July 07, 2018
Battling FOMO and fighting fake news … with robots
By: Philippe Roussiere

We at Accenture Research have been busy developing our Artificial Intelligence capabilities through partnerships with startups, academia and third-party vendors. As part of that work, we successfully adopted 50 new cobots developed by Flint, a Paris-based, award-winning startup.

How does Flint work?

To take a step back, some context: Business researchers regularly suffer from FOMO, or “fear of missing out,” on relevant news. The accumulation of anti-FOMO solutions on PCs or smartphones to monitor news and published reports has become unmanageable. And the fake news fiasco only makes the task more difficult.

Enter AI. Machines can now decipher what is truly relevant for each industry or market analyst. But therein lies a question: Could AI really be trusted to not only do this well enough to inform themselves but their fellow consultants as well? Accenture Research gave it a try.

Flint’s technology relies on machine learning and neural networking—a technology inspired by the human brain (yet a very simple one). AI based on neural networks is trained by a human, so it gets more and more accurate as you train it.

Flint’s AI uses a huge amount of data: millions of articles from thousands of sources, as well as 10 years of data from more than 20,000 reliable expert profiles identified on Twitter. We in turn used this data to model different patterns of behaviors between humans and content. We then designed a "Robots School" as an ongoing and collaborative challenge to help human trainers help robots. The more we get robots trained by smart people, the easier the robots will be to train.

Less than six months into our experiment, 50 Flint virtual personal assistants have improved the daily life of researchers and thought leaders in 20 locations globally. They boost productivity in monitoring specialized news and accelerate the time it takes to review specialized information.

Chart indicates the relevance of the content provided by the cobots and shows how often the content links are clicked on by the user.

Researchers‘ satisfaction rate is high. Click rates to selected articles are high and correlate to how relevant researchers find the news to be.

Alyssa, a researcher in our Frankfurt office, has trained five cobots so far.

“I use my bots both as project assistants, really bringing Human + Machine to life. One of my bots serves a few colleagues focused on gender research in identifying relevant high-quality studies. Another one has helped me stay up to speed on Responsible AI topic, which has helped my high-impact project and microblogging.”

Each trainer creates a personal relationship with his co-bot

How was it to collaborate with Accenture Research?

Said Benoit Raphael, Flint’s Chief Robot officer, "Working with Accenture Research was challenging our young and inexperienced robots with tough experts from all over the world. We were all part of an unpredictable lab project, but with a strong test and learn approach from both sides. So, the first challenge was to prepare researchers to use human-machine collaboration in a practical and fun way, so they could experiment with different ways of training and see what happens. The fascinating part of the project was to see how the AI evolves under this intense training. A few weeks later, the robots freshly adopted by new researchers became faster to train on similar topics."

This is just the beginning. Over time, cobots will evolve into mindful, unbiased colleagues whom our researchers can trust to get rid of FOMO, and thus focus on more value-added tasks.

In the near future, information will find you, not the other way around, but only if you can trust the system to bring you what you need when you need it. We are, in fact, reimagining the processes of business analytics.

Popular Tags

    More blogs on this topic