Preparing for the advent of enterprise AI

Source: Appier. Tu.

The time is ripe for artificial intelligence (AI) to be harnessed, says Appier, which aims to provide AI platforms to help enterprises solve their most challenging business problems.

According to Magic Tu, VP of Product Management, Appier, AI is now a commercial reality through the availability of enough data - 40ZB of it according to IDC - powerful computational hardware such as graphics processing units (GPUs), advanced algorithms and a community with the right mindset.

While Tu said some sources have hyped what AI can do, the technology definitely capable of helping companies to discover new insights, analyse data, predict outcomes, act on intelligence, learn from data and optimise various business processes.

For example, banks starting to migrate from using traditional statistical models to assess personal credit to using machine-learning models, Tu shared. In healthcare, AI applications are used to optimise scheduling and aid diagnoses. E-commerce firms are starting to predict what the customer wants, with as much as 20% of Amazon's revenue coming from the personalised experience, he said.

Given these success stories, companies in the region are interested to begin implementing AI in-house, but are unsure of how to do so. "(Companies) know what AI is but ask 'what can I do with it in my company?'" he said.

There are several alternatives available to companies looking to implement AI, Tu said. If companies are aware of a clear, quantifiable problem they would like to solve, they can look for existing data related to the problem or consider generating the data, then implement an AI-based system to solve it, he said. Companies can also identify problems by reviewing existing business processes, before gathering the data and applying an AI-powered solution.

"Do they have a problem they want to tackle? Is this problem related to data?" he asked.

Either way, it is essential to have the right data and the right data scientists for the job. Many companies would would like to know more about their audience, for example, but do not have a consolidated repository of customer data, Tu said.

At the same time, a "generic" data scientist will not do. "Different problems require data scientists with different skills," he explained. "Do you need natural language understanding skills, computer vision or machine learning skills? Most companies don't even know where to start."

Once the data and the right skills are in place, discovering correlations within the data is key, Tu said. For example, Appier's cross-screen AI technology can look at device behaviour and determine which devices are actually being used by the same person, he said.

If these prerequisites are too challenging for companies, they can also look for an AI specialist which will review business processes together with them to discover problems, see what data is available around the problem, and then provide recommendations on how they can solve the problem, Tu said. Appier's AI-as-a-service can do this in one to two weeks, he explained.

"We (examine) their business processes so we can find potential problems and solve them," he said. "They don't have to hire anyone."

For companies which want to know more about their audiences, as the Commonwealth Publishing Group in Taiwan did, Appier technology can link formerly-anonymous data points into a holistic customer view, Tu said. This allowed the publisher to write content that resonates with their readers and gain more page views while increasing its online subscription base, he explained.

Prior to using Appier's Aixon AI platform, the Commonwealth Publishing Group used humans to predict what readers would like to see and adjusted their editorial scope and advertisement placement by trial and error. Through learning more about its customers via Aixon the publisher has been able to write more appropriate content, match specific advertisements with content more precisely, and gain more subscribers. It now has 4x its previous online subscription audience, Tu said.

Appier offers two product lines under AI-as-a-service, marketing intelligence and enterprise data intelligence, built on real-world audience data, as well as machine learning models built around customer experience, audience behaviour, and predictions that can be applied to different industries.

"We view ourselves as an AI service provider," Tu said. "Our vision is to make AI easy for every industry based on our core strengths. We have audience data and we have an AI engine. We're going to approach other industries to find real problems and resolve their challenges."