– Interview with Ivan Jelić
When supply chains strain and economic forecasts remain uncertain, many executives turn to AI in search of clarity. But while ambition is high, few manage to translate AI potential into scalable, operational outcomes. Ivan Jelić, Group CEO and Managing Director for Switzerland & Germany at Joyful Craftsmen, has spent the past decade helping companies do exactly that.
In the conversation that follows, Ivan shares his perspective on why so many AI initiatives struggle, how to choose the right partners, and what it takes to make AI a long-term lever for growth.
Many AI initiatives never make it past the pilot phase, even though everyone knows the stats by now. From what you have seen in the field, why do so many projects stall out early? What tends to hold companies back?
It does, unfortunately. That failure rate absolutely matches what I’ve seen in many enterprise AI initiatives. One of the biggest issues is the lack of a clear business case. Companies often dive into AI without really defining the problem they’re trying to solve or understanding the value it should bring. And without a solid ROI, it’s hard to get long-term buy-in.
Another major reason is that IT simply isn’t ready. There’s no platform in place, no scalable approach, and what often happens is that teams build something on a shiny piece of software, but then can’t actually bring it into production or integrate it with existing systems.
And of course, we can’t talk about AI without talking about data. If the data quality isn’t there, if it’s incomplete, messy, or siloed, then even the best AI models won’t perform. So yes, the struggle is very real.
Without clarity and trust, adoption just doesn’t happen.
Are the biggest challenges always technical, or do cultural and organizational issues tend to get in the way too?
One thing I see quite often is that companies hire data scientists, give them a task to build AI models, and then just… push those models onto end users. But the people expected to use them often don’t fully understand what the model does, how it works, or whether they can trust it. That’s a huge problem. Without clarity and trust, adoption just doesn’t happen.
Another issue is the missing feedback loop, especially when it comes to data quality. If the people using the model spot issues or inconsistencies, or, more importantly, when they enter data that affects outcomes, there’s rarely a structured way for that information to flow back to the team that built it.
And without that loop, things don’t improve.
But at the core, AI is really about transforming how a business operates. And that kind of transformation only works if you bring people along on the journey. Many companies still underestimate how much of a cultural shift is needed. It’s not just about tools it’s about mindset, communication, and making sure everyone understands the “why” behind the change.
From what you have seen this year around AI adoption, where do companies most often lose strategic focus once they decide to move forward?
What I often see is that companies are still trying to figure out what all the AI hype actually means for their business. Some hand it over entirely to IT, hoping they’ll make sense of it. Others go in the opposite direction making big, bold investments without really knowing where they’re headed. Both approaches can lead to problems.
It’s absolutely critical to tie AI initiatives to strategic business goals. That connection gives you a solid foundation for identifying meaningful use cases, defining a clear path forward, and setting realistic expectations around investment and outcomes. Like with any major transformation, you need a clear vision of something that acts as your North Star and keeps everyone aligned.
And it’s important to understand that you can’t just skip the maturity steps. There’s a learning curve and entering the process gradually matters. Along the way, topics like data ethics, governance, and quality must be addressed not as afterthoughts, but as essential components of any serious AI initiative.
When companies move from AI strategy to actual delivery, where do you see the technical reality catch them off guard?
This is honestly one of the most common issues in fact, I’d say it shows up in about 60% of the cases we see. Companies often underestimate how much foundational work needs to happen before AI can even be generally adopted.
We’re frequently approached by cutting-edge AI startups who are doing amazing things on the modeling side, but they hit a wall when it comes to implementation simply because the client doesn’t have the data infrastructure in place. Sometimes, there’s no data platform at all. Other times, there is one, but it was originally built for something completely different, like financial reporting, and it’s not fit for the demands of AI.
So yes, the “dirty work” building pipelines, cleaning data, creating a scalable, reliable infrastructure it’s not glamorous, but it makes or breaks the AI transformation. Without it, even the best AI solution can’t go far.
There is clearly a lot of excitement around AI, especially with all the buzz around generative models. But when it comes to implementation, reality often looks very different. What kind of disconnect do you see between expectations and what actually happens inside companies?
The reason results often fall short is pretty simple expectations are set way too high from the start. And that’s not just on companies themselves; the buzz created by marketing campaigns, flashy conference talks, and even research papers from big advisory firms all contribute to this urgency and pressure. On one hand, that’s not entirely bad, since it helps push lagging organizations to finally act. But on the other hand, it can also create a false sense of how easy or fast this transformation is.
Personally, I think what we’re seeing isn’t so much a hype cycle anymore; it’s more of a wake-up call. What’s being showcased as success stories usually comes from maybe 15 to 20 percent of companies that are truly leading the way. The rest are still in the early stages.
The reality is, AI is fundamentally more disruptive than previous tech waves because the world today is more digital than ever before. That means literally every computer-based job, in every industry, is the objective of the AI transformation. And yes, that opens the door not just for efficiency, but for real disruption. Thinking about how cloud-first startups managed to challenge massive monopolies a decade ago, we see the samoe happening with AI now, just much faster.
And those “leading” companies? They didn’t crack some magical AI formula. Most of them started their digital transformation journeys ten years ago. They’ve been investing in innovation, building strong data foundations, applying lean and agile principles so adopting AI now feels like a natural continuation. What looks simple from the outside is really the result of years of disciplined work. That’s the gap many other companies are still trying to close.
You have been deeply involved in several projects already this year, whether through client work, events like the one in Zurich, or broader conversations across the industry. Has anything from the first half of 2025 really stood out to you as a lesson learned when it comes to AI adoption?
