娇色导航

Milind Wagle, CIO, Equinix, has his own “customer-churn predictor.” It tells him why accounts are turning over and whether they are likely to do so in the future. That information gives the data center services provider a chance to improve its bookings accuracy by improving service and planning for likely occupancy rates to come. “We are shifting our mindset away from treating analytics as after-the-fact reports toward embedding that intelligence in the business process itself,” he said. With fine tuning, Wagle reports that the churn predictor is approaching 90% accuracy.

Equinix

Wagle’s use of artificial intelligence (AI) with analytics is not unusual. A recent study fielded by Capgemini shows that, of nearly 1,000 AI-using enterprises, almost 80 percent have used it for data analytics and report gaining valuable insights as a result. While that may sound like a surprisingly high number, most of the CIOs we spoke to found it to be credible.

“I’m not at all surprised,” said Joel Jacobs vice president, 娇色导航and CSO, MITRE Corporation. “While I don’t believe that all of the possibilities for AI have settled out, it makes sense that large organizations are recognizing that there’s great potential.”

Machine learning and, to a lesser extent, deep learning are the branches of AI that are being harnessed for data analytics work. Machine learning (ML) works by categorizing data, a basic building block of data analytics, leading to a sort of natural synergy between the two. There is a broad variety of use cases because almost everything touches business data.

Machine learning is also the right tool to recognize and catalog unstructured data, such as documents, images and video, as well as dark data, information that you’ve never accessed (probably because it’s an amorphous part of your big data). Most business intelligence (BI) systems still require structured data. “What about documents? What about images? That’s where we can use machine learning.” said Dan Olley, executive vice president and CTO, Elsevier. “It can both extract information from documents and create annotations that can be further indexed. ML is our key means for extracting knowledge from readable material.”

Although there have been solutions for big data, unstructured data has typically gone untapped because it was more difficult to access prior to AI. The combination of ML and less expensive cloud compute power means that some types of dark data are also within reach. Accessing unstructured and dark data has been the impetus for many data analytics breakthroughs for organizations using machine learning in recent years. The addition of new data sometimes brings new perspective.

Milind Wagle, CIO, Equinix, has his own “customer-churn predictor.” It tells him why accounts are turning over and whether they are likely to do so in the future. That information gives the data center services provider a chance to improve its bookings accuracy by improving service and planning for likely occupancy rates to come. “We are shifting our mindset away from treating analytics as after-the-fact reports toward embedding that intelligence in the business process itself,” he said. With fine tuning, Wagle reports that the churn predictor is approaching 90% accuracy.

Equinix

Wagle’s use of artificial intelligence (AI) with analytics is not unusual. A recent study fielded by Capgemini shows that, of nearly 1,000 AI-using enterprises, almost 80 percent have used it for data analytics and report gaining valuable insights as a result. While that may sound like a surprisingly high number, most of the CIOs we spoke to found it to be credible.

“I’m not at all surprised,” said Joel Jacobs vice president, 娇色导航and CSO, MITRE Corporation. “While I don’t believe that all of the possibilities for AI have settled out, it makes sense that large organizations are recognizing that there’s great potential.”

Machine learning and, to a lesser extent, deep learning are the branches of AI that are being harnessed for data analytics work. Machine learning (ML) works by categorizing data, a basic building block of data analytics, leading to a sort of natural synergy between the two. There is a broad variety of use cases because almost everything touches business data.

Machine learning is also the right tool to recognize and catalog unstructured data, such as documents, images and video, as well as dark data, information that you’ve never accessed (probably because it’s an amorphous part of your big data). Most business intelligence (BI) systems still require structured data. “What about documents? What about images? That’s where we can use machine learning.” said Dan Olley, executive vice president and CTO, Elsevier. “It can both extract information from documents and create annotations that can be further indexed. ML is our key means for extracting knowledge from readable material.”

Elsevier

Although there have been solutions for big data, unstructured data has typically gone untapped because it was more difficult to access prior to AI. The combination of ML and less expensive cloud compute power means that some types of dark data are also within reach. Accessing unstructured and dark data has been the impetus for many data analytics breakthroughs for organizations using machine learning in recent years. The addition of new data sometimes brings new perspective.

Aiming high

Many of the CIOs and their teams wielding AI are thinking big. They’re rolling out projects at scale. They have multiple efforts underway. Some 58 percent of the respondents to the Capgemini survey reported taking on high-complexity, high-benefit use cases. There’s a sense of urgency, because the payback comes in several forms and is capable of generating significant business value. Some organizations are also racing to a competitive advantage. “Usage of AI/ML will continue to increase over the next few years, eventually becoming invisible because it becomes embedded [universally] into business process,” said Les Ottolenghi, EVP and CIO, Caesars Entertainment. “AI functionality is something customers will come to expect from organizations and they will be upset when one does not incorporate the paradigm across all its various touch points.”

