Artificial Intelligence in Debt Collection – Near-Term Applications

Raghav Bharadwaj

Raghav is serves as Analyst at Emerj, covering AI trends across major industry updates, and conducting qualitative and quantitative research. He previously worked for Frost & Sullivan and Infiniti Research.

Artificial Intelligence in Debt Collection - Near-Term Applications 1

According to the Consumer Financial Protection Bureau, Americans filed more grievances about debt collections than about any other financial incident. Of the 316,810 complaints received by the CFPB about debt collection in 2017, the most common was, “Continued attempts to collect debt not owed,” which was cited by 39 percent of grievance filers.

Debt collection in finance is starting to be disrupted by artificial intelligence due to the availability of massive amounts of historical records of customers for banks and other financial institutions. Most AI applications that have real-world business significance for debt collection today seem to be in personalizing communications to customers and identifying clusters of similar debtor profiles.

From our research we have segmented the AI applications for debt collection into the following broad segments:

  • Driving Additional Messaging Campaigns
  • Customer Service (Debtor) Personalization
  • Debt Management Service

We delve further into each of these applications and aim to coax out the need-to-know factors for business leaders regarding AI usage for debt collection.

Driving Additional Messaging Campaigns

Tailoring interactions to debtor habits could be possible with AI today. Virtual assistants are starting to be deployed through channels in which collections agencies can reach out to debtors through email SMS, and outbound dialing, allowing organizations to increase the number of individuals that they’re able to contact on a daily basis.

TrueAccord

TrueAccord was founded in 2013 in San Francisco and claims to be offering an AI-driven debt collection solution. The company claims to be offering debt collection solutions to banks, eCommerce and telecom companies.

TrueAccord claims that their decision engine uses machine learning to create digital interactive experiences that are customized for each debtor. The company claims its platform can create an interaction model for each debtor.

TrueAccord claims this model provides banks with the best possible channel and time to reach out to existing and new debtors which might eventually result in better debt revenue collections.

The company says more than 1.5 million debtors have already been modeled using their platform. Based on these existing debtor profiles, their software claims to predict an individual’s response time, schedule, best communication channel, and type of content that they’ll respond to.

TrueAccord adds that its decision engine can automatically select the appropriate pre-approved messages from banks to deliver to debtors. The software also tracks, in real time, the action events from the debtor, such as interaction with call centers, or email opens, link clicks and browsing patterns on TrueAccord assets. The software then sends this data to employees at debt collecting agencies so they can plan the ideal style and time of their next message.

  • Banks can integrate the TrueAccord software to automate their debt collection team’s communications processes.
  • Employees at the debt collection team can log in to the TrueAccord software and enter debtor information like names, geographical locations, and past repayment data.
  • They will then receive a notification which tells them the best possible time and mode of communication to be used to reach that particular debtor.
  • Once the debtor replies to a message, the software tracks the time and type of reply. For example, it notes whether the reply was by voicemail, email or letters. The company claims it will use this data to suggest what channel the next message should likely be sent through.

We could find no video demos for the TrueAccord software.

In the video below, TrueAccord leadership explains how their platform can help banks and financial institutions with debt collection during the 2015 FinTech & Retail EXPO:

According to a case study, TrueAccord claims to have worked with Upwork (previously Elance-oDesk), a freelancer network which reported having trouble collecting payment for the work agreements that took place on its platform. Upwork needed an immediate recovery solution to ensure payment was received for contracted work on their platform.

Upwork then decided to work with a team of third-party experts to support their internal efforts to recoup payment:

On our own, we weren’t actively pursuing old debt. With TrueAccord’s adaptive solution, we were willing to try and recover as much of it as possible. (Denise Aptekar, Director of Global Payments for Upwork)

TrueAccord claims that the first step towards integrating their platform for Upwork was to analyze the old debts from Upwork’s books. The company claims that Upwork used their platforms to improve their debt collections using personalized messaging. Upwork reportedly added several millions of dollars in old and new accounts after the integration over the past year.

Paul Lucas is CTO of TrueAccord. Previously, he served as the Director of Technology for travel booking firm Expedia for six years before joining TrueAccord. There also seems to be evidence that the company has closed a Series B funding round for $22 million.

In the same post, the company also claims that their clients include technology firms like Yelp! and LendUp.

Customer Service (Debtor) Personalization

Collections agencies usually have a lot of inbound traffic with consumers wanting to make a payment or update their account information and more. While a certain percentage of those calls have the potential to drive revenue, most do not. AI can be used today to help banks focus their attention on calls that are more likely to drive revenue.

CollectAI

CollectAI was founded in 2016 in Hamburg as a subsidiary of German retailer Otto Group and provides AI-based payments and collection services. The company claims to help banks or other financial institutions with managing their account receivables. The company claims to be using machine learning to help their clients personalize communications with every customer for debt collection applications.

Traditionally, banks have used standard letters without any individualization sent through email or post as a way of managing receivables. CollectAI claims that their solution allows banks to digitize the personalization process across all communication channels such as email, WhatsApp, text message, and postal letter.

In addition, the company claims its software can also recognize the right time of day for the message to sent to increase the likelihood of receiving a reply from a debtor. The company says the software might also suggest the proper tone of voice for the communication such as friendly, amusing, remindful or assertive.

