Why AI hasn't yet translated into major productivity gains for businesses
While artificial intelligence can speed up individual tasks by 20–50%, its effect on overall company productivity remains far more modest, according to economist Edgars Čerkovskis. Latvia still lags behind the EU average in AI adoption.
Artificial intelligence can significantly speed up specific tasks, but its impact on overall economic productivity is growing much more slowly, according to economist and AI Leaders School network expert Edgars Čerkovskis. Swedbank estimates suggest total factor productivity could rise by only around 1% per year. In a large international survey of nearly 6,000 business leaders, about 70% said they use AI, yet over the next three years they expect an average productivity gain of just 1.44%.
AI adoption in Latvia remains relatively cautious: Eurostat data show that 12.2% of Latvian companies used AI last year, compared with an EU average of 20%. According to the Central Statistical Bureau, more than 60% of companies that considered adopting AI but did not cite a lack of necessary skills as the main barrier.
Why expectations haven't been met
Čerkovskis notes that many business leaders had hoped AI would quickly automate large parts of their processes, saving time and costs, but such changes are a long-term undertaking. One common mistake is trying to simply "bolt on" AI to existing processes without fundamentally changing how the business operates. The greatest benefit, he argues, comes not from handing a single task to the technology but from redesigning the process itself — comparable to genuine digitalization rather than simply moving a paper form online.
A second challenge is a shortage of expertise: Latvia lacks AI specialists, and companies need people who understand both internal processes and what the technology can actually do. Cost is another factor, since implementing and maintaining more complex AI solutions can require substantial investment.
Where AI already performs well
The expert points out that AI already works effectively for data analysis, text preparation, translation, image generation, and document creation — routine, clearly defined tasks where results can be measured fastest. More caution is needed, he says, before entrusting AI with complex, high-level tasks such as sales or client acquisition.
Čerkovskis recommends that companies measure productivity not by the number of licenses purchased, but by how much time is saved and how that freed-up time is used for higher-value work.
