Global X Research TeamWe interviewed Pedro Uria-Recio, Vice President and Head of Axiata Analytics at Axiata Group. Axiata Group is a leading telecommunications and digital conglomerate with approximately 350 million customers in 11 countries across South and South-East Asia.
Pedro is a senior marketing leader with experience in data analytics, artificial intelligence (AI), product marketing and profit and loss (P&L) management. Pedro leads advanced analytics and AI across the operational companies of Axiata Group. With a cross-functional team of 50 marketing specialists, data scientists and engineers as well as digital professionals, Pedro focuses on marketing analytics, digital growth hacking and new revenue streams. He was previously a senior consultant at McKinsey & Company, and he has had roles at McGraw-Hill, Veolia Water China and Orange France. Pedro holds an MBA from the University of Chicago Booth School of Business. In this interview Pedro shares his personal views, which do not necessarily reflect those of his affiliations.
What are the most useful applications of AI in the business world today? And which do you believe are most promising for future use?
AI is the most general-purpose technology of our time. New products and processes are being developed thanks to better vision systems, speech recognition technologies or recommendation engines based on AI.
There are five main groups of business applications in AI, which correspond to the five kinds of tasks, typically associated with humans, that AI can perform. These business applications, from the least to the most sophisticated, are:
- Robotic Process Automation (RPA) is a technology that can automate rule-based, repetitive, high volume processes. Many back- and front-office processes fit this description. Think of contact centers, where customer files need constant updating of address changes or service additions. Or banks, where a lost credit card means updating customer records across multiple systems and keeping customers informed during the process. Automating processes this way, RPA can potentially cut operating costs by 20-80%.
- Machine vision enables capturing and understanding images and video, as well as other more complex datasets such as biometrics. Self-driving cars that perceive the road, traffic signals and other vehicles is a very clear application. Machine vision can also be applied to handwriting recognition; to medical image diagnostics and cancer detection; and to identification of faulty products in a production line, among other applications.
- Language Processing handles language interactions with employees and clients (e.g. voice, chat, email). For example, a customer service bot can independently handle about 80% of queries, allowing customer service representatives to focus on the other 20% of tasks that require empathy or better contextual understanding. Customer service representatives can focus on higher-value activities that are often more rewarding.
- Advanced Analytics is a support tool that enhances decision-making through predictions and analyses. Advanced Analytics can be applied to every function of the enterprise: In marketing, companies can target digital advertising to each individual customer or predict the products each customer is most likely to buy. In operations, infrastructure companies such as telecom operators, retailers or banks can analyze how to optimize their capital expenditure in very granular ways. In human resources, key attributes of leaders and managers can be assessed to better understand behaviors, develop career paths and plan successions. In back-office functions, machine learning can help detect fraud in real-time in credit card transactions or insurance claims.
- Cognitive Agents are an autonomous intelligent workforce that can interact, execute, analyze and learn like human workers. This latest application is far less mature than the previous ones, but there are some commercial products already, like Amelia from IPSoft.
As we go down the above list of business applications from least to most sophisticated:
- Applications are also less mature and less commercially ready to implement.
- Additionally, costs and time to implement are higher and therefore short-term return on investment (ROI) to the business are lower because benefits are more qualitative.
- However, applications have a higher long-term potential for future use.
Which industries do you expect to see more rapid vs. slower uptake of AI? And how might adoption vary across countries or geographic regions?
There is not a single industry that cannot benefit from AI. Let me choose, for illustration purposes, an example that is considered a highly qualitative and humanistic industry: marriage counselling. A well-designed questionnaire evaluated by Advanced Analytics, coupled with Machine Vision, to infer human feelings from face gestures and Language Processing tools to identify sentiments could help a human marriage counselor to give much better advice to his or her customers.
The degree to which an industry can leverage AI depends largely on the amount of data available. Development of AI use cases requires a massive amount of data, and consumer service industries often find it much easier to get data than other industries, simply because the world has billions of consumers. According to McKinsey, the industries in which AI can have the highest impact as a percentage of revenues are (from higher to lower impact): insurance, high-tech, pharma, travel, semiconductors, telecom and banking. Except for semiconductors, all are consumer service industries.2
Regarding geographies, today there are just a few major hubs for AI development: Silicon Valley and New York in the US, which have pioneered many AI applications; Shenzhen and Beijing in China; Bangalore, fueled by availability of talent, is emerging in India; while the AI scene in South East Asia is led by Singapore with strong government support for entrepreneurs.
