For Business Applications, Artificial Intelligence Needs Intelligent People
- By Richard Seeley
- January 31, 2018
Popular consumer products that rely on artificial intelligence (AI), such as the Roomba vacuum cleaner, work with the touch of a button, but business applications of AI are not plug and play, says Pedram Abrari, CTO of Pramata Corp.
If you're searching for data that will aid business processes, you can't just turn on an AI application and expect it to "find a needle in a haystack," cautions the software industry veteran with a BS in Computer Engineering from UCLA, and a Master's in Computer Science from USC.
Headquartered in Brisbane, Calif., Pramata uses AI to provide Fortune 500 companies with insights into their businesses to help reduce revenue leakage and identify regulatory issues. The 10-year-old company's customers include Comcast Business, CenturyLink, Callaway Golf and Allergan.
AI applications, including expert systems, business rules engines, machine learning, data mining and predictability analytics, are deployed in the service of nitty-gritty business processes.
For example, revenue leakage refers to financial losses resulting from providing customers with services for which you are undercharging or perhaps not getting paid at all. Or it can result from charging subscription rates that are lower than what you're entitled to contractually.
"What can you learn from what you don't know? That's where machine learning and data mining come into play because they go to places where it's difficult for humans to go. That's because we can't digest that much information and we can't find patterns in unstructured data quite as well as machines can."
Pedram Abrari, CTO of Pramata Corp.
AI applications are ideal for combing through documents and finding pertinent information on revenue sources and regulatory obligations that can impact the bottom line.
Using data mining and machine learning, a business can begin to find out what business opportunities and risks are hidden in unstructured data.
"What can you learn from what you don't know?" asked Abrari. "That's where machine learning and data mining come into play because they go to places where it's difficult for humans to go. That's because we can't digest that much information and we can't find patterns in unstructured data quite as well as machines can. A lot of what Pramata does in AI revolves around extracting key pieces of information out of contracts and business systems."
The data collected and organized can provide a business with insights into problems like revenue leakage, which is a primary focus for Pramata, the CTO said.
But despite what we see in scifi movies, machine learning actually requires humans with sophisticated levels of expertise, Abrari explained.
"The challenge of machine learning has historically been that people assume it's this magic thing that magically finds opportunities for your business," he said.
Machine learning is gaining traction, he added, because it can uncover patterns in data and provide valuable insights. This can lead to awareness of business conditions that are not readily apparent to humans, but can help executives and managers understand things that could have a major impact on their business.
"However, to gain those insights you just can't go looking for a needle in a haystack," Abrari explained. "It has to be a controlled search. You have to be able to apply the technology in a controlled fashion. To some extent you have to have a sense of what you are looking for. It can't just be a random search for anything. You've got to be targeting something that you want to optimize."
Determining what to optimize and how to target the machine learning technology to find the data and insights needed is a job for data scientists, who need to train and retrain machine learning models. With refinement by data scientists machine learning can progress from a baseline of perhaps 50-to-60 percent accuracy to 90-95 percent accuracy.
Once 90+ percent accuracy is achieved the AI technology can be used for business decision making, Abrari said. But he notes that data scientists are needed to get to the place where AI applications can achieve success for the business.
While people are needed to make artificial intelligence work, there is currently a skills shortage, the CTO notes.
"Machine learning, Big Data, and data science skills are the most challenging to recruit for and potentially can create the greatest disruption to ongoing product development and go-to-market strategies if not filled," according to a Forbes article. That article was based on an IBM report that said the demand for data scientists is "disrupting the job market."
So more intelligent humans will be required for artificial intelligence to reach its potential for business applications.