Artificial Intelligence and decision support

Or, it's hard to get good help

With seven billion people on the planet, every one of them making choices and having opinions, it may seem odd to say there aren’t enough people to make decisions, and that’s why we need help from artificial intelligence.

Odd, but true.

The increasing reach, power, and affordability of digital technology means more and more processes can be automated and theoretically made more efficient, producing more of the goods and services we want. But we don’t live in a clockwork world where everything, once wound up, happens according to a plan etched on some master cylinder. In the real world, all these technological “servants” need to respond to a changing environment by learning from context and requesting instructions. All the time.

They need examples and input so they can answer questions like: Should the store order more toothpaste?  Do the toilets in the airport’s third-floor men’s room need cleaning yet?  When will the train’s brakes wear out? Is this credit card transaction a fraud? If someone doesn’t like our new cookie flavor, is that a random quirk of taste or an omen that the cookie won’t sell well? How many people will want rides after the big game at a downtown stadium?

It’s a far cry from the science fiction visions of super-smart computers yanking decision-making power out of the hands of puny humans. Instead, life among modern technology is more like one of those British costume dramas where the lord or lady of the manor has to navigate the world surrounded by throngs of servants, some of whom are clever at anticipating their employers’ wishes while others are dimwitted and in need of constant direction.

Software-based staff

The truth is, we need these software-based “servants” because there’s no way human beings could focus enough to make millions of decisions a second. Would you really spend your day (or pay other people to spend theirs) staring at a series of gauges monitoring the brakes of every wheel of every train running between Boston and New York, to see if any of them are overheating ? Or examining every last credit-card purchase made on a single day at a single department store, to make sure none of them is fraudulent? Or reading through thousands of Tweets or Facebook posts that hashtag your company’s new product, to scope out the public’s opinion of it?

Of course not. Even if a person had enough attention to detail to notice microscopic changes across millions of data points (and didn’t collapse from boredom), you could never hire enough people to fill the need. That’s where artificial intelligence (AI) comes in. To use the “Downton Abbey” parallel, it’s the servant observant enough to spot every change in their surroundings and anticipate problems – but smart enough to know what’s worth bothering the master about, and what isn’t.

With the OpenText™ Analytics Suite, we already offer business intelligence and reporting solutions that monitor your organization’s processes, show you what’s going on in visually appealing, easily customized dashboards, and help you make decisions by analyzing results and predicting trends.

Now, with Magellan, we’ve taken the next step by creating an AI-based analytics application that enhances and automates decision-making by learning from, and replicating, human judgment. It can work with most business applications, and can ingest and manage complex and highly diverse data on a platform specifically designed for efficiently crunching massive amounts of information.

The next question: What can Magellan do for me? In some ways, science fiction has steered us wrong. “Artificial intelligence” or “machine learning” summons up images of chatty robots or super-intelligent spaceship computers like HAL 9000 from “2001.”

Machines don’t get hungry

But machines aren’t remotely at that level of sophistication yet. Instead, it turns out they’re good at learning to make simple decisions – often too simple to interest a human – and doing it again and again, millions of times over. Machines lack the nuances that inform human judgments. But that can also protect them from random variation or human weaknesses (like judges whose parole decisions get more negative when they’re tired and hungry), and keep them consistent.

Moreover, they can digest enormous sets of data to spot patterns we don’t necessarily see, and improve their “judgment” as they get more feedback.  All the examples mentioned above – ride-share demand, airport bathroom cleanliness, credit-card fraud, or consumer trend-spotting in social media – are use cases that organizations are using Magellan to derive value from.

The common thread in each of these is that a business process generates a data trail that a human with expertise can spot a pattern in, and make a simple judgment about. For example, a transaction is suspicious or legitimate; a consumer has a positive or negative opinion of a new flavor; this street corner does or doesn’t need another driver to come by. Then that human can “train” the Magellan software to notice the significant elements of that pattern, find it in other examples, and make that judgment itself.

Software deals with the boring, repetitive parts

One interesting example is examining workers’ compensation claims for potential fraud. Everyone involved in this process has an interest in making sure that honest applicants get the medical care and claim reimbursements they need, while dishonest applicants are detected and penalized – as efficiently as possible.

So let’s say we’ve got a well-trained claims inspector. She knows the signs of potential fraud, so her professional judgment is quite accurate about whether a report is honest or not. But it may take her 30 to 45 minutes to read a claim form, make a judgment, and note in her report why she labeled it as valid or invalid.

And then let’s say she has 800 of those claim forms to get through – at least ten full 40-hour work weeks, assuming she does nothing else in all that time.

Now imagine that her company first spends some time setting up Magellan. After a few days’ worth of “training” the software to digest and interpret the text in the claim forms (with the help of a data scientist to codify the claims inspector’s expertise in terms the software can understand ), it can read a form in a second or two… and go through the stack of 800 claim forms in less than an hour.

Magellan is still not the final authority on whether workers get their medical bills repaid. It’s just the helpful assistant sorting out the majority of straightforward claim applications for streamlined processing, while bringing the few sketchy ones to the attention of its “boss.”

For example, Magellan might obey instructions to flag claims where the employee’s description of the accident conflicts with the injury details reported by the doctor or other witnesses, or where the claimant refuses a diagnostic procedure that would confirm the injury’s size or cause. Or it might study archives of fraudulent claims and discover that alleged injuries reported first thing Monday morning have a higher-than-normal chance of being fabricated.

This is just one of the many ways Magellan’s AI-powered analytics can streamline a business process and make it more efficient and profitable. Over the following weeks we’ll showcase different use cases for Magellan. If there’s a particular industry or business function that interests you, please let us know in the comments below.

And if you’d like to learn more about Magellan, click here.

Stannie Holt

Stannie Holt is a Marketing Content Writer at OpenText. She has over 20 years' experience as a journalist, market research analyst, and content marketing expert in the fields of enterprise business software, machine learning, e-discovery, and analytics.

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Close