Customers in the Middle East are becoming increasingly familiar with being greeted by friendly chatbots — virtual helpers that are available day or night for all kinds of burning questions. From Dubai’s sprawling malls to Cairo’s bustling hospitals, Arabic-speaking chatbots are streamlining the customer experience while offering lucrative growth opportunities to businesses that adopt them.
AI-powered chatbots provide 24/7 support in Arabic while saving time and reducing operational costs, allowing companies to optimise their use of talent and resources. They’re not just cost-savers but revenue boosters: Chatbots are reliable digital assistants that guide customers not only to vital information but also can introduce them to new products and services that they didn’t even know that they wanted. The benefits go far beyond the initial interaction: Detailed customer data provides valuable insights in real time that can take your business to new heights.
While some linguistic challenges related to the complexity of the Arabic language initially delayed the deployment of chatbots in the Middle East, technological advances in the field of machine learning have introduced a bevy of new automated chat services that use natural language processing (NLP) to engage customers in organic conversations in local dialects like Egyptian, Emirati, and Saudi Arabic.
Chatbots cut costs for businesses
Studies show that the technology can cut global business costs by $1.3 trillion per year by reducing the resources spent on customer-service agents answering common questions. The front-line support of chatbot technology empowers the remaining customer-service teams by letting team resources be distributed more efficiently and dedicate themselves to meaningful tasks, rather than wasting time answering the same routine questions. They expand a company’s reach, allowing customers to interact any time and from anywhere in the world, overcoming the limitations of traditional customer service hotlines based in a specific time zone.
AI-powered chatbots ask the most important questions that can help businesses identify quality leads. For existing customers, chatbots are powerful tools for cross-selling and upselling, using customer data to make highly personalised recommendations to customers by anticipating their needs and identifying unexplored revenue opportunities.
There is a range of AI-powered Arabic chatbot services available across the Middle East, with notable examples being Arabot, Botpress, Widebot, and Botter. “We use the Botter chatbot for our labs and scan centres, and now we’re able to serve more patients than we used to when we didn’t have an Arabic chatbot,” says the CTO of a leading consumer healthcare group in the Middle East. “Our customers are now able to quickly and easily retrieve their lab tests, enquire about lab test pricing, and request home visits. It’s just a matter of a few clicks. … Our company offloaded our contact-centre agents by 70% and reached out to a wider range of customers by deploying our Arabic chatbot over different channels such as WhatsApp, Facebook Messenger, and our website. Soon, we plan to use it on our Android and iOS app to broaden our reach.”
Chatbot benefits for customers
For customers, chatbots can offer a wide range of benefits that increase their satisfaction. Customers can ask questions and access information and services long after brick-and-mortar businesses have closed for the night. The industry publication Insider Intelligence predicts that retail sales taking place via chatbots will increase from US$2.8 billion in 2019 to $142 billion by 2024.
AI helps chatbots eliminate redundant questions and recognise when a conversation needs to be passed along to a human agent, saving time and reducing frustration for users. A study by Accenture found that 91% of consumers are more likely to buy from businesses that offer a personalised experience, and chatbots can use customer data to deliver customised content, information, and offers that anticipate their interests and needs based on conversation history and contextual data.
Chatbots use two different data set models:
- retrieval-based models, which use heuristics to select preset responses to common queries.
- generative-based models, which also use deep learning and recurrent neural networks to formulate responses while taking into account the length of the conversation and the data being processed.
Chatbots are not just limited to companies’ apps and websites, with many Middle East businesses choosing to integrate chatbot services into popular instant messaging platforms like Telegram, WhatsApp, Skype, and Messenger to process multimedia input like voice, images, and files in addition to Arabic text. Irrelevance detection models know when to pass a conversation along to a human agent, and entity-detection models let users speaking informal Arabic be understood more often.
Developments in recent years have allowed machine learning to take over the tokenisation process (that is, splitting a text into smaller units) to train the AI without human intervention, resulting in the major breakthroughs for Arabic chatbot developers reaching out to a diverse linguistic audience across the Middle East and North Africa.
