Building AI Innovation Labs together with Companies
Abstract: In the future, most companies will be confronted with the topic of AI and will have to decide on their strategy in this regards. Currently, a lot of companies are thinking about whether and how AI and the usage of data will impact their business model and what potential use cases could look like. One of the biggest challenges lies in coming up with innovative solution ideas with a clear business value. This requires business competencies on the one hand and technical competencies in AI and data analytics on the other hand. In this article, we present the concept of AI innovation labs and demonstrate a comprehensive framework, from coming up with the right ideas to incrementally implementing and evaluating them regarding their business value and their feasibility based on a company's capabilities. The concept is the result of nine years of working on data-driven innovations with companies from various domains. Furthermore, we share some lessons learned from its practical applications. Even though a lot of technical publications can be found in the literature regarding the development of AI models and many consultancy companies provide corresponding services for building AI innovations, we found very few publications sharing details about what an end-to-end framework could look like.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Explain it Like I'm 14
Overview: What this paper is about
This paper explains a practical way for companies to create useful AI ideas and turn them into real products and services. The authors call this approach an “AI Innovation Lab.” It’s not just a room with computers—it’s a step-by-step framework that helps a company pick the right ideas, test them safely, and launch the ones that truly help the business and its customers.
The main questions the paper tries to answer
- How can a company use AI to make real, measurable improvements (like saving money, making better products, or understanding customers)?
- Which AI ideas are worth doing, and do we have the right data to make them work?
- What skills, tools, and costs are needed?
- How do we move from a small test to a full product that customers actually use?
How the approach works (in simple terms)
Think of building an AI idea like making a new gadget for a science fair. You don’t build the full thing on day one. You start with an idea, check what you need, make small prototypes, test them, fix what’s wrong, and only then show it to everyone. The framework the authors describe follows that same path.
Two ways to find good AI ideas
- Business-first: Start with a clear business need (for example, “we want fewer machine breakdowns”). Then ask what data and AI tools are needed to solve it. This keeps effort focused on real value.
- Data-first: Look at the data you already have and see what interesting patterns could help the business. This can spark surprising ideas you didn’t plan for.
In practice, the best results come from mixing both approaches: business experts and data experts brainstorm together, then compare options to pick the strongest idea.
The step-by-step innovation cycle
The process is iterative, meaning you repeat small cycles of build–test–learn. Each cycle helps you decide whether to continue, change direction, or stop. The goal is to “fail early” on weak ideas and invest more only in the promising ones.
Here is the typical path ideas follow:
- Potential Analysis: Define the idea and why it matters for the business. No coding yet—just thinking and planning.
- Lab Small: Try a simple prototype with sample data.
- Lab Full: Test with more realistic, historical data to see if it still works.
- Field Study: Pilot it in the real world with real users and real data.
- Roll-Out: Launch it for everyone and set up maintenance to keep it working well.
At each step, the business case (Why do this? What value does it bring?) and the technical plan (How will we build it?) are checked together.
The tools and templates they use (explained simply)
- Value Proposition Canvas: A structured way to list what customers need and how your idea helps them, so you build something people actually want.
- Business Model Canvas: A one-page map of how your business works—customers, costs, partners, and how you earn money—so your AI idea fits the bigger picture.
- CRISP-DM: A common, six-step guide for data projects: 1) Understand the business goal 2) Understand the data 3) Prepare/clean the data 4) Build models (try different algorithms) 5) Evaluate results 6) Deploy and maintain the solution
This keeps the technical work organized and connected to real needs.
What they found and why it matters
The authors combine well-known methods (like Lean Startup, Business Model Canvas, and CRISP-DM) into one complete framework for AI projects. Over nine years working with many companies, they learned what helps AI projects succeed and avoid common failure traps.
Based on their experience, these points stood out:
- Having the right mix of people is crucial. Business leaders, technical engineers, and AI/data scientists need to work together from the start.
- Data is often messy or incomplete. Expect to spend real time cleaning it or collecting better data.
- AI models live inside bigger systems. Data scientists should team up closely with software and systems engineers to make things work in the real world.
- Fail early, learn fast. Small, quick tests save money and time compared to big “all-or-nothing” projects.
- Plan for ongoing care (often called DevOps or MLOps). AI models can “drift” over time and need updates as the real world changes.
Why this research could have a big impact
If companies follow this framework, they can:
- Pick AI ideas that actually help their business instead of chasing hype.
- Reduce risk by testing ideas in small steps.
- Use clear criteria to decide which ideas to invest in or drop.
- Launch AI solutions that are useful, reliable, and maintainable.
In the future, the authors want to spread this approach more widely, create common “patterns” for typical AI solutions (like checklists or starter templates), and train company teams—especially in small and medium businesses—to run their own AI Innovation Labs. This could help more companies build smart, trustworthy AI solutions that benefit both the business and its customers.
Collections
Sign up for free to add this paper to one or more collections.