Among all these hubs, China is becoming the uncontestable global leader in AI as its AI ecosystem is accelerated by a strong technology industry, a mobile-first population and a relaxed approach to data protection.
Just to give you a few numbers: 48% of all AI startup funding in 2017 was invested in China Vs 39% in the US.3 Moreover, between 2013 and 2016, the number of AI related patents in China has been higher than in the US, and it was only in 2017 that the US was able to catch up with China.
What are the greatest challenges or obstacles to further business use of AI?
My answer is going to be focused on emerging markets, which are what I know best. Companies in emerging markets face many challenges deploying AI and Analytics. But these challenges mainly fall under two categories: building data assets and, most importantly, using those data assets to transform the business.
Regarding the first kind of challenges, IT investment in emerging markets has been recently focused on more traditional assets like Enterprise Resource Planning (ERP) with very limited budgets for AI. Multiple industries lack the adequate technological systems to track the operational data flows required by AI programs to make decisions and trigger actions. Even if they have an acceptable data capture set up, many organizations lack the right infrastructure to store data, aggregate it into actionable forms, and make it available to users or machines for decision making.
Regarding the second kind of challenges, even companies in industries at the vanguard of AI adoption struggle to find a programmatic approach to using their data assets to transform the entire enterprise. In many companies, data remains in silos with split ownership. In others, huge data sets are collected but never analyzed. A programmatic approach to monetize AI needs to identify clear use cases where AI or analytics can help solving real business problems; to empower employees and managers to use these systems and tools; to integrate data and AI with operational workflows; and finally, to establish an open culture that embraces making decision through data-driven experimentation.
Good data, both in terms of quality and amount, is generally a pre-condition for successful implementation of AI. What are some key lessons or tips for organizations seeking to maximize the utility of their data with AI?
There are mainly three kinds of requirements to have large amounts of good quality data: systems, governance and a data acquisition strategy.
Regarding systems, companies need to have in place the right technological systems that I describe above to track data.
Regarding governance, companies need to have the right processes to manage their data properly and consistently as an asset, ranging from managing data quality to handling access control or defining the architecture of the data in a standardized way.
I am going to focus on the third group: the data acquisition strategy. A controversy from early January 2019 about the Facebook 10-year challenge illustrates the hacks companies are putting in place to get properly labelled data.
To participate in the Facebook 10-year challenge, you just needed to upload a photo of yourself in 2019 and another photo of 2009, 10 years ago. Hundreds of celebrities participated, as well as millions of other Facebook users.
Soon users started wondering whether Facebook had organized this campaign to get high-quality, properly labelled data for an age progression algorithm. This kind of algorithm can predict how people would look like in 10 years’ time. Facebook officially denied having organized this campaign, but given the sensitivities over data privacy, it is unsurprising that many people were concerned.
The issue is that Facebook might have started this campaign to get large amounts of properly tagged data. Facebook certainly has a massive amount and history of photos, but it is difficult for Facebook to know the dates of the photos exactly because users might not always post them in chronological order or because the dates in the metadata of the images might be wrong. Additionally, people often post low-quality profile photos or photos of their pets. With this challenge Facebook could make sure to have good quality and properly labelled photos. Another technique Facebook has used in the past to increase their amount of properly tagged data was allowing users to tag others.
Companies use these kinds of hacks more and more to get good data. I am not saying that these techniques are necessarily detrimental to users. Not at all. But users should be aware of them and pay attention to privacy considerations.
What is the role of regulation in ensuring proper use of AI, and how does that affect prospects for business and consumer use?
Regulation plays a very important role in AI. Society has regulated land ownership for thousands of years, and we know exactly what it means. We also understand property of machinery, and most people understand intellectual property as well. But society does not know what data ownership really means.
What can third parties do with one’s data? Do companies who use AI on personal data have an obligation to explain how their machines arrive at recommendations that bear on the public interest or personal well-being (such as medical diagnoses)? Should we demand higher standards of ethics from AI than from humans? As a society, we don’t know the answers to all these questions.
