The challenge
Goboony and the merged company Yescapa handle around 450,000 customer contacts a year, across ten countries. At some point a ceiling in scaling the team was hit, while the workload kept growing. The inboxes filled up and proactive follow-up was the first thing to slip.
The work is the same in every country, yet each country had its own team. Some teams had deep, above-average knowledge; others, especially in the smaller or newer countries, had gaps. The job simply demands an enormous amount of knowledge, and it takes a new team member several months to learn it.
On top of that, the work was spread across five tools: Zendesk, Front, Respond.io, HubSpot and Copper. Every case meant switching between systems.
What we built together
Customer support and quality
The AI answers support emails and prepares the tasks behind them: updating payment status, processing cancellations and changing bookings. It checks every cancellation against the terms of that market, chases missing payments and follows up owners who haven’t accepted a booking yet. The team works in its own language cross-market; the customer gets the reply in theirs.
Claims handling
Claims come in with photos, free text and PDFs from several parties, and insurers’ rules differ per country. The AI applies the policy terms for that market and insurer, compares the damage photos and documents, proposes a judgment and drafts the messages to everyone involved. If a photo, statement or document is missing, the AI requests it from the parties itself, and only passes the claim on to a team member once the file is complete. Every claim now arrives fully prepared; the team reviews and approves.
Owner onboarding and listing moderation
Every new listing, the ad an owner places to rent out their camper, used to be checked by hand before it could go live. Now the AI runs more than 21 checks per listing, flags photos with logos and spots attempts to take bookings off the platform. It tells the owner exactly what is missing, and how to make their listing stand out.
A system that keeps getting smarter
None of this was switched on in one go. Every correction the team makes feeds the knowledge base, so answers get sharper week by week. And with every new connection to Goboony’s own platform, the AI can do more on its own: look up a booking, update a payment status, publish a listing. What started as one assistant has grown into the layer most of the operation runs through.
How it works
A new camper listing (the owner’s ad), from submission to live:
-
Trigger
E.g. An owner submits a new listing
-
Analyse
Photos and description go through 21+ checks
-
Assess
Logos in photos or attempts to take bookings off-platform are flagged
-
Follow up
The owner hears exactly what is missing
-
Prepare tasks
Approved listings go live, after a quick check from the human team.
The turning point: flipping the process
One of the first things we built was a strong assistant that analysed disputes between renters and owners, including back-office data, liability and cover. Technically it worked well. But the support member had to press the button, and mostly they didn’t. Doubt about the output, and no habit.
So we turned it around. Incoming email now goes through the AI first, and one process after another followed. The team gives feedback and adjusts the output before it is sent. The AI does the work; people review and train it. Quality became consistent, and the knowledge base started building itself.
What changed
Processes that had stalled for lack of time, like following up owners on bookings, are running again, and the acceptance rate rose 14% in five weeks. The team that used to manage one process now runs around five at once.
