, connected to a real system. 4 to 6 weeks depending on integrations.

Multichannel

€15,000 to €30,000

Several channels with shared context, CRM and business system integration, and routing per team.

Evolution

On proposal

The first 3 months of support are included. After that, ongoing review of conversations and tuning of the rules.

Frequently asked questions about AI assistants

They do when the machine resolves nothing. What annoys people is not that it is an assistant, it is repeating the problem three times and ending up in the same queue anyway. An assistant that knows which order it is and resolves the request beats waiting two days for an email. And the path to a person is always there, with the case already summarised.

It answers from your data and documents, not the internet, and when it does not know it hands over instead of inventing. For actions with impact, such as accepting returns or granting discounts, we agree with you what requires human approval. During the pilot, answers are reviewed by people before they go out.

Yes, through WhatsApp Business, and also on website chat, Messenger and email. Customer context is shared across channels, so someone who started on WhatsApp and continues by email does not have to repeat everything.

In practice that is not what happens. The assistant absorbs the repetitive volume, which is the part of the job nobody enjoys, and the team keeps the cases that need judgement and relationship. What changes is response time and the number of requests left unhandled.

Typically 4 to 6 weeks once the design is validated, and design takes 2 to 3 weeks. Counting from the first conversation, around two months until it is answering customers, with a human review period in between.

Next step

Want to know what this would do in your operation?

Free diagnosis, no commitment. We look at your process and come back with a plan covering scope, timeline and investment. Reply within 48 business hours, NDA included by default.

100% free diagnosis
No commitment to proceed
Reply within 48 business hours
NDA included by default

Request free diagnosis

Two minutes to tell us the essentials. We reply within 48 business hours with a first read on your case and, if it makes sense, a proposed session.

Explore other solutions

Every vertical has its own challenges. See how we approach other sectors where we have proven experience.

Custom CRM