Creating value with AI
by Ole Breulmann, 12/11/2024
Teaser: AI is just one part of your car's engine. The rest of the car must function well to win the race. 🏎️
First of all welcome to my newsletter. This is actually me writing - Ole - and not ChatGPT. Surprise. ⭐️
I am personally bored with ChatGPT-ish, „perfect“ text. LLMs are a huge win for all companies in terms of automation and semi-automation of tasks and in that extend I am so happy about the current technology wave and how it enables us to build innovative impactful solutions, but I need to read text that is more „human“. So I will try to write „imperfectly“ and „more human“ in this format. FYI.
This first issue is about value creation with AI and how to get it right. AI has been around for decades. I have built classification systems, recommendation systems and reasoning solutions around 2010 with the same fundamental concepts which are used today in LLMs: Semantic vector spaces („embedding space“), neural networks and graphs. Since then AI science and engineering advanced significantly - like all scientific disciplines - and now we are where we are: In the total AI hype. I love that. For me it is like closing a circle. I started with AI - neural information processing and computer vision - spent almost two decades as data scientist, software developer, product manager and CTO/CPO and now everything is coming together at the right time. I am very grateful to the universe for this dynamic around me. Being at the right place at the right time. Thanks universe. 🪐
As being somebody who has one leg in the „technology“ basket and one in the „business“ basket, building bridges so that effective innovation is possible, one part of my work has always been to give both baskets a better understanding of the other. I listen. I talk. I draw. Over and over again. To drive alignment. Convergence. For that I have to continuously identify assumptions / beliefs / misconceptions that need attention to make convergence possible. One such misconception that I have recognized when it comes to AI-based value creation is:
„For creating competitive advantages with AI the most important success factor is proprietary data. “
That is an incomplete perspective. While proprietary data is undeniably important, I've found that many companies overlook other equally critical factors that can make or break AI success.
It is true, that for training machine learning models you need data. Acquiring or generating the right data for training models is crucial. Even if you are using foundation models - which you do not train from scratch but fine-tune or just use as they are - you need use case specific data in order to evaluate and optimize performance.
So yes, having use case relevant data can create a competitive advantage. However, the above incomplete perspective is dangerous because it overlooks three equally important success factors for AI value creation and competitive advantage: product discovery, product design, and solution architecture.
These factors go beyond just having data—they are about how you approach the process of creating solutions that deliver real value. Success with AI isn’t just about the technology; it’s about the strategy, design, and execution that bring it to life.
If you want to create competitive advantage with AI ask yourself this:
Do I have an efficient and effective product discovery team and process? Do we validate our assumptions well enough before building? Are we using all the data we can get to make the best investment decisions in product development?
Do I have a strong product design team and process? Do we explore solution opportunities in a cross-domain team before jumping the gun and building something? Does my management know that design is not how it looks, but how it works?
Do I have a strong solution architecture team? Do we build for scaling use cases and clients? Do we take the time to do things right?
Do I have the AI team to use AI effectively? Do we take the time to build evaluation, monitoring, and quality assurance setups?
If your answer to any of these questions is a NO, you know your homework. And yes, this is a lot. And yes, it is not easy. It is about building or transforming your company right. And it is about finding the right people for these disciplines and giving them the right environment to excel. AI is just one part of your car's engine. The rest of the car must function well to win the race. 🏎️
In the last 8 years I have seen multiple, promising companies fail because of one or many of the above aspects. The devil is in the detail. And the detail can break your neck.
At this point this issue is done. Feel free to give me feedback on LinkedIn in a DM.
I need data to find product-market fit for this newsletter ;)
Best, Ole
