Hiring has a visible output.

Add ten collectors and an organization can count ten new employees. Their salaries, training costs, hours worked, accounts handled, and eventually their contribution to performance can all be tracked.

Artificial intelligence creates a less straightforward equation.

An organization can license new tools, automate parts of a workflow, introduce AI-assisted analysis, and describe itself as more technologically advanced. But if the investment is intended to reduce operational pressure and limit the need for hiring more collectors, how does the business know whether it actually worked?

That question sits at the center of my recent Receivables Podcast conversation with Pete Klipa, Chief Client Experience Officer at Harvest Strategy Group. As our discussion moved from AI applications into performance measurement, Klipa made the case for judging technology through business outcomes rather than adoption alone.

AI Adoption is Not an Outcome

There is a difference between deploying technology and improving a business, but that distinction can become blurred when AI adoption itself receives so much attention. 

The stronger test is whether the technology produces measurable operational and financial improvements. Genuine efficiency should eventually appear in outcomes such as lower costs, controlled hiring expenses, revenue growth, stronger productivity, profitability, and overall financial stability.

This becomes especially important when evaluating hiring more collectors.

Suppose an organization handles substantially more account volume this year while adding relatively few employees. At first glance, that appears to demonstrate increased productivity. But technology should not automatically receive all the credit.

A meaningful ROI discussion needs a counterfactual: What would likely have happened without the technology?

  • If the organization previously needed one additional employee for every particular increase in workload, did that relationship change after implementation? 
  • If a process historically required a certain number of labor hours, what does it require now? 
  • If headcount remained stable, did service levels and results remain stable as well?

This comparison creates a baseline against which technological impact can be evaluated rather than assumed.

Cost Avoidance is Different From Cost Reduction

We need to separate these two concepts that are often treated as interchangeable.

Cost reduction means spending less than before. Cost avoidance means preventing an expense that otherwise would likely have occurred.

If an agency employs 100 people before an AI implementation and still employs 100 afterward, payroll has not necessarily decreased. But if increasing business volume would previously have required 15 additional employees and the existing operation can now absorb that growth with only five, the economics have still changed.

The avoided expense may include more than salary. Recruiting, onboarding, training, management time, equipment, software access, benefits, and other employment-related costs can accompany workforce expansion. That makes hiring growth a useful measurement variable even when the objective is not workforce reduction.

For organizations trying to evaluate AI, the relevant question may therefore be less about how many positions disappeared and more about how many additional positions growth no longer required.

Efficiency Cannot be Separated from Performance

Lower cost is only one side of the calculation.

An organization can control staffing expenses by asking the same workforce to manage more accounts, but that does not automatically indicate greater efficiency. Performance and compliance outcomes can also help determine whether technology is creating sustainable efficiency or merely shifting pressure elsewhere in the operation.

If a technology initiative is intended to reduce the need for hiring more collectors, organizations can examine the labor effect alongside recovery performance, compliance results, service expectations, and other relevant operational measures.

The strongest result is lower labor growth without sacrificing the outcomes the workforce exists to produce.

Measure the Workflow, not the Tool

AI ROI can be difficult to isolate because technology rarely operates independently from the systems and processes around it.

The value of an AI tool depends partly on how well it integrates into the broader operation. When connected effectively with enterprise systems, AI can support reporting, analysis, review, and decision-making without creating additional steps or disconnected workflows.

This means measuring the performance of an AI model may tell only part of the story.

A tool might complete an analysis in seconds but require an employee to spend 20 minutes validating, reformatting, transferring, and documenting the result. The speed of the AI itself will look impressive, but the workflow improvement may be considerably smaller.

ROI measurement therefore needs to capture the process before and after implementation. The unit being measured should be the business process, not merely the machine performing one part of it.

AI ROI Needs More than a Productivity Number

There is also a reason not to optimize AI exclusively around speed.

Receivables organizations operate within consumer-protection and debt-collection requirements regardless of whether work is performed manually or with technology. The Consumer Financial Protection Bureau notes that the FDCPA prohibits abusive, unfair, or deceptive debt collection practices, while Regulation F establishes federal requirements governing covered debt collectors and their activities.

Technology does not remove those obligations.

More broadly, the NIST AI Risk Management Framework treats measurement as an ongoing part of AI risk management rather than a one-time performance check. NIST's framework is cross-sectoral rather than debt-collection-specific, but the measurement principle is useful: a system should be evaluated on more than whether it technically functions.

For collection organizations, the business case therefore needs to accommodate productivity and the controls necessary to sustain acceptable outcomes. Ultimately, AI earns its place in the operation when the gains show up beyond the speedometer.

 

Explore more conversations on workforce strategy, artificial intelligence, operational performance, and the changing economics of receivables management on Receivables Info. 

About Adam Parks

Adam Parks, MBA, is the Founder and CEO of Receivables Info and a recognized leader in the receivables management industry. With nearly two decades of experience spanning debt portfolio management, technology, consulting, marketing, and operations, he brings a practical perspective to industry transformation. As host of the Receivables Podcast and a former President of RMAI, Parks regularly explores emerging trends in AI, compliance, recovery strategy, technology, and operational performance.