Accelerate Access to Data and Analytics With AI

To Improve Cybersecurity, Think Like a Hacker

Training employees how to locate and use data insights is a major Big Data bottleneck. A potential solution: AI systems that compress employees’ learning curves.

Even if businesses have made sizable investments in data and analytics, every employee doesn’t necessarily understand how to properly use that data. Learning enough to take advantage of those investments requires time and effort for each employee. While deep reservoirs of data may exist in the organization, the flow through individual employees may be more of a trickle than a cascade.

Artificial intelligence-based approaches may be able to help by enabling each employee everywhere to know what the organization overall knows somewhere.

The “last mile” is a common problem in transferring information. In telecommunications, high-bandwidth channels can move data quickly between central locations. The difficulty is that, at the periphery, data transfer over the “last mile” to individual users is a bottleneck. The rate of information flow in the slowest segment, typically the final one where each communication channel services just a single user, governs the overall rate of information flow. All the massive central bandwidth is for naught when the final link to the user is slow. For example, internet content may travel from all over the world to your local provider without a problem, but getting that to your home office can be stymied by all sorts of problems, such as antiquated cables on your nearby telephone pole, frail wiring in your house walls, or outdated components in your computer. No amount of central investment by your provider can correct limitations at your house — and investments within your neighbor’s house don’t help you, either. Improving the final bottleneck requires an idiosyncratic investment by each user.

Unfortunately, organizations that are ratcheting up their use of analytics must overcome similar last-mile issues. While the organization may have considerable data and analytics capacity, each employee must make investments to understand what is available and how it can be used. Pervasive use of data and analytics by an organization requires pervasive understanding of data and analytics. This is a fundamentally difficult and slow organizational learning undertaking.

One of the exciting possibilities is that artificial intelligence may help businesses accelerate organizational learning. Machine learning can enable faster organizational learning as it can help each employee quickly understand what others in the organization understand — artificial intelligence can distribute learning quickly.

Our forthcoming report on AI and business strategy provides an excellent illustration of this potential in the lead example of Airbus. As Airbus began production of its new A350 aircraft, the company wanted to fabricate the aircraft faster than any prior model — but without any compromise in quality.

One specific challenge was that the new aircraft was, by definition, new. No one had any experience in building it. Airbus faced an organizational learning problem as much as a production problem.

Experience is an important component of the manufacturing process. In a project the size of an aircraft assembly, numerous production difficulties and anomalies were bound to occur. Some problems would be big and require manufacturing to stop until they were isolated and corrected. Other problems would be small and could be quickly worked around without affecting product quality. With time, employees would learn how to handle problems and, most important, learn to discern which problems warranted stopping production and which did not.

To meet the goal of faster production, the whole Airbus organization would need to learn more quickly than it ever had before. Airbus already had the data and analytics infrastructure to collect data related to problems quickly. Yet it would still take considerable time for each of the employees involved to garner the experience necessary — this was the potential bottleneck.

Source From: MITSloan Management Review