At Deep Impact, artificial intelligence isn’t just an add-on – it’s a central part of the strategy, product development, and technological thinking. That also applies to deepico, a Deep Impact venture: the independent company is building a modern cloud ERP platform for the food industry, with the goal of intelligently automating business processes and making them more manageable. CTO Rodrigo Benz explains what AI-First means in practice, how AI-readiness is created, and why experimentation is the key to innovation.
Rodrigo, what does the term “AI-First” mean to you – and how does it fit your current product strategy at deepico?
To me, AI-First means that even at the earliest stage of finding a solution, we consider which tasks a model can take over and what data it would need. Features are designed in a way that intelligence is the core of the solution. This avoids building a chatbot add-on later – instead, it creates functionality whose value is driven by AI.
At deepico, we already rely heavily on digitalization and automation along the entire value chain. AI-First helps us systematically weave future innovations into our cloud ERP products. What’s important for us: AI should solve real problems, not just serve as a buzzword.
How do you ensure your platform is technologically ready for the future use of AI?
We consistently rely on a modern cloud architecture. But more importantly, we’ve built the foundation that allows us to act quickly: whether a model forecasts future customer demand, automatically matches incoming documents to the correct orders, or detects anomalies in production data – we can integrate it into our platform quickly, monitor it live, and roll it out upon success. This ability to experiment quickly and with low risk gives us maximum flexibility for upcoming AI features.
How do you already integrate AI into your product development – even if it’s not yet visible in the product itself?
Today, AI supports us in most phases. During solution design, we discuss ideas with LLMs and arrive at viable concepts more quickly. During prototyping, we generate proofs of concept in a few hours – what used to take days. And when coding, tools like GitHub Copilot significantly increase our speed and test coverage.
What are meaningful criteria in your view for when and where AI adds real value in a digital product?
AI delivers real value where it eliminates tangible bottlenecks in everyday work: for example, when manual data entry and validation are automated, when paper- or Excel-based processes finally flow digitally, or when decisions become more precise thanks to fast analysis. When processes run noticeably more smoothly, the investment practically pays for itself.
What advice would you give tech teams or startups who feel pressure to “do something with AI” – but don’t yet have a clear use case?
Start with small internal experiments – like automated reports or test data generation. That way, you gain experience without involving customers. Keep the experiments cheap – low-code tools are usually sufficient. Most often, good ideas will emerge naturally from these quick wins.
