A day in support
Meet Sofia.
Sofia leads customer support. Sales have doubled in two years; the support team is still three people.
-
140 messages arrived overnight through the chat, email and social networks.
-
Half ask where an order is. Each answer means opening the shop back office and the carrier's website.
-
Questions about sizes, materials and compatibility wait hours — and some customers give up on the purchase.
-
A complaint about a damaged item is buried in the queue. The customer has already written twice.
The challenge
The answers exist. Getting to them takes too long.
Order status, stock and policies are all in the shop's systems. But the customer asks in the chat, and someone has to go and find the answer.
55%
of messages ask for the status of an order
6 h
average wait for a first reply
40%
of messages arrive outside working hours
Estimate Illustrative figures based on typical scenarios. Every assessment establishes your organisation's actual numbers.
-
Customers waiting
Simple questions sit in the queue, and each delay means another message chasing the first.
-
Lost sales
Product questions answered the next day often arrive after the customer has bought elsewhere.
-
Complaints buried
The cases that really need a person are mixed in with everything else.
What changes
A chatbot that knows your shop.
With MCP servers connected to the shop, the carrier and the helpdesk, the chatbot answers with real data and only acts with the customer's confirmation. Refunds and sensitive cases go to the team, with the whole story already summarised.
Today
- Order status looked up by hand, message by message
- Product questions answered the next working day
- Support only during working hours
- Complaints found late in the queue
With a chatbot connected through MCP
- Tracking and delivery estimate answered in seconds
- Answers from the catalogue and stock, with alternatives
- Available 24/7, in Portuguese and English
- Sensitive cases handed to a person with a summary
Guided demonstrations
Three conversations with the shop's chatbot.
Choose a demonstration. Actions with effect wait for the customer's confirmation, and refunds are always decided by a person.
Order tracking — Online shop (simulation)
Processed in the EUIt is 23:40. Rita opens the shop's chat because her order has not arrived.
MCP calls
0- Online shop
- Carrier
Simulation with fictitious data. No real system is contacted.
Product question — Online shop (simulation)
Processed in the EUTiago is about to buy a bedding set, but is not sure it fits his bed.
MCP calls
0- Online shop
- Policies and FAQ
Simulation with fictitious data. No real system is contacted.
Complaint hand-off — Online shop (simulation)
Processed in the EUMarta received a broken vase and wants her money back.
MCP calls
0- Online shop
- Helpdesk and CRM
Simulation with fictitious data. No real system is contacted.
Connected systems
The MCP servers behind this solution.
Connections to the shop platform and the services you already use, through their APIs. The chatbot only sees the data of the customer it is talking to.
-
Online shop
Orders, catalogue, stock and basket
orders.getReadproducts.searchReadstock.checkReadbasket.addWrite with approval
-
Carrier
Tracking, delivery and pick-up points
shipments.trackReaddeliveries.rescheduleWrite with approval
-
Helpdesk and CRM
Customers, tickets and contact history
customers.historyReadtickets.createWrite with approvalconversations.transferWrite with approval
-
Policies and FAQ
Returns, warranties, delivery times and size guides
policies.searchRead
Data processed in the European Union, through our partners. Your data is not used to train models.
Results
Faster answers. A team free for what matters.
Estimates for an online shop receiving around 3,000 support messages a month.
-
First reply to a customer
Before: ≈ 6 h After: ≈ 30 s
-
Conversations handled by the team
Before: 100% After: ≈ 40%
-
Time to pick up a complaint
Before: ≈ 1 day After: ≈ 1 h
Estimate Illustrative figures based on typical scenarios. Every assessment establishes your organisation's actual numbers.
Safeguards
Automation that knows when to step aside.
-
Clearly an AI
The chatbot always introduces itself as an AI assistant and the customer can ask for a person at any time.
-
Verified customers
Order details are only shown after checking the order number and the customer's email or code.
-
People decide
Changes need the customer's confirmation; refunds, compensation and exceptions are always decided by the team.
-
Data in the EU
Processing in the European Union; your data does not train third-party models.
Frequently asked questions
Will it work with our shop platform?
We connect to the most common online shop platforms and carriers through their APIs. The assessment confirms the best route for you.
What happens when the chatbot does not know the answer?
It says so, does not make up an answer and offers to pass the conversation to the team, with a summary of the case. Outside working hours it opens a ticket and tells the customer when to expect a reply.
How long does the first implementation take?
A proof of concept with order tracking and the most common questions typically takes 3 to 6 weeks.
Which questions take up most of your support team's day?
Tell us what they are and which shop platform you use. In the founders programme, the assessment is free and the proof of concept is only paid for at the end, against the agreed results.