Struggling to put AI to work? Start with a pilot
Artificial intelligence won’t wait. Companies that start using artificial intelligence now will, in doing so, build up the expertise that will set them apart from their competitors in the years to come.
Finland ranks second in the EU in terms of the use of artificial intelligence among businesses. According to Statistics Finland, as many as 38 percent of businesses were already using AI technologies in spring 2025, and this figure rose by 14 percentage points in just one year. The enthusiasm is therefore genuine, and the need has been recognized. Nevertheless, in many companies, the use of AI is still limited to isolated experiments and has not progressed to a defined, measurable pilot program.
In this article, we’ll look at why this is the case, how to identify the first suitable pilot project for AI, and how to get started quickly.
Where does the use of artificial intelligence yield results most quickly?
The first pilot project doesn’t usually fail because there is a lack of potential applications. On the contrary: there are often too many options. That is why the best place to start is with processes where the same tasks are repeated regularly, take up a significant amount of time and involve a great deal of manual data searching or transfer. In such use cases, the impact of artificial intelligence is quickly apparent, results are easier to measure, and the risk of human error is reduced.
Suitable pilot projects for Finnish companies’ daily operations can often be found in at least the following use cases:
- Order processing: Entering orders received by email or via forms into the enterprise resource planning system is time-consuming and prone to errors. This use case is particularly common in B2B commerce and the manufacturing industry. Artificial intelligence extracts the data and forwards it automatically – as in the solution we have already implemented for several companies, where an AI order processor handles orders arriving in Odoo ERP in a matter of seconds.
- Searching for documents: Technical documentation, contracts, maintenance instructions – when information is scattered, searching for it eats into working hours every day. A RAG-based AI bot helps you find answers in natural language, even from extensive document repositories, and the answers are based on the company’s own documentation.
- Customer service routines: Responding to repetitive queries takes up employee resources that could be freed up for more complex tasks. An RAG-based AI chatbot added to the website can handle routine inquiries automatically and is able to respond to customers based on the company’s own knowledge base.
- Reporting and data compilation: reports compiled from multiple sources are being automated, freeing up working hours for the interpretation of results, a process in which artificial intelligence also plays a part by identifying anomalies and highlighting key findings.
Why do AI projects get stuck before the first pilot?
Minna Helle had already summarized the key challenge associated with the adoption of artificial intelligence even before becoming CEO of EK. In an interview for ‘On the Management Agenda’, Helle emphasized that the adoption of technology does not, in itself, lead to a leap in productivity. Instead, companies need to encourage their employees to utilize, experiment with and embrace new ways of working. In the long run, she says, it is above all a question of leadership.
Helle’s observation still hits the nail on the head. Often, the first obstacle is not the technology, but a lack of a shared vision and consensus. AI solutions are available, and many companies have already identified situations where they could be useful. Yet the first pilot that would genuinely benefit the business often fails to get off the ground because no shared view emerges within the organization as to where to start.
The IT department is considering integrations, management is reviewing the budget, and the business side is weighing which problem should be tackled first. At the same time, the development teams have other projects underway. When ownership remains unclear, there is certainly discussion about artificial intelligence, but the first concrete step is never taken.
What do successful companies do differently?
If the initial use case is not defined clearly enough, the same problem will carry over into the later stages of the pilot. According to MIT’s ‘State of AI in Business 2025’ report, the problem with AI implementation is often not the quality of the models, regulation or infrastructure, but rather that the solutions do not fit sufficiently well into everyday workflows, learn from feedback or adapt to the organisation’s own context. The report also notes that only a small share of integrated AI pilots generate measurable business value.
According to the Boston Consulting Group, successful organizations take a different approach: rather than trying to roll out AI everywhere at once, they focus on a smaller number of use cases that are critical to their business and measure the impact of these. In other words, AI delivers benefits when the use case is sufficiently narrow, integrated into daily processes and justified from a business perspective.
The problem, therefore, is not usually a lack of AI capabilities or even a lack of budget. The problem lies in the fact that the first area of application must be defined, justified and decided upon. When this issue gets bogged down within the organization, with different viewpoints vying for attention, a structure is needed to take the discussion from the brainstorming stage to a final decision.
Outside help makes it easier to make a decision
The same challenge regarding deployment has also been recognized internationally. AI companies are no longer talking solely about models and tools, but increasingly about how companies can actually put AI to use. In 2026, OpenAI launched the OpenAI Deployment Company to help organizations identify the most impactful use cases and bring AI solutions into production. Anthropic has also strengthened its partner network and deployment support for the enterprise use of Claude.
The trend is clear: the biggest challenge is often not the technology, but deciding where to start. Which use case is sufficiently narrow in scope, yet still relevant to the business? And how can a decision be made so that the matter doesn’t get passed back and forth between different teams?
An outside perspective is helpful here. When a party is involved that is not burdened by the organization’s internal baggage, the discussion can move more quickly toward concrete decisions. Options can be laid out side by side, the benefits assessed in practical terms, and a decision on the first pilot can be made during the discussion itself.
We help identify which processes are ready for piloting and which require further development first. We have implemented AI solutions for companies of all sizes, ranging from order processing automation to RAG bots, which enable employees to access the company’s own data quickly and via natural language queries.
We are not tied to any particular platform or individual solution. That is why we can take a process-led approach: where does AI deliver the quickest measurable benefits, what type of solution is best suited to this, and how should the first pilot be scoped?
A workshop lasting a few hours is enough to get you started
Finding a pilot project for artificial intelligence does not require months of upfront work. When the process has a clear structure and is supported by an experienced facilitator and a technical expert who understand both the potential of AI and its practical implementation, a decision can be made in a matter of hours.
We have developed the AI Sprint service for this purpose. The process begins with a short preliminary questionnaire to assess the current situation. The workshop lasts two to three hours and can be held remotely. The outcome is a written report and a roadmap: what the first pilot project will be, how it will be implemented, and what savings it is likely to generate.
For companies seeking funding for development projects, the ELY Center’s business development grant may, on a case-by-case basis, be suitable for launching an AI pilot. However, it is advisable to check eligibility for funding with an expert in your region before starting the project.
If you’ve already identified a process that AI could help streamline, or if you simply feel that a particular part of your daily workflow is taking up too much time, now is the time to get started. Get in touch – we’ll help you move forward. In just a few hours, you’ll know where to start.
Shall we get started?
"*" indicates required fields
The benefits of Figma prototyping in a software project