How we used AI experiments to revolutionise an entire feature – and what you can learn from it
- in-manas
- Mar 26
- 4 min read
Updated: 6 days ago
Let’s be honest: when you hear “AI”, do you immediately think of innovation and new opportunities – or do question marks and uncertainty come to mind instead? If you fall into the second group, welcome to the club! That’s exactly how we felt when we were developing our Innoboard chat not long ago.
AI is undeniably exciting. But the journey towards building successful, meaningful applications is often unclear, uncertain, and full of challenges. In this article, we’ll share how we turned initial doubts into a real AI success story – and what you can take away from our experience for your own business.
Ready? Let’s dive in.
Our starting point: the AI challenge
The challenges we encountered at in-manas are ones many companies face when exploring the use of AI. It opens up fascinating new possibilities – not just in large-scale projects like our AI research tool or our latest notebook feature (stay tuned for more on that), but also in smaller, more innovative use cases.
And yet, the same questions keep cropping up:How do we integrate AI into existing workflows? How do we identify the use cases where AI can create real value? And how do we unlock AI’s potential without wasting time or resources?
Sound familiar? You might be keen to introduce AI into your business – or perhaps it’s already expected of you. But where do you begin? What tools are right? What solution makes sense? These were exactly the questions we started with.
Why AI projects need to be exploratory
One thing has become increasingly clear to us: AI is fundamentally changing product development. The pace is rapid, with new models, architectures, and approaches emerging constantly. But that also means there’s no one-size-fits-all solution – no universal best practice.
So, how do you navigate a landscape without a clear map?
Simple: you experiment. You try things out. You make space for mistakes. You move quickly and stay bold. That’s exactly the approach we chose.
We intentionally created space for exploratory development, giving our developers the freedom to test new technologies. There were just two ground rules:
All tests had to run within our INNO-VERSE platform.
The experiments had to be based on real use cases with tangible benefits.
Agentic workflows – an exciting new playground
Lately, our focus has been on so-called agentic workflows. If you’re thinking, “Agents – sounds cool, but what does that actually mean?”, here’s a quick explanation:
In simple terms, agent systems are AI-powered tools that can perform specific tasks, answer complex questions, or even take action autonomously. The technology is still quite young – but it holds a lot of promise. Our goal was to empower our customers to use custom AI agents tailored to their specific needs.
But how do you go about building such agents? What real-world problems can they solve?
That’s where Christian enters the picture.
Christian and the 'Innoboard Chat' experiment
Christian is a full-stack developer at in-manas and deeply involved in AI and automation. We gave him a clear mission, along with the necessary time and a dedicated AI environment to support it: experiment with different agent technologies and figure out how we might meaningfully integrate them.
His objectives were clear:
Try out various technologies.
Build internal know-how.
Understand how an AI agent could best handle complex queries.
Identify use cases that would bring immediate value to our users.
Christian chose to turn the existing Innoboard chat into an intelligent AI agent. But why Innoboards?
As Christian puts it:
“Innoboards are great for gathering ideas. But once you’ve collected hundreds, it becomes really tedious and time-consuming to sort through them, evaluate them meaningfully, and identify connections. That’s exactly where AI can shine.”
Specifically:
AI can analyse large volumes of data and recognise patterns or correlations.
An ideal agent would also be able to search the web to answer questions like:
“Are there similar products already on the market?”
“What’s the market potential for this idea in Europe?”
“What technologies would we need to implement this?”
Christian adds:
“Innoboards were also ideal for testing very practical technical questions – like how well can an AI agent really handle a large number of ideas? And how exactly does it need to be structured so it can understand complex queries, respond in the right context, and even work through them step-by-step if needed?”
The potential was huge – but the big question remained: would it actually work?
The outcome: courage pays off
The early feedback from our users was overwhelmingly positive. The risk of trying something new and moving fast had paid off.
In a short space of time, we created a feature that delivered immediate value to our users.
At the same time, we built a strong foundation for improving and expanding the feature going forward.
What can you take away from this?
Our journey into AI taught us four key lessons that we believe can help any business:
Make space for experimentation: The only way to find out what works is to try things.
Focus on real use cases: AI should solve practical, concrete problems – it’s not a goal in itself.
Embrace rapid prototyping: Release quickly, gather feedback, iterate. Don’t wait months for perfection.
Stay up to date: AI is evolving rapidly. Keep exploring, stay curious, or bring in the right expertise when needed.
Now it's over to you!
Are you ready to discover and optimise AI opportunities in your own business?
Our AI team is here to help you identify meaningful use cases and integrate the latest technologies into your processes. You can test and use cutting-edge tools directly via our INNO-VERSE platform.
Curious? Let’s explore what’s possible – together.
留言