Caesars has jumped into AI with both feet. The company has multiple AI efforts either completed or on the way, including contextual customer personalization (customer journey), conversational AI (chatbot, voice), real-time fraud analysis leveraging dark data, photo/emotion recognition to go with their voice interface, a recommendation engine and more.

Caesars Entertainment

For its customer journey project, Caesars created an engine that correlates data across more than a dozen data sources, matching it in near real-time using fuzzy logic ML to confirm whether data from disparate systems is associated with a specific person. When that is the case, the system selectively updates that person’s data elements in a time series table. Caesars uses additional AI techniques to identify contextual offers that it can send to the customer at an optimal time and place and by the appropriate communication method. Although the means, methodology and specific business processes change from company to company, this example is representative of the ways many companies are pairing data analytics and machine learning to generate revenue through upselling or cross-selling in context.

Another context-specific use case for AI and data is being pursued by several companies, including MITRE. “We want to better leverage our existing knowledge,” said Michal Cenkl, director of innovation and technology, MITRE Corporation. For example, let’s say an employee is trying to solve a problem. How did MITRE solve similar problems in the past? Cenkl calls it cognitive assistance. “I think the opportunity is around integrating our unstructured data — for example, project reports and deliverables to our sponsors and customers — integrating those with some of the structured information around the projects that generated them,” he added.

Prediction

MITRE is in the early stages of a similar follow-on initiative that adds contextual and predictive elements. Loosely called “cognitive anticipatory knowledge delivery,” it will seek to provide information to employees in the context of what they are doing. Cenkl used the example of a project manager working on a specific project to describe context. The AI component would filter the available knowledge streams and suggest the most relevant content to the user. Elsevier has a similar functionality that targets researchers in cross-disciplinary environments, offering both context and predictive relevancy.

MITRE Corporation

Many companies using AI report developing predictive analytics that they’ve come to depend on. This usually starts with enterprise-specific key performance indicators. At Equinix, Wagle’s team developed a predictive tool based on machine learning that tracks and forecasts indicators such as power consumption for data centers, bookings, renewal predictions in customer buying propensities. Equinix operates a similar tool that predicts information security breaches, system outages and incidents of compromise. Predictive analytics needs to be tested and fine-tuned over time before it becomes valuable.

Elsevier is using robotic process automation (RPA) in its finance department to spot potential fraudulent behaviors and hosting issues. As a result of that system, “we gained the ability to make predictions and data-driven judgments,” Olley added. It’s not unusual for predictive analytics to grow out of an AI process as a byproduct.

AI and the CIO

The CIOs we spoke to were in universal agreement that AI-powered analytics is a highly significant, likely transformative trend. Blake Hankins, 娇色导航of CyrusOne described it as “the cornerstone to the digital transformation toolkit. It will allow the enterprise to be more efficient and focus on the items which are most meaningful.”

CyrusOne

Caesars’ Ottolenghi said: “AI is driving digital transformation at Caesars and a slew of other companies. It helped us move to a platform [framework] that offers a fail fast, test often, innovate quicker paradigm. This transforms how we work on experiments and proofs of concept. AI also helps us tap into dark and/or unstructured data. And it provides insight across different data elements.”

Almost all the CIOs interviewed for this story report that C-suites are generally bullish about AI/ML and data analytics projects, in fact AI in general. If there is a disconnect, it’s probably related to where and how to attribute the business value of AI. Hankins also points out that most members of the C-suite are unaware of all that’s involved in getting an AI/ML data program incubated. “It’s incumbent upon CIOs to help bring awareness of the potential of such technologies and help create linkage of AI projects and investments to specific business outcomes,” Hankins said.

Many of those who’ve already doubled down on AI have ready advice for CIOs and others looking to do the same. “It’s really hard to do this without data,” said MITRE’s Cenkl. Similar sentiments from others echo this advice.

Olley points out that with machine learning, data becomes the most critical asset because it’s the training set for your machine learning or deep learning models. Your machine learning project is only as good as the data that it the relies on. Both the volume and quality of the data come into play.

Ottolenghi stresses making sure you have management buy-in and choosing use cases that solve real business problems. “That way, you have the backing of the company to do what it takes to make sure the use case is successful,” he said.

Finally, start thinking about hiring talented personnel well trained in data analytics, data science, data engineering if you don’t already have those skills in place.

Parting thoughts

Machine learning and data analytics, when properly developed, can sometimes augment each other in an almost catalytic fashion. Machine learning powers the automation of data analytics, which leads to insights and decisions. That in turn can lead to identifying new data — and the process repeats. There is a power at work in the combination of AI/ML and business data that most CIOs cannot afford to ignore. This story has scratched the surface of the number and types of viable use cases. If you’re still sitting on the fence about AI, it’s time to act.

Related video