The same question, twenty times a day
Opening hours, address, deadlines and terms take up the time of someone who should be preparing quotes.
Customer service and chatbots: the right answer, any time of day
Assistants that answer the questions you hear every day, drawn from your company's own content, and hand the conversation to a person when they don't know the answer.
Who automated customer service is for
Most of the questions that reach a company are the same ones: opening hours, address, deadlines, list prices, the status of an order, whether a given service includes a given thing. Answering them fills the day of someone who should be handling the requests that make money — and out of hours, nobody answers at all.
A customer service assistant handles that layer. It isn't a menu of buttons or a decision tree: it answers in plain language, drawing on the content the company has given it, and when a question goes beyond what it knows, it hands the conversation to a person with the history already written up.
This service is for companies with a volume of repeat questions and written information that answers them. If the problem is instead how requests get routed internally once they arrive, see workflows and processes; if it's the commercial management of contacts, see CRM and integrations.
Problems this service solves
Requests for automated customer service usually start with one of these symptoms.
Out-of-hours requests go unanswered
Anyone who writes in the evening or at the weekend only gets an answer the next working day, and by then they've asked someone else.
Different answers depending on who picks up
Each person on the team explains the same thing their own way, and some versions are no longer correct.
Information that exists but nobody finds
The answer is on the website, in the catalogue or in an internal document, and someone still has to be asked.
Conversations scattered across channels
Website, WhatsApp, email and social media, each with its own inbox and its own way of getting forgotten.
An assistant that makes up an answer costs more than one that says it doesn't know.
How we work
An assistant that makes things up does more damage than no customer service at all. So the work starts with the information, and only ends with the bot.
The real questions
We read through months of conversation history, emails and forms, and group what's actually being asked. It's this list, not an imagined one, that sets the assistant's scope.
The knowledge base
We gather the content that answers those questions — website pages, tables, internal documents — and correct anything out of date. The assistant can only answer from this.
Tone, limits and handover to a person
We define how it speaks, what it never answers (prices outside the price list, deadlines still to be confirmed, legal matters), and the point at which it hands the conversation to a person, with the history and contact details already collected.
Testing with real questions
We test with the real conversations from phase 1, including the badly written ones and the ones off topic, and fine-tune until wrong answers disappear from the sample.
Going live and review
We publish one channel at a time. In the first few weeks we read through the conversations with your team: whatever went wrong becomes a new rule or new content, not an excuse.
What you get
Beyond the assistant itself, you get what you need to keep it running without depending on us.
The assistant on the agreed channels
On your website and, where it makes sense, on WhatsApp, email or social media messages.
An editable knowledge base
The content behind the answers, kept somewhere your team can correct without coding.
Handover to a person, with context
Rules for when to hand over the conversation, to whom, and with what data already collected.
A history of conversations
Everything that was asked and answered, to review, correct and understand what customers want to know.
A usage report
How many conversations there were, how many were resolved without a person at all, and which questions keep going unanswered.
Legal notice and record-keeping
The notice that customer service is automated, and the handling of conversation data, as the law requires.
Common questions about automated customer service
The questions that reach us most often before putting an assistant in front of customers.
Can the assistant make up an answer?
It can, if it's built to answer everything. Ours is built the other way round: it answers from the content it's been given, and when a question goes beyond that, it says it doesn't know and calls in a person. It's less impressive in a demo and far safer with real customers.
Will my customers know they're talking to a machine?
They will, and that's how it should be. The assistant identifies itself in its very first message and says how to reach a person. Hiding this isn't just against European transparency rules in practice — it also sours the conversation the moment a customer suspects.
Does it work on WhatsApp?
It does, with a WhatsApp Business account and the approval Meta requires for automated messages. The same assistant can run on your website and on WhatsApp from the same knowledge base, so there's never two versions of the truth.
What languages does it answer in?
It answers in whichever languages we set, and it usually replies in the language the question was asked in. What limits it isn't the language, it's the knowledge base: if the content only exists in Portuguese, answers in another language are translations of that same content.
How much does it cost to run each month?
There's a usage cost that depends on the volume of conversations and the provider chosen, and a cost for keeping the content up to date. In the quote we give an estimate based on your current volume of requests, worked out for double that, so there are no surprises if things go well.
What if I want to switch it off?
You switch it off, and customer service goes back to how it was. The knowledge base stays yours and stays useful, if only as an internal manual for the team.
Show us the questions you get asked most
Give us a month's worth of conversation history and we can tell you what percentage of it can be automated, and what that's worth in your team's time.
Other services in Artificial Intelligence