01The challenge
HelloPrint is Europe's largest marketplace for personalized print products. It operates in more than 30 countries, works with a network of over 300 manufacturers, and serves more than 800,000 customers. With that volume, their support team handles a very high daily flow of queries — both by email and through web forms.
The challenge was clear: they needed an AI solution capable of automatically answering support queries, but one that went far beyond a generic chatbot. The AI had to access HelloPrint's internal system, know the real status of every order, and in some cases, make changes directly. It also had to work inside HubSpot, where the support team works day to day.
02What I did
01
AI connected to the internal system
I configured the AI to read information from HelloPrint's internal database — order status, customer data, transaction history — and make changes when needed. This wasn't about giving generic answers, but resolving real queries with real data.
02
Forms turned into conversations
HubSpot has an important limitation: tickets generated from web forms don't open a conversation in the sense the platform understands, which stopped the AI from being able to interact with them. To solve this, I built an API integration that redirects every response received through forms and turns it into a conversation inside HubSpot.
03
Automatic response loop
Once the query comes in as a conversation, the AI processes it and generates a response. If the customer writes back, the system detects the new message and sends the conversation to the AI again, creating a continuous response loop with no human intervention.
04
Multichannel coverage
The same system works both for emails the support team receives in HubSpot and for queries that come through web forms, unifying both channels into a single automated flow.
03Results
The implementation cut the daily volume of queries the support team had to handle manually by almost 30%. Those queries are now resolved automatically, with answers based on real data from the system — not generic templates. This freed up the team's capacity to focus on the more complex cases that genuinely need human attention.