One thing that’s become really clear in the first half of 2025 is that no AI system works perfectly from day one. It’s just not how these things operate. It’s a matter of statistical viability and a big part of the data scientist’s job is to iterate, improve the model over time. If you don’t have enough data, or if the data quality is poor, the predictions will be off. It’s the same story regardless of whether we’re talking about GenAI, classic machine learning, or anything else: garbage in, garbage out.
We’re seeing that even with public GenAI systems, they can do certain things well, but because they’re trained on broad and often random data, they fall short when it comes to deep, industry-specific value. So, if you want AI to actually drive business outcomes in the enterprise, it needs to be purpose-built and trained in the right context to deliver meaningful results.
What really stood out for me over the past few months in client projects, at the Zurich event, and elsewhere, is how most use cases are still centered around cost optimization. That’s fine as a starting point, but I think there’s a missed opportunity here. Very few organizations see AI as something that could eventually become part of their intellectual property.
The reality is, investing in AI now could unlock massive, unforeseen benefits down the line. But it requires a structured, long-term approach, and patience to let your organization evolve. Take Amazon for example, it started as a platform to sell books, but now it sells everything. Along the way, it became one of the world’s biggest players in air freight logistics. That kind of business model innovation doesn’t happen overnight. You need a vision, and you need to trust the process — because real strategy often emerges as you go.
For leaders who want to approach AI seriously without getting lost in the buzz, what would you say are the most important first steps? How should companies set themselves up for success, especially if they’re just starting out?
Based on what I’ve seen, I’d group companies into three categories when it comes to AI readiness.
First, you have roughly 20% that know exactly what they want. They’re already on their way, and they mainly need support to scale whether that’s deep technical help, architectural guidance, or just enough capacity to keep things moving. Often, they’re juggling multiple AI projects and simply need help optimizing or managing the complexity.
Then, there’s the majority, around 60%, who are running AI proof-of-concepts but still don’t have a clear path to scale. Their efforts tend to be more opportunistic than strategic. With these clients, we usually start with an envisioning workshop and a structured assessment. We look at the current environment and categorize the gaps we see, typically across organizations, technology, and governance, and then create a plan to address them.
And finally, there’s the group that’s just getting started. They don’t have AI in place yet, but they want to understand how to begin. Here, our role is more about helping them build a vision and create realistic business cases, one that makes sense given their resources and capabilities. It’s about showing them a clear, doable path forward.
So regardless of where an organization is on the journey, the key is to be honest about your current state, set realistic expectations, and focus on building the right foundation especially around data, culture, and governance. AI isn’t magic, but with the right structure and mindset, it can deliver real value.
A lot of companies have now gone through that first phase: pilots, internal AI teams, strategy documents. What do you think comes next? Are we finally heading into the phase of real integration and scaling?
As I mentioned earlier, most companies are still in the phase of trying to figure out what AI really means for them. They’ve done some pilots, maybe formed internal teams, but now they’re asking: What’s next? That’s where we’re seeing a growing demand for Data & AI Strategy services. These companies need a structured, realistic plan, and that’s where we come in.
Typically, we work with them to build a three-year roadmap. It starts with a sharp six-month focus, then outlines clear objectives for the first year. What’s key here is that we help them shape this roadmap in a way that fits their internal budgeting process, which usually happens in the fall and winter. That way, the strategy isn’t just conceptual, but also something they can act on, invest in, and build into their planning cycle.
For most of these companies, the first steps in 2026 will involve starting to really work with their data, running initial pilots but also focusing heavily on building the right platform and governance foundations. It might sound like a slow process, but in reality, that’s what makes it sustainable. That’s what sets them up to scale successfully down the line.
Look beyond the buzzwords.
Ask if they can actually build.
There’s been a noticeable rise in firms calling themselves AI consultancies promising end-to-end solutions, from strategy all the way to implementation. What’s your take on that trend? And how can companies tell who is actually equipped to deliver?
That’s a great question and honestly, it’s something I’ve been thinking about for a while now. Even during my MBA the last years, we already had classes and exams on statistical modeling and training AI systems. That experience made it clear to me that awareness and capabilities around AI are growing and not just on the technical side, but increasingly within business functions too.
As someone from a data-focused consulting and engineering background, I can say this with confidence: AI consultancies won’t be able to survive in the long run if they don’t have strong Data foundations and practices. It’s not enough to just talk strategy or run a pilot on exported data. You need to understand how to build and operate data platforms data warehouses, data lakes that can handle real-time, complex data flows. That’s where the real impact of AI happens. It’s one thing to build a flashy prototype, and completely another to make it run reliably across real systems, month after month.
For us at Joyful Craftsmen, the challenge wasn’t the tech, since we’ve always had strong engineering capabilities. The real shift was learning how to bring more business acumen into our consulting work. Over the past few years, we’ve deliberately built that muscle, so that we can not only build the system, but also speak the language of our clients, understand their needs, and help apply AI where it matters, within their business units!
That’s also why we’re able to keep our promises: on budget, on time, and without smoke and mirrors. So, when companies are choosing partners in this space, my advice is simple: look beyond the buzzwords. Make sure the partner can actually deliver, technically, operationally, and with a solid understanding of your business.
Closing Thoughts
There are no quick wins or guaranteed shortcuts when it comes to making AI work, and that has been clear in every project Ivan has seen up close. Success doesn’t come from the next tool or trend; it comes from doing the steady work that earns trust, both inside and outside the system.
For most companies, that means accepting that real transformation isn’t just about technology but rather about the people, the pipelines, the governance, and the patience to keep building when the hype fades.
In a world that is always hunting for the next shiny fix, the companies that succeed with AI are the ones that treat it as a discipline and not a checkbox. That discipline is the difference between a pilot that looks good on slides and a system that works when no one is watching.
The post A field note: AI Ambition vs. Operational Reality in 2025 appeared first on SQLServerCentral.
