Nyhet -
Interview with LVI Numero Oy (LVI-INFO) CEO Magnus Siren regarding AI Proof of concept
In the past couple of months, Seavus has been working on implementing an AI solution that will drastically cut down on costs, time and human resources. In this interview, Magnus Siren, the Managing Director at LVI-Numero Oy (LVI-INFO), shares his experience with AI implementation and the long-term implications of having an AI member in the team. Read more about the whole process and the challenges of designing a solution that would simultaneously comply with the new standards and regulations, and act as a bridge between all parties involved.
The AI proof of concept turned out to be a great success! What was the main reason to implement an AI solution in the LVI-INFO business environment? What were your expectations?
I’ve heard so much about Artificial Intelligence and Machine Learning recently, and I was so excited to have a chance to implement it in a real business environment, to make a small test and see what it is capable of and how we can use it in our business database. I came across Seavus and realized that they have the know-how, and so we tried to figure out how we can come up with a solution that would deliver good results. That’s how Leavanny was created.
In the beginning, I thought “well, okay, this is nice but is it really useful?” So I didn’t have any high hopes. On the other hand, Seavus presented me an impressive use-case at our first meeting so I thought “well, I’ll give it a chance”. The end result turned out amazing!
Good and user-friendly product information is an important factor for the company. How can AI help in classifying the products in the database?
Our database has data for 160.000 products (even a little more than that) and, in parallel, there is a new technical standard called ETIM. It is a separate system of classifying and presenting technical information about products in a non-linguistic format, which is a completely different classification system from what we already had. In our own classification system, we divide products into sub-groups that needed to be matched to the products’ description to the ETIM classification.
Now, you can imagine what it would take for a person to sit down and do 160.000 of these matches manually. I don’t even think that one year for one person would be enough. On top of it, imagine how boring the whole process would be. Plus, I don’t think the accuracy would be so high. The Seavus team told me “this is not a problem, we have the ETIM system” and they matched it with the data from our database. We already had manually classified 10.000 products and Seavus used half of them to train the process and the other half (without the products being classified) to test it and see the results. What really was amazing, in my opinion, was the accuracy of the rest 5.000 products that we hadn’t classified compared to those 5.000 that we manually had. The accuracy was something like 99%. The numbers are really impressive.
And where is the LVI-INFO database on the HVAC market in Finland?
Our database is used as a link between the product suppliers, the manufacturers, the importers and the wholesalers that distribute the product on the market. This means that our purpose is to collect products’ data and then send it to the wholesalers and to the whole supply chain. More or less, all products sold on the Finnish market have their product data put into our database so the manufacturers don’t have to send it to everyone in the supply chain. So they only put it in one place and everyone can get it from there, from our database. In that sense, we are basically like a connector in the middle of the network, an important bridge between all parties involved.
What are the main advantages of having Leavanny as an AI member of the team?
First of all, I’d like to point out that this was the first phase of a pilot test to see whether the classification system would work. We tested Leavanny only the first 10.000 products. We still have a second phase where we’ll implement it on all 160.000 products. This just a one-time-work because all you have to do is put Leavanny at work and include all the data in the database. However, the difficulty is that along with all standards and products there are daily changes that need to be updated. This entails work on a continuous base – and it’s exactly where Leavanny steps in: to track and monitor these changes and automate the updates so that we could have a perfect match all the time. Not only it’s difficult to manually update all 160.000 products but it is also time-consuming to manually monitor and update the changes for each and every one of them.
Do you think AI will influence productivity levels and work dynamic, and what are the long-term implications of having AI in the business process?
Well, of course, it will but there are many new things we could build with this AI as well. Leavanny is just a small part and it already saves us a lot of time and resources. In the future, I see many opportunities for AI development and expansion into new business areas. We’re using, more or less, the same technology but we could focus it on other areas and tasks, which means that there are many things that could be done in order to improve the quality of the data in our database, which is crucial for our business.
An interesting fact: all product information in the database is in Finnish yet none of the team membersthat worked with the AI speaks the language. However, the AI reached an impressive 99% matching accuracy which is an excellent example of AI used in practice. Do you plan to fully integrate AI in your business environment?
Yes, that’s correct and it’s also quite amazing if you’re asking me! At the very beginning of the project, I thought “oh, well, we might have a problem here” but it turned out that there wasn’t. It’s a machine learning process, so it was the machine that learned Finnish. There are a lot of neuro-networking translation engines – you just need to know where to find them and implement them in the system.
Surely, we’ll continue with Seavus in other areas and I think we have a good future of implementing AI on a more continuous base in our business. We don’t have the time anymore to do all the updates, especially because the implementation of ETIM has doubled the information for each product and this is why AI is a perfect solution!
