Customer support
EnergyTech
High Accuracy
Zonneplan
Zonneplan is a Netherlands-based energy company that helps households generate, store, and optimize renewable energy through solar panels, smart home batteries, and intelligent energy management.
Industry:
Energy Technology
Use Case:
Customer support
Employees
201-500
Location
The Netherlands
Company Overview
Zonneplan is a Dutch energy tech company focused on helping people produce, store, and use their own energy. With over 215,000 customers, it is likely that if you see a solar panel on a Dutch roof, Zonneplan installed it.
The company experienced recurring workload peaks in its customer service operations, particularly during predictable periods, such as monthly invoice cycles. Each billing period triggered a surge in repetitive customer questions, overwhelming their support teams.
To maintain response quality and service levels, they had to scale staffing rapidly. Yet, hiring and training new support staff quickly proved impossible. As a result, operational capacity became a bottleneck.
The consequence was clear: either slow down business growth to keep up with customer expectations, or find a more innovative way to handle recurring demand. For a fast-growing company, slowing down was simply not an option. That is when Zonneplan decided to build Kiki, their AI-powered chat support.
Zonneplan learned that successful AI adoption starts with teamwork, not technology. Their mantra became clear: “Don’t let your AI team work in isolation.”
When developing their chatbot Kiki, they began with a small, multidisciplinary team: Glenn, who deeply understood the customer, a senior developer, and customer service specialists who brought domain knowledge from day one. This ensured that those using the AI also helped shape it.
Rather than over-engineering, they built a quick proof of concept and let domain experts evaluate the answers. Their feedback on what was clear, on-brand, or confusing guided each iteration. AI Engineers fine-tuned prompts, models, and retrieval logic accordingly.
As usage grew to over 200,000 conversations per month, manual review became impossible. To maintain quality, the team implemented systematic evaluation using Orq.ai, combining offline and live testing against a “golden dataset” curated by experts. They also introduced an LLM-as-a-judge system to automatically assess responses at scale.
By combining domain expertise, experimentation, and structured evaluation, Zonneplan transformed AI from a side project into a scalable capability.
Timo Verbeek
Applied GenAI Engineer
Before Orq.ai, our team relied on Excel sheets and custom scripts, a lot of manual work that slowed us down. Now we can ship new features much faster, especially for complex use cases like voicebots. The real value is in the speed and ease of testing; it multiplies our output and gives us room to be creative.
Kiki quickly became an essential part of Zonneplan’s customer service operation.
By combining AI with real domain expertise, the team handled recurring customer questions instantly and consistently without increasing headcount.
Response times dropped dramatically, while customer satisfaction improved thanks to more accurate and on-brand answers. Support staff could now focus on complex cases instead of repetitive requests.
With Orq.ai, Zonneplan gained full control over quality and reliability. They could test new model configurations safely, monitor performance, and continuously improve without disrupting live operations.
What started as a small proof of concept scaled into a production-grade AI assistant that now processes over 200,000 messages per month, setting a new standard for how Zonneplan uses AI to support growth, not limit it.
Orq.ai empowers AI development making it fast, reliable and collaborative, so that teams can innovate without limits.
