BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
Indian IT services firm Wipro believes it is "very critical" today for all enterprises to move beyond traditional AI and looking at historical data. Instead, Wipro said, it's time for companies to embrace advanced AI capabilities, which means ensuring an AI system can recognize emerging patterns.
With AI permeating all aspects of technology, building an AI practice has become a logical step for channel partners. Over a year ago, Wipro developed what it calls the STRL -- or sense, think, respond and learn -- framework. The STRL framework aims to help customers use AI across their entire data and analytics value chain to identify patterns, make predictions and shape the outcomes of what they're looking at, as well as understand customer behavior, said Manoj Madhusudhanan, Wipro's global head of cognitive technologies. The framework also seeks to help organizations understand the behavior of their end customer.
Taking AI to the next level
Initially, Wipro focused on applying AI from an automation standpoint. The company sought to optimize customers' existing processes to boost operational efficiency and create cost savings, Madhusudhanan said.
"While cost savings are important, applying AI as a revenue enhancement was an important ask from our customers. This made us apply niche and emerging AI concepts" like automated machine learning, meta-learning, Graph neural networks and others to take AI to the next level, he said.
"We strongly believe that we should not just be focusing on research around AI," Madhusudhanan said. The STRL framework concentrates on looking at patterns as well as concepts, he said. In contrast, the traditional analytics approach is to have a data scientist look at the data more from a patterns-only perspective. However, because new patterns continue to emerge, the machine must constantly evolve, and that can be done by applying new lenses to the data, he said.
For example, a machine might be trained to play chess but not to understand the game as a concept. "In this case, the machine will never ever know that it is playing a game known as chess," Madhusudhanan said. "How do you make a machine understand a concept rather than training it with all the patterns?"
So Wipro has opted to focus the STRL framework on three areas: outcome shaping, behavior inference and design for behavior change. Wipro is applying the framework in a number of vertical markets, he said.
Outcome shaping, for instance, is being used to help a large U.S. retailer predict its annual revenue forecast with 99% accuracy, according to Madhusudhanan. The framework "is helping the customer improve their COGS [cost of goods sold]," he said.
The STRL framework is also helping banks to identify fraud rings from data that on the surface appears to be normal, he noted. "Our Intelligent Fraud Prevention IP uses concepts like behavior inference to identify emerging frauds and fraud rings," Madhusudhanan said.
In this case, Wipro uses Bayesian network techniques, a popular statistical method for making predictions, because "it's not enough to just tell a customer, 'This is a potential fraud ring.' … We have to explain why the algorithm thinks that. So we combined a neural network with Bayesian techniques to come up with a Bayesian neural network and actually give an explanation why an algorithm feels that way," he said.
In Gartner's Magic Quadrant
Wipro's investments in analytics helped it land Challenger status in Gartner's Magic Quadrant for Data and Analytics Service Providers. "As a key player in the analytics space, Wipro has accelerated its investments through multiple acquisitions over the last two years," said Jorgen Heizenberg, senior research director at Gartner, in a statement.
Manoj MadhusudhananGlobal head of cognitive technologies at Wipro
Wipro's analytics assets include more than 110 apps and solutions across different vertical industries. The firm also has a number of other AI and analytics offerings, including a Data Discovery Platform, a business intelligence transformation framework and deep learning, he said. Additionally, Wipro offers an open-source AI and cognitive computing platform for automation, called Holmes.
Heizenberg also cited Wipro's diverse partner ecosystem and the fact that the company "has strengthened its collaborative ecosystem through partnerships with academia, investments in startups, and alliances with [data and analytics] vendors."
Gartner also gave Wipro high marks for its technology business consulting. "Reference customers scored it highly on technology expertise, for providing quality solutions, and for its AI and data science skills," Heizenberg said. "They value its guidance on [data and analytics]-related strategy as well as its industry expertise.''
At the same time, Heizenberg noted that there is room for improvement. While "reference customers praised Wipro for its technical capabilities and industry expertise, they suggested that there is room for improvement in business consulting."
Some reference customers also informed Gartner that Wipro is "too process oriented." A few raised concerns over inconsistent project resourcing, which caused delays, Heizenberg said.