“You have to be smart and creative regardless of all obstacles that come in the way” – Charis Sfyrakis, Algomo
As a part of the CAST project, we have been supporting creativity driven innovation in tourism and, more recently, have sought to mobilise a response to the impact of COVID-19 in this space via our creative solutions from CAST hackathon last July. We are now conducting a series of interview-based case studies to examine closely how businesses are coping with the current situation. Our second interview, directed by Creative Business Network, EBN – European Business & Innovation Centre Network & WestBIC is with Charis Sfyrakis from Algomo.
Algomo was the runner up of the CAST hackathon. They decided to apply to the hackathon, because COVID-19 had a major impact on the founder’s holiday accommodation family-run business. More specifically, there was a radical increase in repetitive client requests which were made in a variety of languages. Algomo tested their bot on the website and found that the overall operations load was significantly reduced as a result. They understood that many other property owners could also benefit from this technology and wanted to see if industry experts would agree with that!
“It all started 2 years ago when I used to be the Lead Data Scientist at the team that created HSBC’s global chatbot framework. The ultimate purpose of this framework was to enable the bank to easily roll out chatbots in different markets and languages”, recalls Charis Sfyrakis, Chief Executive Officer. HSBC is a global bank with worldwide customers regularly asking the same type of questions: How do I activate my credit card? Will I be charged for cash withdrawals? How do I transfer money internationally? Etc. It became apparent that a similar technology could also be used in other industries, such as tourism.
The ideal scenario back then was to somehow centralise the business knowledge and then be able to serve the corresponding responses in any language. At that time, despite recent advances in Artificial Intelligence (AI), there was no chatbot solution that could create a single chatbot for queries in all languages. In other words, one had to create a new bot for every language, providing each with its training data and the corresponding responses separately. According to Charis, replicating the process 5, 10, 20 times is not the biggest problem. The real challenge arises when you want to update all those bots at once, and especially when you need it urgently.
In February 2020, Algomo was born with the ambition to revolutionise customer service operations across multiple languages. Based on the most recent breakthroughs in Natural Language Processing (NLP) and more specifically in transfer learning, they created a multilingual chatbot engine, capable of learning from every single human interaction to automatically update its understanding across all languages. There are 2 main benefits of this technology. Firstly, the bot engine can be trained with a relatively low amount of data (1 question, 1 response type) which allows them to build the bot in a very short amount of time. Secondly, a single question pair is enough to make the bot understand similar phrasings of the same type in 16 languages. The underlying mechanism is that semantically similar sentences in different languages are represented with similar vectors (which are used by language models). For instance, “I want some water” in English and “Tengo sed” (I am thirsty) in Spanish are semantically closer compared to “I want some water” and “I want some time” which are both in English and have more common words. It therefore has nothing to do with direct translations.
Coding everything in a language agnostic way allows Algomo to maintain a single bot for all languages. “If you teach the bot something in English, you don’t need to retrain it in French”, explains Charis. It therefore has nothing to do with translation. While Algomo is responsible for centralising the understanding across all languages, it is then up to their clients to define the response types that they want to provide for each idea.
Algomo concentrates on automating customer support across multiple languages, focusing on 4 main markets: international companies (e.g. large banks), companies that aspire to become international, companies that operate in multilingual countries, and companies that are not necessarily international but have international clients (e.g. hospitality). Algomo’s mission and long-standing ambition is to democratise access to information and documentation to everyone, no matter what language they speak. Their initial focus is South-East Asia and Europe as these regions speak a multitude of different languages.
Right now, their product is undergoing beta testing, with Algomo running 2 pilots. The first one is a multinational distributor of industrial supplies. Having presence across all Europe, they want to enable their employees to search for information in any language, without having to maintain each article in all languages. That way, someone in Poland could search in Polish and source documentation about discounts on exhaust pipes from someone in Spain. This has very promising growth perspectives. The second one is with an Arabic neo-bank which is starting their customer support in English and wants to seamlessly roll out their customer support in Arabic as well. Algomo is also in early stage conversations with some travel companies in South-East Asia and has some non-disclosure agreements with some smaller companies such as a hedge fund and a gaming company.
Communication and the Ways of Working is the real Challenge
According to Charis, the biggest challenge posed by the COVID-19 pandemic is not the business itself: “it just happened that we had to operate in this environment, we started with COVID.” The real problem is communication and ways of working, as employees (8 currently) cannot meet in person in the same office space. To remedy the situation, they are heavily investing on capturing documentation through a wiki on ClickUp to document everything they do to facilitate knowledge transfer and onboarding of new hires.
“In terms of financing, I don’t know much about government support schemes nor whether they are applicable to us. When we set up the company, we had a runway of at least 12 to 18 months. So, it’s not like we’re hit by the crisis, as we were not expecting to get any income for at least 12 months which is quite standard for a startup. The issue is not financing, it is about getting access to the market”, clarifies Charis.
Algomo is supported by 3 different accelerators. First, NatWest, an London-based accelerator, provides an amazing office space (that is currently not accessible unfortunately) and gives them credibility and recognition, mentorship and legal support, etc. Secondly, the NVDIA Inception program, a virtual accelerator, grants them access to their computing infrastructures (i.e. GPUs, AWS and TCP Google Cloud Platform). Thirdly, skyincubator, which is based in New York, helps them with fundraising and supports them from the idea phase into revenue generation.
For the time being, Algomo has decided to postpone their fundraising activities. At first, they validated Algomo’s value proposition by interacting with numerous potential customers and leads to understand if they were impacted by the issue of multilingualism and to see the possibility of scaling up internationally. “The major risk that most startups have is creating something that nobody cares about”, highlights Charis. Since the end of last summer, they switched their focus from business development to product development to get more traction and customers, etc. This is because their product is not fully mature, and companies have been hesitant in adopting Algomo despite expressing interest in its large potential. Have a look their demo video!
“One thing that I realised about entrepreneurship is that you have to be smart and creative regardless of all obstacles that come in the way. And you have to sustain a balance between believing in yourself and the long-term vision, while adapting and listening to what is happening outside.”
Algomo believes that their chatbot can help hospitality & tourism businesses quickly resolve their most common queries around COVID-related or other topics across any language, and it is designed in such a way, so that it can be set up in minutes. This addresses the needs of both big OTAs, as well as smaller independent providers. By being able to help their customers with instant resolutions to common requests, businesses can have a 24/7 customer service, across all languages, ensuring that their visitors will get the answers they need, in particular reduce traveller anxiety and uncertainty. This helps businesses become more credible, increase their sales, reduces their churn rates, and also helps businesses understand what their customers are looking for and mark areas for improvement.
To learn more about the company, visit https://www.algomo.com/.