CollectAI claims that their algorithms measure the effectiveness of its customer communication decision-making (such as analyzing time sent or content in relation to email opens). The company claims their software can profile a specific customer, or debtor, and also cluster similar debtor profiles to help banks send personalized communications to new individuals.

According to Mirko Krauel, CEO of CollectAI, the platform measures the level of success its decision are having by answering questions like:

  • “Did the debtor open the email?”
  • “Did they go to the landing page?”
  • “Did the individual attempt to make payment in the last month?”

The software can then update the ‘best responses’ for each type of debtor profile based on the information the algorithm has.

A bank might deploy the CollectAI software for their internal debt collection team. After the integration, the software can automatically prompt debt collection employees with the best channel for communications. CollectAI claims the bank’s debt collection team can also gain notifications on what type of content to send to each debtor for maximum chance of revenue collection.

Mirko also adds the integration of the collectAI platform usually happens through a RESTful API although they have also enacted direct integration into bank systems more rarely. Their platform also supports some ERP systems via a standard integration and serves banks, eCommerce platforms, digital publishing houses, insurances, utilities, etc. with automated debt collection personalization.

We could not find a suitable demo video for how the CollectAI software functions although below is a 7-minute interview with Jorge Davila­ Chacon, CollectAI where he explains the company’s debt collection solution for banks:

COO Steve Emecz presented a live case study at the EXEC 2018 in March where he claims the company worked on projects with the Otto Group and Hanseatic Bank. We discuss their collaboration with Hanseatic Bank in more detail below:

Hanseatic Bank, a 75-percent subsidiary of Société Générale, deployed CollectAI’s B2B2C solution for debt collection. The CollectAI platform could potentially notify Hanseatic’s customers with payment reminders in the segments of credit cards and consumer credits.

CollectAI also claims to have developed a dynamic landing page for each customer who owed a payment which leads to an immediate payment option for the overdue amount. The platform also identified the best timing and digital communication channels of each debtor by coaxing out patterns from historical re-payments data.

In the first six months of the collaboration, the case study noted that Hanseatic Bank’s collection rate on debts increased by 24 percent in total. This was 14 percentage points higher the rate measured, at an unspecified time, before implementation began.

CollectAi also claims that they helped reduce the bank’s expenses for receivables management by 88.6 percent in one particular area of operation. Further details on measurable results for Hanseatic were unavailable.

Michel Billon, CEO of Hanseatic Bank said: “Customer-centric collections are key for our digital transformation strategy. Thanks to CollectAI’s solution we have increased the efficiency of our account receivables management and optimized customer retention with a higher satisfaction. Further key results were an overall higher cash flow, a faster execution as well as reduced communication costs.”

Mirko Krauel and Michel Billon explain the case study in more detail in the video below in their Money 20/20 Conference Keynote Speech:

We could find no evidence of previous robust AI experience from the leadership team at CollectAI or any client projects, apart from those mentioned.

Debt Management Service

Summit LLC

Summit Consulting was founded in Washington DC in 2003 and has around 90 employees. The company claims to provide analytics and data science solutions for commercial clients and government agencies.

From our research, we could find no product demos or details on how a bank or governmental agency might use Summit’s analytics solutions. Details on how the integration might work or how individual users in debt collections agencies might access their software insights were also unavailable at the time of writing.

Summit claims to have worked in a project with the US Department of Treasury to develop a debt management service (DMS) for their fiscal service team. The Department of Treasury needed to identify, collect and resolve debts owed to other government agencies such as those relevant to state child support and delinquent student loans.

Summit claims to have received databases containing information on debts and collection activities from the government agency and used machine learning techniques to classify different clusters of debts based on customer characteristics. The company claims that their software determined which debts would likely result in the highest collections using the DMS’ active collection tools and campaigns.

In addition, their software also might have identified the best timing and holding period for conducting debt collection communications in the form of calls, emails or letters. Summit claims their post-integration the Department of Treasury’s new collection strategies led to a seven-percent increase in collections while reducing the costs of the DMS program to taxpayers. Further details on this projects and its results were unavailable.  

Ed Dieterle is the Director of Data Science at Summit Consulting was previously a Researcher at Harvard and Senior Program Officer for the Bill and Melinda Gates Foundation although it was unclear if he worked specifically with data science projects while there.

Summit LLC claims to have worked on analytics projects for US Governmental agencies such as the Occupational Safety and Health Administration (OSHA) and US Department of Labor (DOL)

Near-Term Applications and Trends

Through our research, we came to the following key points that business leaders looking to apply AI for debt collection processes might need to know before getting into a project.

  • Personalization of debtor interaction seems to be the largest trend for AI applications in debt collection today as seen with solutions from companies like TrueAccord. Furthermore, based on some of the case studies noted, it seems that this personalization has resulted in some increased and quicker repayment rates.
  • Small and medium-sized businesses might need to understand that they may lack the scale needed in terms of the number of customers for AI solutions to be truly effective. Many of these noted solutions claim to predict best outcomes after analyzing a large pool of current debtors and past customers that are similar to the debtor. Because of this, customer histories, or large amounts of current data, may be needed for the application’s success.
  • With the applications above, we’ve seen that integration of debt management systems could take two to three months or more. Additionally, it seems that the involvement of internal debt collection officers being vital in developing an AI model. Financial institutions will need to take time and money spent on these applications into account and weigh these factors against a program’s benefits.

 

Header image credit: Joinnus.com

Subscribe