Artificial intelligence has moved from experimentation to everyday business discussions at an unprecedented pace. Yet despite the hype, many organizations still struggle to turn AI initiatives into real, measurable value.
In a recent ATR Soft webinar, Account Executive Juhani Siro, together with AI Specialist Iidaliisa Pardalin and Software Engineer Janne Marjalaakso, explored why so many AI projects fail, what successful projects do differently, and how AI is already reshaping the way we work.
Do AI Projects Fail?
Recent studies have claimed that as many as 80-95% of AI projects fail. These numbers are striking, but they also raise an important question: what does failure actually mean? In many cases, AI systems work technically as intended, yet the project is still labeled a failure because it produces no clear business impact or measurable return on investment.
A frequent mistake is focusing on technology, not the problem. Organizations often select AI tools during peak hype phases, hoping that AI itself will create value. This phenomenon is often referred to as the “shiny object syndrome”. It is like having the key before knowing which lock it should open. Without a clearly defined problem, even the most advanced AI solution is unlikely to succeed.
Another recurring issue is data. Despite years of discussion around data-driven organizations, many companies still struggle with incomplete, messy, or scattered data.
“The decades old principle of Garbage In, Garbage Out still stands today, even more so with AI. What we often still see is that data is missing, it’s messy or scattered around the place. Data is the backbone of AI,” says Iidaliisa.
If the data foundation is weak, the resulting models will inevitably reflect those weaknesses.
Finally, organizational silos and unrealistic expectations often undermine AI initiatives. Different teams may build similar solutions in parallel without coordination, while expectations of AI “fixing everything” create disappointment when reality proves more complex. Resistance to change and lack of AI skills can further slow adoption.
How to Make AI Projects Succeed?
Successful AI projects start with clarity. Before writing a single line of code, organizations must define what success looks like. Is the goal to save time, reduce costs, improve quality, or enable entirely new services? Without clear and measurable success criteria, it becomes impossible to evaluate outcomes objectively.
Equally important is finding the right problem to solve. The most successful projects often emerge when business and technical experts work closely together, building a shared understanding of real needs. External partners can also play a key role by bringing specialized expertise and helping organizations see beyond internal silos.
In many cases, the best starting points are the most mundane tasks; repetitive, boring processes that consume time and energy. These are typically easier to automate and often deliver the highest return on investment. Once these “low-hanging fruits” are addressed, organizations can move on to more ambitious goals.
“When you get processes optimized with AI, you’re essentially freeing up people’s time from those repetitive boring tasks to something more fruitful and they can be more innovative. They can be happier with their everyday work if they don’t have to focus so much on that stuff that nobody wants to do as much,” says Iidaliisa.
“With AI you should solve the most mundane, repetitive tasks because that is usually the most valuable approach.
These are usually considered the so-called most boring problems, but they also have the lowest difficulty to solve and they usually have the highest ROI,” adds Janne.
More mature AI users go beyond efficiency gains and start using AI to create entirely new business models and sources of growth. According to shared experiences in the webinar, organizations that redesign workflows and rethink their operating models tend to be the most satisfied with their AI journey.
Leadership also matters. AI adoption is not just a technical change but a cultural one. When leaders actively use AI themselves and clearly signal its importance, it significantly lowers psychological barriers and encourages broader adoption across the organization.
How AI Is Changing the Way We Work?
Beyond business metrics, AI is already transforming everyday work. By automating repetitive tasks, AI can free employees to focus on more meaningful, creative, and value adding activities. This shift has the potential to improve both productivity and job satisfaction.
At the same time, AI introduces new challenges. Research discussed in the webinar highlighted how employees may start working longer hours or multitasking excessively simply because AI makes it possible to do more. Without clear norms, this can lead to overload rather than empowerment.
To address this, organizations should establish clear AI practices: guidelines on when to use AI, when not to, and how to combine AI support with human judgment. Humans must remain responsible for final decisions, with AI acting as an assistant, not the decision-maker.
“In a lot of organizations, there are some people who are more scared about AI or more resistant to it and then there’s people who are really excited about the things it can bring to their work and they want to try new stuff. Harnessing those people’s enthusiasm is really key,” says Iidaliisa.
Communication and involvement are crucial. Openly explaining why AI is introduced, offering practical training, and creating feedback loops all help reduce fear and resistance. Many organizations also benefit from appointing e.g. AI champions, enthusiastic employees who support others and share success stories.
Moving Forward with Confidence
AI adoption is an iterative journey. Setbacks are inevitable, but they should be treated as learning opportunities rather than final failures. Persistence, realistic expectations, and a healthy organizational culture are ultimately more important than perfect planning.
At ATR Soft, the key takeaway is clear: successful AI projects are not about chasing hype, but about solving the right problems, building strong foundations, and keeping people at the center of change.
“The most important thing is to not let AI take the wheel. Humans are still the ones making the final decisions and while AI can support us and make our work more enjoyable, it shouldn’t replace human judgement. It would be good if organizations had clear rules for the use of AI, when to use it, when not to, but also when to trust AI and when not to,” sums up Iidaliisa.
