The Most Expensive Mistakes Organizations Keep Making With AI (Since 5 Years)

Today’s newsletter is co-authored by Dat Tran, former head of AI at Axel Springer SE and an experienced AI entrepreneur.

You don't have to sell us on AI. We're convinced AI will revolutionize countless industries, bringing efficiency gains and innovations we couldn't have imagined just a few years ago. And in some industries, this disruption is already in full swing.

Yet, despite all the advancements and buzz, some fundamental mistakes in AI adoption seem to stick around. We've seen it firsthand: the same mistakes made five years ago are still being repeated today, and it's crucial to understand why.

That's why today we want to cover what these mistakes are and which strategies you can apply to avoid them.

Let's go!

Some things in AI adoption (sadly) never change

So, what's going on here? Why do companies keep tripping over the same issues with AI? The answer is simple: fueled by the hype, many are diving headfirst into AI without a solid strategy and—above all—a solid understanding of the fundamentals.

It's reminiscent of past hype cycles like "Big Data" and "Data Science". Success stories from big tech companies are making waves and leaving everyone mesmerized. "If Google does it, we should probably do it as well!" The obsession with the latest trends and the tendency to follow trend-setters lead to a shiny object syndrome and a repeat of past mistakes.

But today, we don't want to point fingers. Instead we want to do a reality check and cut through the noise with a focus on what really matters: integrating AI in a way that makes sense for your business. Because when approached from the right angle, AI can truly unlock 10x productivity gains and give you a 5-year advantage over your competition. 

So let's get into the specifics of where companies are still going wrong and how you can avoid these pitfalls to truly leverage AI’s power.

Mistake 1: Not Treating AI as part of A larger ecosystem

Too many companies are fixated on AI, thinking it's the end-all solution. But in reality, AI is only one part of the equation (even though a big one in some use cases). You also need great software, an excellent user experience, and a solid business model to make it work. We've talked to countless companies that are so blinded by the GenAI hype that they invest solely in LLMs without a clear strategy on how to integrate machine learning into their products effectively.

Think about it: you wouldn’t buy a fancy new engine for your car without ensuring the rest of the car is in top shape, right? The same goes for AI. It needs to be integrated into a well-rounded product ecosystem to truly shine.

Mistake 2: The Persistent “PoC” Syndrome

Everyone loves a good proof of concept or prototype. Us, too! They're easy to build, show early value, and often make you look good in an organization. And make no mistake—PoCs or prototypes are vital in your AI strategy because it wouldn't make sense to go all in on a solution which you haven't tested on a small scale first. But some organizations are taking PoCs to the other extreme—churning out PoCs without ever bringing anything into production.

We're not saying every PoC must work. There's a high rate of failure in PoCs by definition. But for every PoC, you should have a solid strategy for what happens next if your assumptions are validated. This means:

• Preparing for agile projects

• Setting up implementation teams

• Building internal projects and gathering support

• Creating a product roadmap detailing feature releases

Too often, if you ask about the next steps for a prototype project, the answer is unclear. So yes, do prototypes, but only if they're part of a larger roadmap.

A PoC is a great way to test the waters, but it shouldn't be the end goal. It's like dipping your toe in the pool but never diving in. The focus needs to shift from endless experimentation to actual implementation. It’s time to move beyond the PoC phase and start thinking about how to scale these projects into real, impactful solutions.

Mistake 3: Just “Buying”, Not “Making” 

Another recurring issue is the classic "make or buy" dilemma. Companies are great at making "buy" decisions—they'll happily purchase AI solutions off the shelf. But when it comes to building and integrating these solutions, many non-tech companies hit a wall. They suddenly find out that integrating vendor solutions into their own systems requires deep technical expertise and product management skills. When it’s time for implementation, this knowledge is often lacking within the organization, leading to two problematic outcomes: hiring or contracting outside help (which is expensive and undermines the business case), or putting the AI project on the back burner until the initial excitement fades and people forget about it.

Building a strong product and engineering team is tough, especially in large, legacy-driven organizations where the business isn't tech-focused. However, organizations need to recognize that having a solid technical foundation is essential in the 21st century. This starts with having a competent CTO and extends to capable internal engineering teams. Additionally, CTOs should be on an equal footing with the CEO, rather than just reporting directly to them.

While buying AI solutions can be a quick fix and an easy start to your AI journey, it often leads to a lack of deep integration and understanding of AI within the company. Building in-house expertise might be harder initially, but it pays off in the long run by creating more cohesive and effective AI-driven solutions—especially if you’re operating in an industry that is highly disrupted by AI.

Mistake 4: Ignoring the people aspect - a.k.a. Culture

Lastly, and perhaps most importantly, is the need to create an AI-friendly culture. We're deliberately saying not an "AI-culture" because that's not the ultimate goal. The ultimate goal is to build a culture where fast innovation, experimentation, and curiosity are encouraged, putting your business in a position of accelerated growth and not "managed decline" (as Sam Altman puts it).

Fostering this culture doesn't happen overnight and it doesn't start from the ground up as a grassroots movement. It needs to start from the top.

Executives might know about ChatGPT, but their understanding often doesn't go much deeper. You need to lead by example and simultaneously invest in your employees to drive your digital and AI agenda.

"But training people is costly and the outcome is uncertain," you might say.

You’re right. But it's a prioritization problem, not a budget problem. Companies are spending massive amounts on AI projects on external consultants—Accenture alone has booked $2 billion in AI projects by the first half of 2024— instead of investing in their own teams. Imagine if that money were spent on upskilling employees and paying them adequately. Consultancy isn't bad, but it should complement, not replace, in-house expertise.

Put your people first, lead by example, and the culture will follow. And yes, that also includes top compensation and working conditions for your top talent.

Conclusion

The key to successful AI adoption is to avoid the hype traps, while integrating AI thoughtfully into your broader ecosystem, and building in-house expertise over time.

Let’s break the cycle of repeated mistakes.

When done right, AI can deliver incredible value and set you years ahead of your competition. But this requires a shift in mindset. It's not about chasing the latest trends or mimicking big tech companies. It's about understanding your unique needs and crafting a strategy that fits.

Invest in your people, cultivate a culture of innovation, and take a high-agency approach to building and integrating AI solutions. Avoid quick fixes and shiny new tools that look good on PowerPoint. 

Real success comes from a deep, sustained commitment to getting it right.

AI is a powerful tool, but it’s only as effective as the vision and execution behind it. Commit to smarter, more strategic AI adoption.

Your business’s future depends on it.

Good luck!

Dat & Tobias