Certis Cisco: understand the problem before applying AI

Chua Chwee Koh, Chief, Group Technology and Operation, Certis says that implementing artificial intelligence (AI) is not easy. If it were, self-driving cars would be mainstream by now, he said at NVIDIA's recent AI Innovation Day in Singapore.

"There are a lot of products out there but sorry – the product itself is not enough," he said. "You and your ops team have to make it happen."

Chua.

The problem is that the technology alone cannot deal with everything that happens in real life, he explained. "You have to look at it from the operational perspective," he said.

Specifically, businesses need to understand their customer and their own operations well to map out the customer journey and identify real problems to solve. "Be prepared to re-engineer and do what I call ops-tech integration," he said.

Both processes and solutions will need to be analysed iteratively and reshaped as required, he added, noting that the journey can be difficult.

"If you want something you may have to modify it," he said. "As you get more data, you can improve the technology itself."

Certis Cisco has that operational experience, and has been working on combining operations with technology for some time. The company safeguards leading airports in Singapore, Sydney and Qatar, and is also the largest closed-circuit television (CCTV) installer in Singapore, Chua said.

The combination of operations and technology, which Certis Cisco calls Security+, is successfully protecting shopping malls such as Jewel in Singapore and can bring manpower deployment down by up to 30%.

Jewel has over 5,000 sensors, more than 700 CCTVs and 12 different systems spanning 135,700 sq m of space. It runs successfully because there is orchestration in place, Chua said. "You need to connect all these sensors and infomation, you need to digitise everything that's at the heart of it," he said.

Operations and technology also means understanding people and how to get the right information effectively, as well as implementing business process re-engineering (BPR), he added. "If you continue to do the things you were doing before, you are not going to get much headway," he said.

"You need people to do operational technology... understand what customers need, what problems they need to solve."

It is important to look at data and modify the technology as needed, as opposed to force-fitting the technology, he stressed.

To scale up, Certis Cisco will need to implement AI, Chua said. The technology could boost video analytics in surveillance so that the company can automatically identify people in wheelchairs or using walking sticks and offer them assistance as an example, Chua said. A human watching CCTV screens is unlikely to process as much video information at once, could well miss those who need help, or make mistakes, he explained.

"Computer vision now is better than human vision in most situations today. When that happens and you bring more and more AI to it, all those tasks that require humans to look at the screen... the AI is going to do much better and the human can focus on more innovative tasks, the more empathetic parts (of the work)," he said, pointing out that people still want the human touch when they need help or want to complain.

This drove the company to launch the Certis Centre for Applied Intelligence in July 2019 and build an ecosystem to realise AI benefits successfully. Some of the company's partners for the centre today include Hitachi Vantara, Microsoft, Cisco, NVIDIA, Asus, and Xjera Labs.

"You need partners, you need academia...customers to bring the data, you can't do AI without the data – we hope that this is the major catalyst for us to take the industry to the next level," he said.

"At the end of day it's important to bring in the customers and the partners who are prepared to say 'Here, you can have my dataset'," Chua elaborated. "At the end of the day we get better solutions."

The company has invested in a GPU farm capable of over 500 teraflops of performance, Chua said, sharing that this is the largest such farm for the security industry in Singapore. A teraflop refers to 1 trillion floating-point operations per second (FLOPS). Many FLOPS are needed for AI operations.

The many applications of AI that the company is working on include identifying different mosquito species, predicting energy consumption, and anomaly detection to uncover safety violations at construction sites.

Certis Cisco also has an auxiliary police robot deployed at Jewel to study how robots might augment or replace the human security guards that Certis Cisco deploys. Looking for operational use cases and incorporating the robot into current operations may sound simple but is not easy, Chua said.

"You need to bring in people with deep operational experience who are familiar with bringing this together to get that AI (benefit), it's not just the technology," he said.
 
The company's key markets are Singapore, Australia and Hong Kong, and Chua observed that Singapore is way ahead in AI readiness, whereas Australia and Hong Kong are somewhat behind.

"In terms of money, Singapore is ready," he said.