By David Hewitt, Senior VP of Sales
Right now, every recovery vendor in healthcare is saying the same thing: AI-powered, machine learning, intelligent automation. Read five websites and you’ll see the same promises five times. And here’s the uncomfortable truth… most of it is true. The tools are real, they’re widely available, and they’re genuinely good at the mechanics: requesting statements, extracting data, matching line items at scale. Heck, we use all the same tools too.
Which is exactly why the tools no longer tell you anything. When everyone has AI, AI stops being the differentiator. What separates a recovery partner from a recovery disappointment is what sits underneath the technology. So before you sign with anyone, including us, ask these five questions.
- What is your AI actually trained on?
An AI system is only as good as what it has learned. A model trained on generic AP data knows what an invoice looks like. A model trained on three decades of healthcare-specific supplier behavior knows which vendors suppress credits, which ones require escalation, and what a real anomaly looks like in a hospital’s payables. Ask the vendor to describe their training data. If the answer is vague – the answer is generic.
- How do you find money at suppliers we’ve never paid?
This one surprises people. Some of the largest recoveries sit with vendors that show zero spend activity in your system, because they’re related to a vendor where your activity is significant. Mergers, acquisitions, name changes, merged and purged vendor records… the corporate family tree behind your vendor master is invisible to any tool that only reads your data. Most automated approaches rank effort by spend, chasing the 20% of suppliers that carry 80% of dollars. Ask the vendor how they find the credits sitting where your records say there’s nothing. If they can’t answer, those dollars stay lost.
- What do you ask for when the statement isn’t enough?
Every supplier speaks a different language. One vendor’s statement of account is another’s aged trial balance, account ledger, or remittance history. Some suppliers suppress credits from standard statements entirely unless the right question is asked the right way. Automation takes whatever document arrives at face value. Ask the vendor: when the statement doesn’t show the full picture, how do you know, and what do you request next? Experience answers that question – software doesn’t.
- Who picks up the phone when automation gets ignored?
Automated statement requests are easy to send and easy for a busy supplier to ignore. Identification is not recovery. Someone still has to reach the right person, escalate past the unanswered emails, research the discrepancy, and carry the claim to an actual credit memo or refund check. Ask the vendor who does that work, how many people do it, and how they know who to call. If the answer is “the platform handles it,” ask what happens when it doesn’t.
- What did you recover before it was easy to say you could?
Recovery claims are easy to make and hard to substantiate. Any firm can promise AI will find what others missed. Ask for the history: how many health systems, how many years, how many suppliers engaged, and references who will take your call. A track record that predates the AI boom is the one credential that can’t be spun up in a pitch deck.
The bottom line
AI is a great tool. Alone, it is not enough. The right partner brings all four: the process, the technology, the people, and the history of success. At SpendMend, we use modern AI as heavily as anyone in this market… and we run it on thirty years of accumulated knowledge, more than 300 health systems served, and relationships with over 100,000 healthcare suppliers. That’s our answer to all five questions. Ask everyone else the same questions.