There’s plenty that businesses could learn from Estonia’s digital transformation. For years, the country has allowed residents to file their taxes online, and to vote over the internet using government-issued smart ID cards.
It has a nationwide system of digital health records, and its land register is also online. These systems are decentralized but interconnected through an open-source data exchange platform, X-Road, which has also been adopted in Finland and Iceland.
Now Estonia is setting out its national strategy on artificial intelligence. While it’s not the first country to do so, its pragmatic approach could serve as an example to enterprises figuring out their own approach to AI — as well as opening up business opportunities for some.
A pragmatic approach to AI
Estonia sits on the shore of the Baltic Sea, with Russia to the East and, a short hop across the water, Finland to the North and Sweden to the West. The country is tiny, about half the size of Maine, but with the same population, around 1.3 million.
“Coming from a small country like we are, we have to do the most with the little resource we have. That’s why we started experimenting with digital tech in the first place,” says Estonia’s government CIO Siim Sikkut.
Sikkut has been in his post a little over two years and says the importance of AI was apparent early on: “We have to do more with emerging technologies, and AI is in the forefront of that.”
Last fall, Sikkut’s team began working with EY (Ernst & Young) on developing a three-part national AI strategy, which will soon be submitted to the government and published in English later this year. The first part of the strategy focuses on making the most of AI in government, an area in which Sikkut has been trying to introduce a conscious strategy of experimentation.
Estonia is not aiming to be disruptive with its AI strategy, rather it is taking a pragmatic approach. “Our aim is to solve practical problems in our government, in service delivery, efficiency and so forth,” Sikkut says. “If we can use what somebody else has done, perfect. If nobody else has done it, then we will innovate.”
Estonia already has a few AI-based use cases in the works, including:
- bringing chatbots into government interactions with citizens;
- using image recognition technology in agricultural inspections to reduce the need to send people out into the fields;
- speech-to-text systems that can provide court transcripts automatically;
- optimizing police patrols.
Predictive policing is a controversial practice, notably for the risk of reinforcing racial bias, but Sikkut insists, “We want transparency and responsibility built-in.” For now, police trials are focusing on things such as how weather conditions influence traffic congestion in order to better deploy traffic patrols.
Broadening the AI skills pool
The second part of Estonia’s strategy focuses on boosting the uptake of AI across the country’s economy. That’s the industrial and innovation policy side of it, Sikkut says.
For that, Sikkut’s team has sought the views of business. Estonia has a strong technology industry and a thriving startup scene, and a key theme to emerge there is the need for more skilled workers. “Give us the talent and we’ll do wonders,” they told Sikkut, in essence.
To meet that demand, Estonia has increased the number of places available in higher education to study computer science. It has also targeted a particular bottleneck for businesses: data science. “We have brought in some new funding for data science, so they have researchers and some new curriculums around that,” Sikkut says.
The country has also focused on attracting talent from elsewhere, streamlining its immigration process to make it easy to bring in technology specialists. Other measures include special visas for founders to build startups in Estonia, and a government program to attract candidates from outside Estonia to fill technology industry vacancies. “We have relatively good returns on this,” he says.
Estonia’s traditional industries, including forestry, metallurgy and industrial equipment manufacturing, face other challenges on the path to improving productivity with AI. “There, the biggest issues really are about how to get started in the first place, how to govern the data, the AI, the machine learning in the companies. Much more basic stuff,” he says.
Multinationals haven’t been forgotten. Estonia is also looking at how to make itself more attractive as an investment destination, allowing businesses to draw on the country’s pool of skills in particular industries. “We’ve had some initial explorations in the health space. There’s promise,” says Sikkut.
Going all in on AI
The third part of Estonia’s AI strategy is about the regulatory environment: “We want to make full use of AI, not just go halfway,” he says. That means asking, “Is there something in our law that we need to fix, or provide some clarity on, so that fully autonomous software systems could be used freely?”
Clearing up such ambiguities is necessary to encourage the development of autonomous systems, whether they are driving taxis or deciding who can obtain a bank loan. But it remains a delicate balance, as businesses need to understand their exposure to risk and citizens want to know who is culpable if an autonomous system has a harmful impact. “We do it in order to assure people that look, you can trust them and if stuff happens there will be a recourse like there is in ‘real’ life anyway, in the physical world,” he says.
Based on his experience, Sikkut has one key piece of advice for CIOs considering how to bring AI into their enterprise: Just do it. “Quite often in big corporate structures, there’s like a big corporate strategy process and then you perhaps get to do something. I think that needs to change, clearly, especially the more tech moves ahead faster and faster. That would be my advice: Just start doing it, and then through that build up a strategy based on what works.”
And don’t be afraid to emulate what others are doing. “Let’s copy more. Be very pragmatic about it. I would like to think that AI, at least as it stands, doesn’t have to be deep, deep science. There’s a lot of deep science involved, and a lot of great companies who do this and those who get it, it’s awesome,” he says. “But even the existing models, libraries, off-the-shelf stuff, there’s so much stuff you can reuse and if you just apply it wisely then you can move ahead so fast in terms of being an adopter of this tech, you don’t have to be the innovator of it necessarily. So, just get going.”