Strategic Innovation with AI
After the initial hype around generative AI, we're now seeing a wave of reports highlighting high failure rates of AI projects and the reasons behind them. This isn’t surprising — with any new technology, early enthusiasm inevitably gives way to the reality of steep learning curves and missteps.
The shift we’re experiencing with AI right now is one of the biggest we’ve seen. But even with this quite extraordinary new hammer, not every problem is a nail. We are still figuring out how to best use this new, shiny tool, that is rapidly evolving. Thus; experimentation is necessary. Recent headlines makes us think that it is time to repeat some fundamentals around innovation, experimentation and failure.
The experience from a number of innovative and leading tech companies, has taught us that the recipe for innovation is to try a lot of different ideas in order to find the ones that truly push the boundaries. This goes for both technical explorations and validating business ideas, and includes practices of e.g. proof-of-concepts and A/B-testing. However, since money isn’t exactly cheap nowadays, it is necessary to be strategic and purposeful with the investments you put into innovation. The recent emphasis on starting with a business case before you pour money into AI explorations is absolutely necessary, under pretty much all circumstances. But that doesn’t mean that we can, or should, expect every proof-of-concept to go into production.
How do you decide which ideas to try - and how do you know if you should move forward?
Failing fast -and cheap - has never been more important. In order to do so, you need a strategic approach to innovation. What ideas are you trying, how much do you invest into them, when do you move forward and when do you abandon them and try something else? The fundamentals are the same for AI projects as for product development and innovation in general. With AI it’s slightly more complicated because the technical landscape is moving so fast. An idea that doesn’t work today might work in 3 months, when the tech has evolved. Add to this the fact that a lot of AI development right now is fear driven. Fear can create a sense of urgency, which is not always a bad thing, but it is seldom a good growing ground for innovation.
Adding a bit of structure will help you make informed decisions. There are some learnings from past decades around innovation that we can apply:
1. State Your Hypothesis
Before you start a proof-of-concept or experiment, make sure you have a clear and shared understanding of what you are aiming for. What hypothesis are you trying to prove, or disprove? Set up clear success metrics and stick to them. Whether it is to test a new business or product idea or to gain technical experience or both, the goal should be clearly expressed, as well as how much you’re willing to invest to get there.
2. Build vs. Buy
Make strategic decisions on what to experiment with in-house, what to outsource and what to buy off the shelf. These choices vary from business to business, methods such as Wardley mapping can help you figure out what truly differentiates your business. Adding AI capabilities in those areas will take longer than implementing the low-hanging fruits, but the strategic value is much higher. That doesn’t mean that you shouldn’t be using AI also in other areas, but look into buying it from someone else to avoid re-inventing the wheel, and adding a long-running maintenance-burden. Very few companies need to train their own LLMs for example.
3. Avoid Sunken Cost Fallacy
If your AI initiative doesn’t give you the result you were looking for, i.e. if your hypothesis is proven wrong, abandon it and move on to something else. Not all experiments will be successful. If you believe the idea might still benefit from AI, but the tech is not mature enough yet, pause and try again later.
4. Invest in Learning
Success with AI starts with competence. Ensure your internal teams have the right skills, or partner with experienced external partners to bridge gaps. You need both a good business understanding and technical skills to succeed with AI, but in many cases, you don’t need to have all expertise in-house.
5. Stay Curious, Not Fearful
It’s easy to become fear-driven in the face of uncertainty and rapid change. To be successful, try to resist that urge. Instead, stay curious, experiment thoughtfully, and embrace the learning process. When studies report that 88% of genAI pilots are not reaching production, it’s not a reason to panic — it’s a sign that we are still early on the journey.
GenAI has started maturing, but it’s far too soon to say what we can make of this new technology. As Amara’s law goes: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”
