The Flemish government created Flanders Investment & Trade (FIT) in 2005 to promote sustainable international business in Flanders as a key element in the socio-economic development of our region. FIT therefore has a twofold mission: to support Flemish companies with their international activities, and to attract foreign investors to Flanders. The agency has its headquarters in Brussels and 70 offices all over the world.

One of the Flanders Investment & Trade (FIT) agency’s main tasks is to conduct market research – for consultancy purposes and to identify global economic trends. It needs to be able to make this knowledge about Flanders and global markets available quickly and efficiently, and the agency is now using machine learning to facilitate this. AI company Arinti, which is part of De Cronos Groep, has developed FIT search – a search engine that doesn’t just search for specific terms but also understands additional, contextual information which it converts into new keywords. ‘The system has already evolved into a knowledge tool. Our initial aim was to refine our market research to make it faster and more efficient, and process commercial opportunities, but now the entire organisation benefits from having an intelligent knowledge database,’ says Peter Bulckaert, Head of the IT, Human Resources & Finance Department.

• FIT uses artificial intelligence to build smart search engine
• Tags and keywords are created automatically, including from photos and videos
• Experts can distil market research and process commercial opportunities faster

FIT supports Flemish export companies and facilitates investment projects in Flanders. It provides advice about commercial opportunities and performs market research on a daily basis, for example about our foods sector or the Flemish ecosystem of quantum computing, alongside many other topics. It should come as no surprise that the expertise at FIT is very widely distributed. It is written down to certain extent, but the FIT SharePoint archive already comprises 15,000 documents… and counting.

‘Structuring, categorising and archiving data is a very labour-intensive affair,’ says Yves Ruland, Data Project Manager at FIT. ‘We needed a way to access the right information and knowledge faster, which is why we developed a program together with Arinti – FIT Enterprise Search – to serve as a kind of smart search engine. We use the artificial intelligence from Microsoft Azure Cognitive Search, which the search engine is built with, to automatically extract keywords from documents so we can then use these tags for fine filtering. The smart tool can even recognise photos too, which is how AI creates extra metadata.’

Directory structure replaced with automatic indexing

The automatic creation of tags also raises the question of whether it still makes sense to organise documents using the classic tree structure with folders. ‘Everything is so fleeting now; if you structure things too much, the future catches up with you. That’s why we wanted to automate the indexing and searchability, so we don’t necessarily organise things in folders anymore, but we do bring some structure to the metadata. This means you still need to think about metadata a bit, but it takes less time than before, so it’s a combination of both for now,’ explains Ruland.

Knowledge system as an extra brain

‘We find the tool to be very powerful,’ says Bulckaert. ‘We all have lots of work to do, and artificial intelligence helps us deliver information faster. I can look details up quickly and easily when I have a meeting with a company, for example, so FIT Search acts as a virtual, extra brain. We’re also planning to expand the database accordingly.’

‘Refining the information faster was only one aspect, after all,’ adds Ruland. ‘The system has now evolved into a knowledge tool, so in addition to the reactive responses we’re also putting reference works online proactively. Our foreign network is our eyes and ear in these markets, and we use their feedback to draw up reports which are available on our website. The improved search function means we can now add to this research faster and more intelligently. And the “graphic” search results mean we can see connections between certain pieces of information too. The development of FIT Search as a knowledge tool is a key project for FIT’s AI strategy.’

Flanders accelerates

The project with Arinti fits within Flanders Accelerates! – the five-year plan to speed up development of the Flemish economy that is supported by the European Regional Development Fund (ERDF). Bulckaert: ‘Our ambition is to turn FIT into a data-driven organisation, and this machine learning project fits within this plan. We initially wanted to experiment with it, while at the same time also encouraging Flemish companies to start using machine learning and accelerate innovation. This didn’t require a huge financial investment; we’re really picking the low-hanging fruit here.’

FIT started researching AI and big data as early as 2017. According to Bulckaert, this mainly showed that AI works, but also that you can’t just use it arbitrarily. It therefore defined a strategy last year, with five projects planned for 2020-2022. One of these projects is knowledge retrieval, including FIT Search.

Even more AI in the future

FIT didn’t take any chances and scanned the market for AI providers. ‘Arinti stood out thanks to their pragmatic approach and innovative character. They’re at the cutting edge of innovation and efficiency, and they delivered the project quickly, which is essential in our competitive market. All countries want to raise their economic profile and are allocating resources to do so: our counterpart in the United Kingdom already employs around fifty data scientists, for example. Following on from this knowledge tool, we also want to automate conversations with organisations: personalised marketing and lead generation are our main points of focus for the coming year. We’ll of course be looking to Arinti for help with all these AI projects too.’

• 10,000: enquiries received from customers every year
• 15,000: documents searched by the new FIT search engine
• 5: strategic AI projects scheduled for 2020-2022