Skip links
KI Roadmap

Create your own AI roadmap: 5 steps to the first experiment

Everyone is talking about AI, and yet most companies are struggling with it. And no, AI is not just a trend like NFT, Web 3.0 or Pokémon. AI has been around for much longer than Chat GPT. And yet no one has shown us the limitless possibilities as clearly as Chat GPT did in November 2022. And they’re doing it again with “Sora”.

Let’s not kid ourselves. The integration of AI services is new territory for most people. Where is the best place to start? First and foremost, we think it takes courage to take the first step. For example, with an AI readiness check with us. But now to you. We try to bring some order to the chaos with a general guide. A well-thought-out AI roadmap is important in order to optimally utilise the diverse possibilities of AI and at the same time master the challenges. Below we present our perspective and approach to developing a customised AI roadmap for your company.

Step 1: Understand the basics

Before you dive into the world of AI, it’s important to understand the basics. The 5 AI archetypes are:

Machine Learning (ML): algorithms that learn from data to make predictions or decisions.
Audio and Speech: Includes text-to-speech and speech-to-text applications.

Natural Language Processing (NLP): Processing and understanding of human language by machines.
Computer vision: Enables machines to “see” and interpret images and videos.
Generative AI: Generates digital content from different types of data, including text and images.

Step 2: Identify needs and goals

A clearly defined goal is the starting point for any successful journey. By identifying specific business problems or opportunities that can be improved by AI, you lay the foundation for your roadmap. Questions you can ask yourself:

Where are we currently losing a lot of time?
Which simple tasks could be taken over by AI in the medium term?

Where do we enter data multiple times or type it in?
How can we increase customer satisfaction through personalised recommendations?

Are there opportunities to make our supply chain more efficient through predictive analyses?
Can we use AI to recognise and prevent fraud attempts at an early stage?

How can the energy consumption of our plants be optimised using intelligent systems?

How can AI-supported tools improve occupational health and safety?

How can we improve our support through automated customer interactions?

Can we identify new business opportunities through AI-supported market analyses?

Can we improve our product quality through real-time monitoring and analysis?

How can AI help predict maintenance needs to minimise downtime?

How can AI optimise our marketing and targeting?

🚀 Categorise and prioritise the topics. For a mapping with possible AI solutions and further food for thought, we recommend the AI maps from, for example.

Step 3: A design sprint makes ideas tangible

A design sprint is a dynamic and focussed process that is specifically designed to move from abstract problems to concrete solutions in the shortest possible time. By working together in a multidisciplinary team, challenges are identified, innovative solutions are developed and immediately visualised. Testing these solutions directly with your target group enables you to gather valuable feedback and adapt your strategies accordingly. In addition to 2-3 visualised ideas, a design sprint can also result in a clear roadmap for your AI initiatives. While the design team develops prototypes, the technical team can evaluate existing AI services in parallel for a proof of concept. Incidentally, our specially developed GPT can support you during the design sprint.


If you also need personal support in the preparation or implementation of a design sprint, Deep Impact will be happy to assist you with its expertise and experience.

Step 4: Develop an MVPoc

The development of an MVPoc (Minimum Viable Proof of Concept) is a strategic method for quickly demonstrating the feasibility of an idea while developing the minimum scope of a product that allows valuable user feedback to be collected. It focuses on the core features needed to test the project’s key assumptions and gain insights into its viability and market need. This approach helps companies utilise time and resources efficiently by validating early on whether their AI solution actually addresses the problem it was developed for and whether there is demand before investing in full product development cycles. And the first hurdle of AI integration is also already technically validated.

Step 5: Scale and customise

Once your proof of concept provides feedback and data, it’s time to scale and customise your solution. The AI roadmap is not a static plan; it evolves with your business. At Deep Impact, we believe in agility and flexibility to respond to changes in the market and new insights. Don’t do everything at once, it’s worth prioritising the features clearly and reviewing the prioritisation regularly. This allows you to react to the market and pivot.

Tools: MoSCoW Prioritisation or Impact-Effort-Matrix (with Excel template for download)

Foundation: culture and leadership

Such a transformative technology requires a supportive culture and strong leadership. Leaders play a critical role in creating an environment for experimentation, learning and growth. At Deep Impact, we foster this with hackathons, mini-presentations and lots of proof of concepts. But that probably needs its own chapter and another blog post. Stay tuned…


Developing and implementing an AI roadmap is an ongoing and rewarding endeavour. With a clear understanding of the fundamentals, a strong focus on your business goals, rapid prototyping through design sprints, developing MVPocs, you can reap the benefits of AI. And more importantly: gain initial insights and grow from them.

At Deep Impact, we are here to guide you every step of the way.