Jay Tchakorav will give a presentation at the CRF Forum & EXPO in Seattle, WA on August 11, 2015. His presentation is titled, “Five Ways Artificial Intelligence is Transforming and Automating Accounts Receivables.”
Levelset is visiting the CRF Forum & EXPO next week. In preparation, we’re brushing up on one of the topics that we’re most excited about: artificial intelligence in credit receivables.
Policies: The Original Automation
Levelset‘s CEO Scott Wolfe, jr. wrote about automation and artificial intelligence in a July 2014 LinkedIn post titled, “Is Automation Dead, or Just Getting Started?” In it, he dubs policies “the original automation.” Company policies were implemented to facilitate streamlined and consistent decision-making. A policy removes the need to weigh options. It dictates what to do in a given scenario. Thus, a policy automates. (Not sure how to come down on a decision? Consult your company’s policy. The decision has already been made.)
Advancements in computer power and the advent of the internet enabled businesses to electronically automate their policies, such that machines could implement a policy without the assistance of a human.
Take, for example, Hawkeye technology used in Tennis Matches. Tennis has a policy (the rules of the game), which say that if the ball lands outside of the line, the point is lost. Referees and line judges implement this policy using their best judgment. Hawkeye automates that policy by doing exactly what a judge does: it determines the position of the ball. (Of course, Hawkeye doesn’t yell “in” or “out”, so it falls short of actually implementing the policy, instead aiding referees.)
Can Computers Make Exceptions?
The primary shortcoming of automation is its weakness at making exceptions.
Human beings make exceptions to rules and policies. Exceptions are unique — if they weren’t, they would be included in a policy to begin with. It’s not always clear why somebody makes an exception. Usually, it just feels right. That’s one reason we still have human referees. (Could a computer decide when to give a yellow card in soccer? Could a computer decide when to give a pitcher in baseball a warning?)
Computers can write symphonies, and computers can write news articles.
Can a computer learn when to make an exception? If a computer knows when to make an exception, is is still an exception? The Atlantic thinks not. See their 2011 article, Why Computers Will Never Replace Us. Whether or not a computer can entirely replace a human being, it can certainly aid us in making better decisions.
Implications for Credit and Receivables
Credit departments have all sorts of decisions to make, and they often are not straightforward. To name a few, these include: when to give credit; whom to give credit to; the size of a credit line; payment terms; collections tactics.
Credit departments in the construction industry have the added task of making decisions about security instruments: when to send preliminary notices, when to use mechanics lien rights, etc. There are three A’s that will continue to impact credit and receivables: Automation, Analysis, and Artificial Intelligence (AI).
1. Automation
Most credit departments have policies that dictate credit decisions. For large businesses, it can be a burdensome process to review every decision. This is where modern technology can help.
Teach a computer your credit policy. Give the computer the same information a human would use to make a credit decision, and the computer will compute and output the decision. The biggest difference: a computer can make thousands of decisions in minutes, or seconds. The difference in speed allows you also to include more data to make smarter decisions. Feed a computer any amount of information: this could range from a prospective customers’s credit history, to receivables data in your industry, to your own receivables history.
For construction businesses, this could also include mechanics lien laws.
Some credit managers may fully automate this process, allowing computers to make decisions without review. Others will require every decision to be reviewed. There does exist a happy medium. One option is to set parameters that identify when a decision needs review.
2. Analysis
The analytical possibilities are endless when a credit department enlists the help of computers. Compare any number of factors against each other to determine which reaped the most positive benefits: Which terms or which customer profiles led to faster payments, what led to increased repeat sales, whether exceptions performed better or worse than policy decisions.
With enough data, you can hone in on granular situations to completely maximize receivables performance. For example, you might come to the conclusion that, “if a prospect fits profile X and they want credit in August, then we should agree to terms Y, and use collection tactics Z, as these decisions will lead to the highest probability of timely payment.”
3. Artificial Intelligence
AI and machine learning can take automation and analysis even further. In some ways, Machine learning is automated analysis.
As you continue to make credit decisions and collect data, the decision-making machine analyzes this data to identify what leads to best outcomes. It then adjusts its algorithm to make smarter decisions based on its learnings.
Conclusion
All of this likely seems very in-the-clouds and far off, but the reality is that automation and machine learning are here. They’re present in most of the apps that we use everyday (Google Maps, Spotify, Netflix), and they can be utilized in credit departments as well.
Even small investments in automation software can save your department dozens or hundreds of hours of work, and minimize human error.