Viewpoint: Practical Applications of AI for Billing and Collections

Introduction

At last, another writeup on the power of artificial intelligence (AI). For the past year, we’ve all been on our own personal journey of AI adoption. In both business and personal applications, generative AI solutions have progressed from mind-blowing revelations (think of your first ChatGPT prompt) to daily companions that help with the little things. With the rapid pace of innovation and real-world application, it’s easy for our imaginations to run wild about what the future of AI holds.

Below, we promise no “art of the possible” talk (as fun as these conversations are) and will instead focus on practical applications of AI for billing and collections teams. We’ll discuss real world business outcomes these tools currently deliver, and how it’s moving the needle for system administrators in today’s top SaaS billing solutions.

First, let’s align on a several concrete statements about AI in August 2024:

1. AI’s value proposition is not equal across all landscapes. The value of AI models heavily favors text-based outputs, which doesn’t do a whole lot for finance teams. Said differently, current tools are much better at writing a bedtime story than they are at prepping and auditing financial statements.

2. AI models are highly data dependent. Poor data leads to poor output, but perhaps AI can help cleanse the dirty data we live with today (stop, no aspirational speak).

3. AI offerings and their underlying large language models are expensive to develop. To be monetized, this expense must be passed somewhere, likely the consumer.

Enough Fluff, How Can AI Help Me?

Vendors are introducing AI in two meaningful ways: to provide better value-driven outcomes for their customers, and to make their application more user friendly. We’ll explore these two threads in isolation.

How AI is improving business outcomes?

1. Improved Revenue Recovery: The concept of “smart retries” or “dynamic collections” is a well-baked offering in today’s billing and payment processing solutions. Leveraging years of transaction data, these LLMs are optimizing when failed payments are retried to increase likelihood of successful collection, improving cashflow while decreasing involuntary churn.

2. Proactive Collections Planning: AR automation software intersects with AI to determine what variables lead to past-due balances, and proactively predict and generate collections plans to prevent delinquency. Bolt on co-pilot gadgets to draft emails and summarize calls also provide efficiency wins.

3. Payment Matching and Reconciliation: AI helps accelerate all steps in the payment matching process:

   1. Matching payments to customers

   2. Matching payments to remittances

   3. Matching invoices to remittances

4. Anomaly Detection: Consumption analysis and historical billing data is used to spot anomalies in recurring billing. With anomaly detection, billing teams can catch bad bills before they’re sent, correct, and analyze the root cause to decrease the likelihood of future errors for the same reason.

5. Pricing Evaluation: A data point unique to your billing system is the price you’re charging for a service. Ease of use tools like natural-language queries help teams understand the price charged for the same service across all customer segments and regions. In turn, this is helping data science and pricing teams optimize pricing for top-line growth, margin health, and other price-optimization missions.

6. Adaptive Acceptance: Adaptive acceptance allows processors to try multiple submissions to the card network when the original submission is declined. These retries occur in real-time and are only attempted if the transaction remains profitable and is not likely to result in a dispute. Processors claim these adaptive models can increase authorization rates by up to 10%.

How AI is making system administrators’ lives easier?

1. Natural Language Assistants: Natural language helpers are quickly becoming an administrator’s best friend across three applications:

   1. Data Queries: At the risk of double-dipping, natural language queries are becoming commonplace in billing tools, reducing turn-around time on inquiries dependent on billing data. Strong offerings in this area intake a prompt, produce the necessary SQL, repeat back (in layman terms) what the query is doing, and then ask for feedback.

   2. Documentation Queries: Learning the ropes of a new application can be difficult, especially when you’re mostly/entirely dependent on vendor documentation. Natural language queries curate responses to specific questions not easily tackled by the conventional Search function in developer and administrator docs.

   3. Configuration: It’s still a bit early here, but for the sake of wanting this piece’s lifespan to outlive the current fiscal quarter, let’s talk about it. Prompting tools that create draft automation/workflows are a real thing, allowing users to describe a rule or configuration, and have AI do the build.

2. Less Reliance on Developers: It’s common for top-shelf billing and collections tools to be extended with code. This may be for traditional automation use cases (such as creating a workflow or sending a message to another system), or when developing customer facing templates for invoices, statements, credit notes, etc.. ChatGPT won’t make you the next Linus Torvalds, but odds are it can help you tweak that integration mapping or rearrange content on your invoice template.

That’s it. Intentionally short (I hope?) and pointed at the finance aficionados. As these trends shift and offerings emerge, we’ll keep you abreast on what’s real and fodder.

Thanks for reading,

Mitch Colyer