Governments and regulators should steer public debate towards them, however, to ensure that AI contributes to inclusive business growth and constructive social outcomes. Business leaders and civil society also have valuable input into the questions AI will oblige us to make.
Moreover, data protection regulations, such as the European GDPR (General Data Protection Regulation) define requirements for corporate data governance. Requirements vary from regulation to regulation but are centered in three main areas:
- Customers’ rights: including consumers’ consent to allow companies to collect their data; their right to request access to their data or to correct errors; or their right to have their data erased or transferred to another competing business.
- Obligations of the companies handling data: such as informing customers of the purpose for collecting, using, or disclosing personal data; making reasonable efforts to collect accurate data; arranging reasonable security measures to prevent unauthorized access; and informing supervisory authorities and customers of privacy breaches promptly.
- Restrictions to the companies: such as not keeping personal data beyond a certain period (after which it must be deleted); or not transferring data outside of national or corporate limits.
Additionally, governments can play a critical role to deliver the benefits to society by:
- Making government data available to businesses in machine-readable formats and deploying AI in government use cases such as detecting tax fraud.
- Developing AI talent and making long-term investments in education, including continuing programs to help mid-career workers to keep pace with the digital economy.
- Supporting the development of AI hubs as epicenters of talent, entrepreneurship, development and commercialization.
What areas of AI are you most excited about for the next 3-5 years and beyond?
By far the business area of AI that I am most excited is driverless cars. Driverless cars have the potential to change so many industries, like for example:
- Automotive: driverless cars should be a catalyst to replace old cars, resulting in growing demand.
- Insurance: demand for insurance should decrease due to fewer crashes and fewer drivers.
- Trucking and taxi: self-driving fleets should reduce cost and improve route efficiency but could generate unemployment.
- Ride-hailing: ride-hailing companies should not have to pay drivers anymore but would likely have to shoulder the burden of owning the cars.
- Airlines: cars should become more convenient than planes for shorter overnight trips while passengers sleep comfortably inside the car.
- Telecom: driverless cars should require a dense high-speed network, which is an opportunity for telecom operators.
- Public transportation: driverless cars should be able to service out-of-the-way locations easier than before.
- Hotels: demand for single-night stays at roadside motels should decrease since people will travel more overnight.
- Parking and garages: cities should allocate less land to parking since cars can drive alone back to the garage.
- Real estate: faster and easier commutes should increase the value of property in suburban areas.
- Brick & mortar retail: prime locations should become less relevant due to easier commutes.
- Urban planning: most modern cities are built to cater for the needs of cars. Cities of the future should be designed in totally new ways.
- Fast food: passengers just set the destination and the car selects an optimized route. This should reduce the chance of detour for an impulse food purchase.
- Deliveries: delivering food or groceries home should be easier than ever.
- Energy: demand for gasoline should decrease since most driverless cars will be electric, but demand for electricity should soar.
- Media & Entertainment: passengers should consume more entertainment because nobody will be driving.
- Driving schools: this business should disappear.
- Traffic enforcement: it should be unnecessary.
- Interior design for cars: a new industry should flourish to make cars more habitable and more comfortable to sleep in.
Driverless cars are a technology that is going to change the world as we know it. This is what makes it so exciting!
Related ETFs
AIQ: The Global X Future Analytics Tech ETF (AIQ) seeks to invest in companies that potentially stand to benefit from the further development and utilization of artificial intelligence (AI) technology in their products and services, as well as in companies that provide hardware facilitating the use of AI for the analysis of big data.
BOTZ: The Global X Robotics & Artificial Intelligence ETF (BOTZ) seeks to invest in companies that potentially stand to benefit from increased adoption and utilization of robotics and artificial intelligence (AI), including those involved with industrial robotics and automation, non-industrial robots, and autonomous vehicles.
DRIV: The Global X Autonomous & Electric Vehicles ETF (DRIV) seeks to invest in companies involved in the development of autonomous vehicle technology, electric vehicles (“EVs”), and EV components and materials. This includes companies involved in the development of autonomous vehicle software and hardware, as well as companies that produce EVs, EV components such as lithium batteries, and critical EV materials such as lithium and cobalt.
For current fund holdings, please select the fund ticker above.