Over the past few weeks, I have been running a simple experiment. There are three ways to search the databases contained in ScoutAsia and while all three methods have their specific uses, I wanted to see which was broadly the most useful. The clear winner -- as I hoped and expected -- is our new AI-driven search, which we are calling "Scout AI".
What I did was the following: I chose a topic, in my case "unicorns", private start-ups that are valued at more than $1 billion. First, I built a Target List. Using our company database of over 760,000 companies across Asia, I researched and found the top 15 unicorns, put them into a Target List and set up a daily News Alert, allowing me to receive any new news articles about these companies by email. I didn't expect this to be perfect since Asia has around 250 unicorns (China alone has over 200 -- see this previous blog post for further analysis). But I assumed that the top 10 or so, including Ant Financial, Bytedance, GoJek, Grab and Didi would attract the bulk of daily news coverage.
Second, I set up a Saved Search across ScoutAsia's 40-plus News Sources, using "unicorn" as the key search term. Again I opted for a daily News Alert by email.
Finally, I created a customised Scout AI search for "unicorn". This is based on semantic search rather than keyword search (as used by Google and most traditional search engines). The difference is that semantic search uses context and association to find relevant articles rather than just robotically searching for one or more keywords. Once your Scout AI is trained you can start to use it...and as you do, it will continue to learn and refine its article selection. For a full explanation, please click here.
Setting up the first two searches took just a few minutes each. Creating and training the Scout AI took a little longer, but while I started the training by selecting around 20 articles myself and labelling each as either relevant or not relevant, ScoutAsia's human curation team in Singapore then took over and continued the training. Together, we have now fed the AI algorithm over 2,000 articles and it is getting pretty good.
Just how good was shown by my daily News Alert email this morning: my Target List produced 6 articles about unicorns; my Saved Search just 4...and my Scout AI, 56 articles. Naturally, it included all of the pieces that the first two methods surfaced. Today, these included articles from the Financial Times, Nikkei Asian Review, Deal Street Asia, KRAsia and others about the Line/Yahoo Japan merger; Alibaba's Hong Kong IPO, expansion moves by Tencent that mentioned Bytedance; and earnings from Asean unicorn Sea, which runs the Shopee e-commerce service.
However, my Scout AI search also called my attention to an article about Amazon's strategy of creating private-label brands in India, which could threaten local rival Flipkart (a unicorn, naturally); news about Singapore digital payment platform Nium expanding into Indonesia; and several pieces about fund raisings by investment firms that have successfully backed previous unicorns. On top of that, there were multiple takes on the day's 'big' news from Line/Yahoo and Alibaba, some of which provided fresh details.
Of course, not every user will want to read, or even scan, more than 50 articles per day -- despite it being a topic they are interested in. But we will be introducing the ability to overlay filters onto Scout AIs, both by geography and by industry or sector, allowing you to really refine your search down to essentials.
What my little experiment proves, though, is that there is a great deal of news out there and while getting to the key information quickly is valuable, the really critical thing is not to miss an important signal in the first place. Our new Scout AI function is designed to solve both of these challenges for you.