Six months ago, we set out on an ambitious journey with Franz, our AI-powered document processing startup. Today, I want to share our story - the challenges we’ve faced, the pivots we’ve made, and the exciting innovations we’re working on. It’s been a whirlwind of learning and progress, and I’m thrilled to give you an insider’s view of our experiences.
When we started Franz, we had a clear vision: to significantly improve document processing with AI. As former freelance developers, we saw firsthand the inefficiencies in how businesses handle documents. We knew there was room for substantial improvement.
Initially, we envisioned creating a pure SaaS solution - a one-size-fits-all platform that could handle any document processing need. But as we dove deeper into the problem and started working with our first clients, we realized something crucial: every company has its own “ism,” its unique way of doing things.
This realization led to our first major pivot. Instead of a monolithic SaaS platform, we decided to build a flexible toolbox/SDK model. It wasn’t an easy decision - we had to scrap some of our initial work and rethink our approach. But we knew it was the right call. Our new strategy allows us to:
It’s a more nuanced approach, but we believe it offers far greater value in the long run. Plus, it’s been incredibly rewarding to see how our clients are using our tools in ways we never imagined.
Now, let me geek out for a moment about our technology. We’ve implemented what we call a multi-modal approach to document processing. It’s not just about using multiple technologies for the sake of it - it’s about leveraging the strengths of different tools to create something greater than the sum of its parts.
We’re using OCR, Named Entity Recognition, large language models, and few-shot learning, all working in concert. One of my favorite features is our handwriting correction system. By cross-referencing inputs with the full document and company data, we can catch and correct most misspellings. It’s like having a super-smart, context-aware spellchecker.
Another big decision we made was to lean heavily into open-source models. This was a bit nerve-wracking at first - there’s a lot of hype around proprietary AI models. But we found that when properly implemented, open-source models can perform just as well as closed-source solutions, often at a fraction of the cost.
This approach has paid off big time. We’ve managed to get our PDF processing costs down to about €0.001 per page. But more importantly, it’s given us more control over our tech stack and reduced our dependence on major AI providers.
As exciting as the past six months have been, I’m even more thrilled about what’s coming next. We’re focusing on two major areas:
These aren’t just cool technologies - we believe they’re the key to truly transforming document processing and B2B communication. Imagine a system that can learn and adapt to how your company, your suppliers, and your clients work. That’s the future we’re building towards.
The past six months have been a rollercoaster of challenges, breakthroughs, and constant learning. We’ve laid a solid foundation, but we know we’re just getting started. There’s still so much to do, so many problems to solve.
But every time I see a client streamline their processes using our tools, or when I hear feedback like, “This has cut our processing time in half,” I’m reminded of why we started Franz in the first place.
We’re not just building a product - we’re working towards a new approach in how businesses interact with their documents and with each other. It’s challenging, sure, but we believe it’s a journey worth taking.
Thank you for your interest in Franz. We’ll keep pushing forward, one line of code at a time, to make document processing smarter, faster, and more efficient for businesses across Europe and beyond. Stay tuned for more updates - we believe the best is yet to come!