Over the past three years, many organizations have gained momentum in leveraging AI, and the benefits are now showing up in everyday operations. What have we learned about best practices during this time? What works – and what doesn’t?
The release of ChatGPT three years ago sparked an unprecedented technological shift, prompting nearly every organization to consider how AI could be leveraged. In both customer conversations and public discourse, one question has stood out: how can AI deliver real, measurable value in everyday operations?
At Digia, we take a longer-term perspective. Businesses have been utilizing various AI applications long before the emergence of generative AI.
Here’s a distilled summary of what we’ve learned from years of customer projects and conversations: what works with AI – and what definitely doesn’t.
Everyone knows an AI project that didn’t go as planned
International studies report that over 70% of AI pilots fail to reach production. In August 2025, a report from MIT made headlines by revealing that 95% of them produce no measurable results at all.
The root causes of failure are rarely technical. More often, the challenges stem from organizational practices or even strategic-level issues.
Based on our observations, the most common pitfalls include:
- Shiny toy syndrome: AI is bought “because others have it.” Projects are launched without a clear business goal or ownership. The focus is on technology, not problem-solving.
- Poor integration: AI solutions are developed in isolated silos and not connected to core processes.
- Unclear ROI metrics: Benefits are not defined in advance, making success impossible to measure.
- Organizational challenges: Resistance to change, lack of skills, and siloed operations hinder effective AI adoption.
In other words: Don’t buy AI. Solve business problems.
Success is only 10% technology – so what does it depend on?
Time and again, it becomes clear that successful AI adoption is more than just technology.
So what determines success in AI projects? In our view:
But first… Strategic choices you can’t avoid – and leadership you can’t outsource
Before any of the above matters, or success is even possible, a few strategic decisions must be made about what you aim to achieve with AI and where to focus. AI requires investment, and to invest wisely, you need a clear view of where AI can deliver the greatest value for your organization. In short: AI demands leadership.
After the first experiments and easy wins, the biggest strategic choice becomes: do you optimize what exists – or build something entirely new?
- AI-aided business: AI is used to enhance existing processes and services. It brings efficiency and improves quality, but doesn’t fundamentally change the business.
- AI-driven business: This is the more ambitious path. AI is placed at the core of the business, enabling new services, processes, and business models. It requires careful evaluation – but bold transformation can unlock greater competitive advantage.
Summary: Do this to get measurable value from AI
1. Identify strategic use cases that generate measurable value. Only those are worth investing in.
2. Lead the change and engage people from day one. This is much more about ways of working than an IT project.
3. Ensure AI solutions move from idea to production – and deliver results. Moving from pilot to production is one of today’s bottlenecks. However, it is only in production that AI creates value for the organization. Additionally, ensure compliance with AI-related regulations and data protection laws.
4. Manage the full lifecycle of AI through continuous development and governance. AI remembers and learns. An AI-based application is a dynamically evolving system in production – unlike traditional software – and requires a different kind of maintenance.
So, to sum it up: don’t expect AI to be your company’s all-knowing genius.
It’s more like an overly eager intern who’s read the entire internet but has no clue what your company actually does. Without proper leadership, onboarding, and clear tasks, it might confuse the break room with the boardroom – and endlessly print cat videos. Meanwhile, the bill keeps running.
Want to hear more, ask questions, or join the conversation?Sign up for the AI Morning on January 20th, which will challenge your thinking on how to unlock the benefits of artificial intelligence. The event is in Finnish.