“We started out in the travel technology space helping our clients make sense of the large amount of both structured and unstructured data involved in the distribution of hotels,” explains Zumata CEO Josh Ziegler. “While improving our processes, we began leveraging artificial intelligence (AI).”
The problem scope was enormous, Mr. Ziegler observes, when one considers all hotels, amenities, points of interest, concepts and combinations that a user could ask for. Zumata gained access to millions of images that didn’t have tags or labels to describe or identify their content. As a result, retailers sometimes ended up showing images of a toilet or a conference center as primary pictures, when they may have a fantastic location and facilities. “When we saw this missed opportunity we turned to AI specifically image recognition, to tag our images,” Josh says.
Tagging allowed for highly effective use of images, which in turn boosted hotel reservations. Zumata also used AI to determine the quality of images and improve their descriptions, showcase details of hotel facilities, and even make the supply chain more efficient. It is easy for consumers to find out if a hotel has a gym, but whether or not it is appropriately equipped is another matter, reasons Mr. Ziegler. “With image tagging, we can now show pictures of the gym in the search result so you can quickly see if the hotel has a treadmill or free weights. The end result is more sales for our retail partners and a much better user experience for their customers,” he adds.
As Zumata told the story of what they have been doing to help customers find their perfect hotel, banks and insurance companies began approaching them for custom technology application. “As we dug deeper, we discovered our real core competency was the ability to train cognitive systems, and as a result we now working with clients in these and other verticals.” Today, Zumata provides customer experience solutions to companies that wish to help customers find the products or services that meet their needs, as well as companies looking to automate their customer service interactions. “Our AI offerings cut across industries; we are engaging with insurance companies, banks, telcos, retailers, and many others,” says Mr. Ziegler.
Zumata’s commitment to sustained innovation and pushing technical limits has won for them industry awards and recognition. It has been named one of ten most disruptive companies in travel, and has won Digital Innovation Asia Award and the IBM Partner of the Year award. It is also among the first companies to release a natural language API on the IBM Bluemix Catalog.
Training Cognitive Systems
Training cognitive systems calls for a balance between identifying the appropriate use case and supervising the outcomes. Zumata starts by identifying the client or business ‘pain point’, and working to find ways to automate the process to either provide a better user experience or reduce the client cost.
“From there, data is king,” Mr. Ziegler explains. “The more data we have, the better solutions we can provide. This data is used to supervise the training. Driven by our engineers, supervision is a manual process involving code, algorithms, as well as machine and deep learning. It involves a lot of repetition and practice before the solution can be put into our client’s hands.” Even as the supervision continues, as end users throw new examples, scenarios, questions, etc. at the system. While AI systems can learn, Zumata insists on steering this learning to ensure desirable results.
“We’ve seen creativity from AI tools in specific use cases, but we feel that an advantage of an AI system is its ability to provide uniformity to customer experience in terms of both tone and information.
“For example, when we’re training our systems to identify hotel imagery and tag or label the photos, we want to ensure there is a consistent outcome. A picture of a pool, should always be labeled as a pool, and not as water, pond, river, lake, etc.”
When Zumata considers a customer service chatbot solution, it compares the outcomes to human agents. Chatbots have the ability to display consistently a positive attitude, and provide accurate information. Human agents, on the other hand, often have outside influences that impact their mood, their delivery to end users, and may provide the wrong information, Mr. Ziegler highlights. “The tasks where we use AI to perform are creative in the sense that they’re not entirely structured, but rather than uniqueness we strive for consistent experiences.”
Zumata develops new products and services in house. “I am fortunate to have a great team. All our products and services are proprietary, and have been developed in-house.
“We have deep expertise in delivering accommodation inventory to travel retailers. For these clients, we have identified several areas to address based on industry knowledge. We began tagging and labeling images, relating accommodation information to user requests through natural language processing, and providing automated customer service.”
Deep Tech Focus
At its core, Zumata is a tech firm that focuses on deep tech, and uses technology to simplify what is complex for its customers. It began using AI to deliver premium content for hotel bookings and deliver personalized search results. Quality and relevant images alongside personalized results have been able to lift conversion rates by 50 per cent. This has led to the company’s hotel API being adopted by some of the largest travel players in the world.
Mr. Ziegler believes that what they have achieved in travel has started gaining attention from other industries. “Banks, insurance companies, retailers, and others started to reach out to us to help them develop solutions to the challenges they faced, leveraging on our deep expertise in AI, and training cognitive systems. We began developing customized solutions for clients based on their specified needs.”
Take for instance ‘Jiffy Jane’, an insurance chatbot Zumata created for NTUC INCOME. The objective was to allow users a better way to buy travel insurance, so Zumata designed a user-led conversational chatbot that incorporated important, related, and relevant insights such as currency conversion and weather information. Ultimately, through existing historical data combined with active supervised training of how customers wanted to engage with Jiffy Jane, Zumata was able to deliver a solution that answers customer queries, showcases and highlights comparison data, provides relevant tangential information, and facilitates the purchase of travel insurance – a one-stop shop that provides a seamless and quick experience for customers.
“Our AI led products can be applied to various business units within banks, telcos, retailers and insurance companies,” Mr. Ziegler explains. “They can be either customer facing or internal facing. The keys are that they typically revolve around advisory or sales related functions.”
One example that is applicable to any industry is deploying AI-led technologies for customer service. “Advising customers on their enquiries in real-time, through many channels, in any language, and at any time of day, is a dream for most companies. Our data shows that we can address well over 70 per cent of the enquiries for our existing clients. The financial savings from automation are incredible, and what’s often not talked about is how the customer receives a better experience at the same time. Gone are holding times and long chat delays – AI is simply a better customer service option for most enquiries.”
In addition to customer service capabilities – typically post-sale questions – Zumata can also deploy pre-sale advisory services. They run the gamut, from a simple text advisory for products like insurance, wealth advisory, shoes, event marketing, to just about anything else. In more complex forms, these AI-led solutions can incorporate things like image recognition for retailers. If a customer sees a shirt she likes on her favorite celebrity, she can simply take a photo and share it with our chatbot. The chatbot can query the retailer’s inventory to find the best match, and even advise which belts, pants, shoes, or other accessories match the particular items. Online shopping then takes on a completely different experience.
The use cases are wide and varied, Mr. Ziegler points out. Companies are now starting to realize that they have internal customers as well. “We’re seeing a lot of interest for internal applications from large enterprises. The scenarios range from simplifying the process of leave application and expense claims submissions to information depositories for sales or service staff to access information quickly. The cognitive capabilities we have developed can be used in many different ways and applied to any industry.”
Zumata Biz Plan
Zumata began with backing from venture capital firms alongside other angel investors. Today, it derives incomes from two principal sources: For its travel products, Zumata charges a transaction fee for every accommodation booking. For its customized solutions like chatbots, it charges setup/development costs, as well as an ongoing license fees.
Although unable to disclose its financial figures, Mr. Ziegler says that their turnover is expected to grow tenfold in the coming year.
“We remain focused on executing upon our signed contracts, which are with some of the world’s largest travel retailers. This calls for a relentless focus on innovation, reliability, and delivering business grade solutions. We have found delivering top quality products generates more business with existing clients and leads to quality inbound interest from new clients.”
Article was published in http://www.portfoliomagsg.com/article/why-simple-is-best.html