From retail to e-government: potential applications
Today, chatbots are being introduced in a wide range of industries in the Middle East such as retail, real estate, banking, hospitality, government, and education. Financial services, health, and insurance industries are key areas where chatbot deployment is expected to grow in the region over the next few years.
For the insurance industry, chatbots are a powerful tool that can significantly reduce costs for providers. When users file claims or submit a First Notice of Loss (FNOL), chatbots facilitate processing the claim by requesting documents and photos, using machine learning image recognition technology to verify evidence of damage in context, and communicate information to the user regarding reimbursements. Middle East-based insurance firms like Qatar Insurance Company and Oman Insurance Company have adopted the technology in 2021, extending their service to platforms like WhatsApp for greater customer reach.
Banks on forefront of chatbot tech
Juniper Research projects that operational cost savings from for the banking industry will reach US$7.3 billion by next year, 30 times higher than projected savings in 2019. Users are no longer restricted to the limited opening hours of their local bank and use of technology in the financial services industry can save up to four minutes of time per enquiry — saving banks US$0.50 to US$0.70 per interaction. Financial institutions like Emirates NBD, Liv, Commercial Bank of Dubai, and Mashreq currently use the service, offering insights to customers about their spending in different shopping categories, reminders about payment deadlines, information about balances and interest earned on savings, and tailored recommendations on savings management.
The technology has many applications in the public sphere, particularly in the Gulf region. With a regionwide push toward smart cities, digitalisation, and e-government, chatbots can be a valuable resource for residents as they go about their daily lives, from paying bills to registering for a driver’s licence. The UAE leads the Gulf region in the implementation of AI-powered automation for public services, offering public services related to business licensing, visas, and transport via the Rashid chatbot, named after the late Prime Minister Sheikh Rashid bin Saeed Al Maktoum. The COVID-19 pandemic has highlighted the power of mass messaging for public safety: in emergencies, chatbots offer round-the-clock information and support to citizens, alleviating the strain on responders and allowing representatives to focus on solving complex problems to keep citizens safe.
Arabic dialects and structure pose a challenge
One of the greatest challenges that chatbot developers face is a diverse array of more than 30 recognised Arabic dialects. There are more than 60 ways that Arabic speakers can express the word ‘want’, posing a unique challenge to linguists and computer scientists collaborating on NLP project development. Developers grapple with morphological ambiguity, when one word has many meanings, and syntactic ambiguity, when a sentence has more than one possible structure.
“The most challenging thing about Arabic chatbots is the synonyms,” says Youmna Yasser at Botter, an enterprise chatbot company headquartered in Egypt. “The Arabic language is the hardest language to learn on Earth. It has 12.3 million words, and with this huge number comes the challenge. One word can be written and said in multiple ways according to the geographical culture, education level, and heritage.”
The second-most-challenging thing about Arabic chatbots is the dialects, Yasser says. “One word in the Egyptian dialect can mean the contrary in the Saudi dialect, like ‘salam’. In Egyptian, it means ‘bye’, while in the Saudi dialect, it means ‘hello’. We even have the same issue within the one country: The northern Egyptian northern dialect is totally different from the southern and western ones.”
How chatbots are trained
With the help of AI trainers trained in customer data integration (CDI), Botter built customised and business-centric models which classify users’ inputs and get continually refined over time. “While making sure that the training and evaluation process is a continuous one, we enhance the models on the go with real-time inputs,” Yasser says. “We templated these business models and kept reusing them with other customers within the same business scope. The second thing we developed are models for smalltalk and chitchat, the small conversations that take place aside from the business goal of the chatbot. These models we built can be easily customised based on the business requirements and deal with these conversations in different dialects with the help of the powerful speech and text understanding tools.”
Some businesses prioritise dialect-specific Arabic chatbots when introducing the service to a specific region: Abu Dhabi Islamic Bank, for example, introduced chat banking in 2020 that can process Emirati Arabic. But Arabic speakers from other countries must submit inquiries in classical Arabic.
Although the development of Arabic chatbots were challenged by the relative scarcity of resources needed to train the learning model, the recent emergence of many Arabic virtual assistants offers an optimism outlook for the development of chatbots that are even more responsive, accurate, and conversational as the technology continues to evolve.
