To AI or not to AI? Define the problem first
The KL leg of Trescon’s World AI Show organised a panel discussion about impact of Artificial Intelligence (AI) and Machine Learning (ML), upon organisations on their transformation journeys.
The most futuristic, Star-Trek type idea came from Zespri’s head of strategy, Keerthi Kumaravelu who painted the possibility of AI making day-to-day decisions for you. That also includes shopping decisions. “If algorithms are what that is going to buy products for you, the nature of marketing changes. That’s what exciting for me,” she said.
But what could be the steps to enable something like this, or enable the digital vision of any organisation in any industry, for that matter?
Head of Business at Tokopedia, Galuh Adifani who moderated the panel, had pointed out that the simplest things can be the biggest challenge to address in an organisation. So in summary, what are the barriers to digital transformation, that panellists see in their respective organisations?
Essilor Group’s regional director of technology, Juliana Chua admitted that the end goal of an organisation is important. Once that is established, smaller, achievable goals can be carved out. “Then execute it, roll it out, then test and learn from it,” she said.
What she described could be said to draw upon principles of the Agile development method, that proposes minimum viable products or smaller achievable goals. She also shared a CI/CD approach, a continuous loop of integration and development that is based upon feedback from customers.
This approach ensures the product is always being improved upon and always being enhanced to be still relevant to customers.
With these useful methods in mind, she also cautioned about the end user requirements gathering process where over-analysis could happen. “That tends to be downfall of any project. It makes the project so big that it never seems to be implementable as well.
“There comes a point in time when you have to say ‘No’ or ‘Stop, let’s push this out to market first to get the immediate reaction and see what is the next best format to proceed with’.”
How do you justify spending money for digital transformation?
Mudah Malaysia’s CEO Gaurav Bhasin said that in terms of challenges when it comes to implementing technologies like machine learning (ML), the biggest is figuring out whether the problem is worth solving via ML or not.
A lot of times, individuals and teams are very excited by machine learning as a buzzword. Everybody is excited about using technology to solve a problem, he observed.
The second problem is related to the fact that the problem may not be related to machine learning, but everything else around it, like data points to collect to train the system, the monitoring infrastructure from an execution perspective, the culture of experimentation and when to kill an experiment while not ‘killing’ it.
These challenges are to do with what makes (a project) a success or failure, and in turn goes into defining how to invest money in these technologies too.
“Ninety percent of the time, people come with solutions versus actually trying to define the problem better and understanding whether this is indeed a solution we should experiment for that problem,” Gaurav said.
AI for optimisation
With car utilisation rates tied to the availability of cars, GoCar’s CEO Alan Cheah, said the solution they come up with must be able to consider utilisation, availability, and also other measurables like how often and for how long cars are ‘offline’ when taken for repairs, and how fast cars can be deployed to locations.
“For us, these are always of immediate concern. Everything we build right now is always geared towards solving these and the turnaround time for those cars.”
He does admit there is an ongoing challenge of finding the right knowledge base for them (to refer to) to build their solutions.
“In Malaysia we have limited (expertise) in terms of skillsets and knowledge base when it comes to machine learning and AI.”
Local talent, a local knowledge base and cost of investment are some of the challenges that the ride-sharing service faces. But these could easily be the challenges that other industries face as well.
There are AI and ML benefits to be had, no doubt. Juliana who is in the visual health industry sees that AI can assist in providing diagnosis and suggesting procedures for patients. Gaurav recognises the benefit of analysing traffic behaviour to his e-commerce website, and opined voice-based search as one exciting future.
But a majority of the panellists also concur that problems have to be understood and defined first, to ascertain the suitability of AI/ML technologies to solve those problems.