ARTIFICIAL INTELLIGENCE: THE TRAVEL INDUSTRY’S SECOND WAVE OF DISRUPTION

Harsha Subasinghe is positioning CodeGen, a travel industry e-commerce business supplying the world’s largest brands, to disrupt the travel industry

For a technology that only recently made it out of the laboratory, artificial intelligence, or AI, is making a splash in the travel industry. The Internet making it possible to book travel online has shuttered many travel agencies, and a new wave of technology led by AI is now threatening a second round of disruption.

For hundreds of years, people have consulted experts about travel and trusted their judgement about destinations, hotels, ocean liners and later, airlines. These days, more and more people turn to the Internet for advice.Industry experts say around 25-30% of those who speak with a travel agent or adviser end up booking. However, online, the conversion or those booking travel following a search is as low as 0.5%. There can be many reasons for this, including fleeting behaviour online at websites to book air tickets, hotel stays, tickets for attractions and other travel-related booking. Often, people are exploring for purchase at another time: office workers are whiling away their time online, not enough information to make a decision or some issue in the booking systems turn people off from completing transactions.

Harsha Subasinghe, founder and chief executive of CodeGen

Online travel agencies, which are replacing the ones on high streets, are also struggling as they often don’t have enough information to understand their audience. In contrast, in a face-to-face interaction, the salesperson is able to relate to the potential buyer and offer personalized advice.Even when an online outfit has some knowledge of the potential buyer exploring the options, they might not know how to make the best use of the data. As a result, they cannot offer the optimal product at that point.

CodeGen, a successful Sri Lankan tech company, is a developer of travel-related software solutions that help solve these problems for global clients ranging from major international airlines, online travel agencies, hotel booking services, theme parks and cruise lines. CodeGen software covers every step from booking to customer support, and some of its products are now using AI technology to improve user experience and convert more browsers to buyers.

For decades, AI was known for promising more than it could deliver. CodeGen, founded in 2000, expects great leaps using the technology on top of what it’s already achieved for its customers.

“If you improve the conversion rate to 2% from 0.5%, that’s a massive achievement,” declares Chief Executive Harsha Subasinghe, who is also the company’s founder. He overlooks the company from a glass-walled office the size of a small apartment. Right next to his office is a brainstorming space, with artificial grass, a flat screen TV, flip boards and other paraphernalia associated with the activity.

CodeGen’s non-disclosure agreements prevent it from naming its main customers. However, Subasinghe says it has travel booking sales of up to $7 billion annually. CodeGen’s revenue is not linked to the value of the booked business on its various travel reservation systems. However, its chief executive suggests the number of customers served has a loose association with revenue, although he declines to provide any figures. CodeGen approaches the conversion issue with a three-pronged strategy, but it’s unclear how many of its customers use the entire suite of products.

First, it tries to discover as much information as possible about potential customers by exploring the site of one of its clients. In real life, it happens when a potential buyer logs on to a site by sharing their name and other basic data. Once online in a CodeGen-built site, software scans public social media pictures and posts to gain an understanding.

They do this with the help of image recognition and natural language processing, two AI techniques designed to analyse and understand what we post. Image recognition enables software to identify a picture and the people in it, situations and text. For example, based on a picture of five people on a beach, a travel site may suggest a beach holiday to a client with a discount voucher for five. Natural language processing is doing the same by helping a computer understand the human language, written or spoken. In this case, travel agencies can take information from hotel reviews across the internet and categorize each hotel by cleanliness, convenience and so on.

Second, it seeks the best available products. They use the newfound personal information for a process called optimization, which searches through all available results, analyses, ranks and refines them based on preferences and interests, to offer an optimal product.

“If people have too many options, they won’t be able to decide. If they have too few, they will look somewhere else. Our job is to find the right algorithms to offer a few well-considered choices.”

Yohan Welikala, Vice President of Research at CodeGen

Third, it helps customers choose. It’s now possible for chatbots to engage a customer in conversation online, without a customer care executive having to sit opposite you. They are meant to simulate human interest by answering enquiries online. They often appear as a pop-up messaging offering assistance. Their responses are mostly taught, but researchers are working on more human-like conversations. “I personally think that, in a few years, websites will be dead. People will go and talk to a chatbot,” predicts Subasinghe.

