AI and the Need for Speed

Artificial intelligence (AI) holds substantial promise for organizations to reduce costs and increase quality, but how AI affects organizations’ use of and relationship to time — in reacting, managing, and learning — may be the most jarring.

Certainly costs and quality changes are nontrivial:

• After initial development, AI-based systems may substantially reduce variable costs. Organizations can replace or increase productivity of expensive knowledge workers as AI supplants many tasks. With AI, a thousand radiologists cost the same as one. With AI, each customer can have his or her own customer service representative. Currently expensive bottlenecks may dissipate.
• With machinelike precision (pun intended), decision making can be consistent. Without those pesky humans adding entropy, processes can work the same every time. Each of the AI radiologists performs exactly the same as the others, reaching the same diagnosis. Each AI customer service representative recommends the same resolution. And with this uniformity in place, organizations can incrementally refine and improve, ever increasing quality.

But while costs and quality are important, improved AI also heralds changes to a fundamental business limitation — time. For example:

• Current mortgage approval processes typically take 30-45 days. “Getting a loan, even a preapproval, doesn’t happen overnight.” Why couldn’t it happen overnight? With data increasingly available to support all the information that goes into a loan application, AI approaches may be able to dramatically reduce the time required to get a loan. If you find a house you like in the morning, the mortgage application process doesn’t need to be the holdup to getting the keys that afternoon. Mortgage brokers (if they continue to exist) will simply need to be able to swipe credit cards.
• Lawsuits can take considerable time to resolve. On average, doctors spend nearly 11% of their careers with an unresolved malpractice claim. Some of the time is in gathering data. Some is in delays due to court congestion. Some is due to deliberation and settlement. How much of this could be shortened with AI reducing discovery and decision-making time?
• Medical diagnosis itself can also take substantial time during which the patient is not being treated. In primary care settings, the diagnosis of childhood cancers can be difficult, particularly due to the low incidence rate; but “shorter lag time could improve the prognosis, and … prolongation of the diagnosis period will badly affect the prognosis.” The potential for improved health outcomes through faster diagnosis is undeniably appealing.

Improvements in speed through AI have the potential for both monetary (e.g., worker time) and nonmonetary (e.g., customer satisfaction, reduced anxiety, improved health) benefits.

However, my sense is that most organizations are used to current timings and aren’t ready for them to change substantially. As the pace begins to increase, what do organizations need to watch out for?

• Too fast to manage: The pace of current processes allows time for monitoring and management. If (when!) something doesn’t work as expected, there is usually still time to correct and manage the process and outcomes. But human managers may not have time to manage at a machine pace. They may not have the opportunity to intervene before the money is gone from a risky loan, the guilty walk away, or the patient suffers from incorrect treatment. In the race to optimal speed, the breakneck pace may reward risky behavior as organizations excel — until they crash. A breakneck pace is called “breakneck” for a reason.
• Too fast to react: In an ideal competitive scenario, your organization will be able to use AI to be fast, while competitors remain slower. However, it won’t likely work out that way. Competitors won’t sit still. As you are building AI based on the environment from yesterday, your AI systems will have to compete with the competitive environment of today. The potential for rapid change by others in the environment may prevent processes from settling down into a steady state.
• Too fast to learn: Much of the improvement in AI in the last decade stems from a transition from a rule-based approach to a data-based approach. Instead of coding rules for all expected scenarios, AI models are trained on data and learn from observing. But in an increasingly AI world, the value of data may decay quickly. The rate of decay may exceed the rate at which new data is available or the AI model can ingest it. The Netflix prize is a classic example; the company offered a prize for a substantial improvement in its algorithm that recommends movies based on historical customer data. After awarding the prize, Netflix found the algorithm was not as useful on their current data — their historical data on DVD-by-mail rentals was much less useful in understanding video streaming behavior. When AI depends on data, the rapid decay of that data may be debilitating.