.svg)
DURATION
.webp)
Kober, Romania’s leading manufacturer of paints and varnishes, needed a solution that could simplify customer navigation across a broad, technically complex product range. From decorative paints to industrial coatings, each product required clear usage guidance based on surfaces, conditions, and intended applications.
To solve this, Linnify developed a conversational intelligent agent capable of replicating the interaction with a real consultant. The AI-powered assistant provides personalized product recommendations, application instructions, and document lookups (technical sheets, compliance files), making technical data accessible in real time.
Operators and admins have access to a centralized dashboard that monitors and analyzes all user interactions. Through conversation labeling and structured analysis, we can continuously improve the agent over time, leading to smarter responses and an even more delightful customer experience for its customers.
The paint and coatings industry faces multiple challenges in customer experience:
The conversational agent was developed for the Romanian market, supporting both end consumers and business customers who rely on Kober’s wide product portfolio. The assistant helps non-technical users quickly identify suitable decorative products while guiding professionals through detailed technical specifications.
Because Romania’s market mixes professional contractors with individual DIY customers, the conversational AI had to be intuitive and user-friendly, while still providing advanced technical depth for expert users. Aligning the assistant’s tone of voice with Kober’s brand identity was also critical to ensure trust and familiarity.
.webp)
Kober faced a recurring challenge common to companies with extensive and highly technical products. Their diverse customer base, from professional contractors to DIY homeowners, often struggled to understand the differences between similar products and to identify which one best suited their specific needs.
Each product came with dense technical sheets filled with color codes, chemical compositions, and precise usage conditions, making it difficult for non-experts to identify the right product for their specific needs.
Traditional chatbots were not designed for this level of complexity. Most existing solutions could only provide generic answers or redirect users to static catalog pages.
They were unable to interpret nuanced questions, recommend products for unique surface types or environmental conditions, or guide customers through proper application steps.
For operators, this gap created inefficiency and lost opportunities. Customer service teams were burdened with repetitive queries, manual document retrieval, and time-intensive guidance, slowing response times and making it harder to scale personalized support.
.webp)
.webp){kind=small}
Rather than one general-purpose chatbot, Linnify built Gustav as a network of specialized agents behind a single chat interface:
Gustav is a multi-model system, with each agent running on the foundation model best suited to its task. Its most demanding and highest-value capability, the Color Selection Agent (the function most central to Kober's market), runs on Anthropic's Claude Sonnet 4.6 on Google Cloud's Vertex AI.
Gustav rolled out in phases: internal beta (April to May 2025), soft launch (June 2025), and production with real customers from July 2025. A performance review showed the Color Selection Agent needed higher accuracy and responsiveness than its original model could reach, so Linnify migrated it from Gemini 2.0 Flash to Claude Sonnet 4.6 on 1 July 2026, reaching production-grade performance far faster than continued iteration would have allowed.
.webp)
Linnify acted as the end-to-end AI partner, from architecture to delivery, keeping Gustav aligned with Kober's brand voice and business goals. We built specialized agents that interpret nuanced, domain-specific questions and grounded them in Kober's technical documentation and compliance records.
We manage the full AI infrastructure (model orchestration, integration, monitoring, and cost and performance optimization) through Vertex AI.
Evaluation metrics for technical accuracy, recommendation precision, domain restriction, complaint routing, and knowledge-base access each averaged over 97% accuracy.
The result was not just a chatbot, but a scalable, adaptive intelligent assistant capable of transforming how Kober engages with its customers online.
We applied expertise in AI-driven conversational systems, UX/UI design, and product strategy to create a solution tailored to complex sales environments. Our team ensured the assistant could deliver brand-aligned communication, handle nuanced technical queries, and empower operators and admins with actionable analytics.
We also focused on crafting a seamless user journey, where customers received expert-level guidance through natural interactions, while administrators gained actionable analytics to continuously optimize the overall customer experience.
The clearest proof point is the Color Selection Agent, Kober's core differentiator. On its original model it failed on more than half of live customer conversations. Since switching to Claude Sonnet 4.6 on 1 July 2026, it has recorded zero customer-reported errors in production.
Gustav turned a static, complex catalog into an interactive, AI-powered experience. By guiding users to the right products and application steps, it reduces friction, lifts satisfaction, and improves conversion potential, while positioning Kober as a forward-looking leader in Romanian paints and varnishes. The project deepened Linnify's expertise in conversational AI for technically complex industries and showed how multi-agent systems can replicate expert consultation at scale.
This project strengthened Linnify’s expertise in building conversational AI tailored for industries with highly technical and complex product ecosystems. It showed the power of multi-agent systems to replicate expert consultation digitally, while providing high value to customers.
Companies that benefit from this conversational AI solution are the ones selling complex technical products, such as paints, coatings, or industrial materials, where customers need guided product selection and usage support.
This solution differ from traditional chatbots because, unlike generic chatbots, this solution uses multiple specialized agents to handle product recommendations, application guidance, and document retrieval with accuracy.
The system is powered by knowledge sources such as public website data, technical documentation, compliance certifications, and CMS insights.
The experience is customized for the brand through the assistant, which uses brand-specific tone, integrates seamlessly into the company’s ecosystem, and reflects the company’s communication style.
For admins and operators, the features available are advanced conversation analytics, real-time sentiment tracking, and customer experience statistics.
Gustav is a multi-model system running on Google Cloud's Vertex AI, with each agent matched to the model best suited to its task. Its core differentiator, the Color Selection Agent, runs on Anthropic's Claude Sonnet 4.6, with Linnify managing model selection across the system to balance performance and cost.