Finally, in case real-time advice and recommendations are not enough, CodeGen is making predictions by analysing large booking engines and understanding their effect on the market and prices. Using AI technology, they are trying to predict what happens next, as an accurate prediction would benefit the customer. Factors influencing people’s behaviour include weather, cultural events, war and so on. “Those things happen randomly; no one can predict when it happens, but we try to understand what it does to the market and how it changes people’s decisions. We are building a map of where people are travelling to based on their bookings, and trying to see the trends,” says Vice President of Research at CodeGen Yohan Welikala.

Researchers at CodeGen are building this technology into their products. Their most recent research is focused on ontologies. An ontology is like a map of real-world objects and situations written in a way a computer can understand. It can see the context and relations, which makes it possible to find answers to general questions. For example, asking Google what the weather is like. It will find a weather forecast, identify what day it is and answer the question. CodeGen uses this system in their travel software to answer specific enquiries about their clients’ products. Their ultimate goal is to create a dynamic ontology: one that would learn and update itself.

In fact, self-learning technologies or machine learning is the major challenge in AI. “We’re at the verge of a breakthrough. If we tackle that, we’re one step closer to building the general intelligence computer,” says Yohan excitedly about  advances in AI happening around the world. Much has changed since Yohan Welikala joined the company as an intern 15 years ago. CodeGen has grown to be the biggest investor in AI research in the country. Its chief executive Subasinghe says up to 40% of revenue is invested in research some years.

“We embrace these challenges. It drives us to work on new areas,“ says Welikala with a spark in his eyes. However, deploying AI in a technology laggard is a gradual process. According to Welikala, some of the greatest challenges come with internal change management. When computers became mainstream, people took time to get used to them. Now, they are being asked to let computers think for them. “They’re not always comfortable. There is a lot of resistance.” In this case, CodeGen shows them how the technology works and what it can do.

“At first, we run the programme side by side to see if it’s doing a good job. Sometimes, it does a better job and the employee feels insecure.” Most often, people are worried for their jobs. CodeGen’s office at the heart of Colombo’s tech hub is buzzing with young people of all backgrounds, engrossed in conversations or scanning their smartphones. When graduates join CodeGen, there is a lot they have to learn to bridge the gap between theory and practise. “It can be difficult for them to understand, to put themselves in the consumer’s shoes and to bring the theory to life. We encourage them to travel, see new trends and devices. Once they come back, they’re more innovative and come up with their own ideas.”

Code-Gen-2

Vegetables grown on a vertical structure on CodeGen’s rooftop are monitored by computers to figure out optimum growth cycles

However, once the initial excitement fades, many start looking overseas for opportunities. This poses a challenge for companies like CodeGen that invest years into training their employees, especially in AI, but also for the country. “That’s why I’m trying to change the mindset of these guys. They were educated in this country and they have to pay something back before they leave. They can rescue the economy of Sri Lanka if they stay,” says Subasinghe.

Out of 400 employees, 40 work in research and development, the unit headed by Welikala. They spend most of their time working on CodeGen’s travel products.

In the middle of a cramped, messy garage; wire, nails and early designs scattered all over is the Vega project, CodeGen’s first and most popular side affair. The first Sri Lankan all-electric supercar, which aims to reach 0 to 100 km/h in less than 3 seconds, is being designed and built here, across the driveway from CodeGen’s main office.

On its rooftop is a greenhouse of vegetables grown on a vertical structure, like a high-rise garden. Without soil, computer-regulated water and fertilizer feed the vegetable patch, which grows tomatoes, cucumbers, beans and lettuce. The plants are monitored by computers to figure the optimum growth cycles by ensuring the right amount of water, nutrients and humidity. One day, Subasinghe wonders if this experiment could be a solution for the dry, unfruitful land in some parts of the world.

CodeGen’s interest in autonomous cars and drones seems as matter-of-fact as the velvety pool table in the lobby.

To Harsha Subasinghe, “it is again an optimisation problem that you need AI for”—as if every problem in the world could be solved with good mathematics, smart leadership and bold ideas. To a man who long surpassed his youthful dreams while running a gigantic business, nothing seems to be impossible.