AI technologies used to be of quite a low standard ten years ago but have made a quantum leap over the last decade. What are your predictions for next ten years of AI technologies?

AI has evolved a lot in the last ten years thanks to Deep Learning. What Deep Learning is very good at is taking a lot of very complicated data and using that data to make very accurate predictions. Deep Learning has been able to predict things very well. And going forward, it is starting to have the ability to generate things. What do I mean by generate things? I mean that AI is now at the point where it can create a sentence that sounds like a human created it. It can create a picture that looks like a photo. We are at the point where you are going to start seeing AI generate things. So, in the next ten years, I believe that’s going to happen.

How would you comment on the practical adoption of AI among startups in Vietnam, with many having a tendency to put an AI label on many of the products they provide? What are their successful achievements?

I can talk about what we have seen from Appier’s side. We have had some successful customers in Vietnam. First is Nivea Men, a man’s skincare company.

Firstly, they care a lot about brand safety. They don’t want to be sending ads to any one person too many times. They want to be able to target people on the right channels. In order to help them, we used AI to predict what devices were used by the same user. Once we know which devices were owned by the same user, it stops us from sending too many ads to them. Also, AI helps us to reach the right device at the right time, because there are a lot of people browsing on weekends on a certain device but on weekdays they use a different device.

On top of that, we want to understand what customers’ interests are. Using AI, we can predict and know that, in general, people who like Nivea Men also care about sports and current affairs. The combination of all this allows us to improve the performance of their advertisements, with a 22 per cent increase in the click through rate (CTR).

Another client wanted to re-engage their customers. The first thing we did was use this information to segment their customers into a lot of tiny little segments. Using AI, we can break it down to more than hundreds and thousands of segments. For each segment, we then predict their LTV (lifetime value). This allows us to pick the best segments to target. Once we have picked the segment, we can again use AI to decide which products are the best to send them. So, each different person might see different products. This has had significant results, with an increase of about 20 per cent in conversion.

Corporates now have two options: setting up their own AI department or choosing to contract a third party to develop AI. In which cases would you advise companies set up their own AI department and in which would you advise they contract with a vendor?

Generally, companies do both at the same time. Some might build their own teams. But when you start doing this, you need people who know how to do it. Then, it is going to take a lot of investment. So, companies that are able to follow this trend are companies with a lot of resources. But even in these cases, you might want to consider working with vendors on specific problems. Because vendors do not consult on all of the problems. They tend to be very good at what they are doing, so there is room for a hybrid model.

Appier has availability in all of the Asia-Pacific, which has also been a hot spot for AI development in recent years. How would you comment on the competition?

There are many AI companies in the Asia-Pacific region. One thing that I think we do very well is we don’t just provide AI solutions - we provide an end-to-end solution that includes our data. It’s very important that AI works with the right data. A lot of time you get into a situation where you have to enter your own data and it doesn’t match, it doesn’t work. It’s not our problem, vendors say, it’s your data’s problem. We don’t do that. We make sure our solution is end-to-end.

The second thing that we have is the full switch. Our solutions don’t just cover one problem. We try and attack the entire lifecycle of the customer - all the way from prospecting and how you get your customers in to how you engage your customers more closely, better and more efficiently, all the way to really understanding your customers.

What are Appier’s plans for 2020?

We have just received $80 million in funding, at the end of November. The idea is that we will use this to improve our products and expand to new markets. Generally, we are constantly looking at new applications of AI. We want to tackle problems where we have advantages. Our advantage is our expertise in AI and our expertise in data.

Building up new products might mean you build something from scrap, or we could end up buying a company like we did recently with Emin (Emotion Intelligence). The reason we bought this company is because it has something that uses AI. In particular, it uses AI to predict if someone is hesitating to buy. Once we know someone is hesitating, we can send them a coupon. These are the kinds of things you can look forward to over the next few years.