The Medical Records Chase: Why Your PI Firm Pays a Human $52K a Year to Send Faxes

A managing partner in Texas told me his firm employs three full-time medical records clerks at about $52,000 each. With overhead, call it $210,000 a year. Their job, reduced to its essentials: fax HIPAA authorizations and call providers to ask where the records are.

The medical records chase is the most automatable function in a personal injury firm and, somehow, the least automated. The reason is fear: no partner wants to walk into mediation missing an exhibit. Fair. But fear of a bad outcome is not a reason to keep paying humans to do machine work. It's a reason to build the machine carefully.

What the records chase actually is

Strip the job to its loop and you get five repeating tasks: generate a templated request, send it through whichever channel that provider answers, log it in the case management system with a follow-up date, call or re-fax when the deadline passes, and when records arrive, tag them and flag the gaps (missing dates of service, missing imaging, missing billing records). Not one of those steps requires human judgment.

So why hasn't it been automated? Because the data is plumbed into eight different systems: the CMS/CRM, the fax service, phone logs, scanning, billing, email, and the spreadsheet where somebody tracks which providers prefer which channel. The clerks are the integration layer. Their real job is holding eight systems together in their heads.

Why most "records automation" products miss

Two categories of existing tools, two ways to miss. Request generators automate the first 90 seconds of a six-week process and declare victory. Portal networks like ChartSwap and ChartRequest work well for providers already in the network, which by definition are the responsive ones. The providers driving your timeline are the ones who answer nothing but a fax machine and a persistent human. The real solution has to behave like your best clerk: know each provider's channel preference, know when to escalate, know when a subpoena conversation starts, and keep one eye on the statute.

The build that actually replaces the chase

• Day 0. The AI reads the case type, incident date, provider list, and statute of limitations, generates HIPAA-compliant requests, and routes each one by that provider's historical send-and-receive record.
• Day 3. Requests logged in Litify or Filevine with a 14-day follow-up clock running.
• Day 14. Inbound channels checked; no response means an automatic follow-up on the provider's preferred channel.
• Day 21. Escalation to voice: the AI phone agent calls the provider's records department.
• Day 28. The case flags for human review with a subpoena recommendation attached.
• On arrival. Records tagged, gap analysis run, missing dates of service and imaging flagged.

All of it happens in the background, with no human keystrokes until an escalation or decision point actually needs one.

What the supervisor catches

Every outbound request passes a Communications Supervisor that checks for missing fields, date mismatches, wrong provider names, and incomplete redactions. A wrong request pauses for approval; proven categories auto-send. This is the part I insist on: 100% draft mode on day one, with auto-send phased in by category as accuracy proves out. Any vendor promising full automation on day one is asking your firm to be the test environment. That's the difference between AI Fusion and a liability with a dashboard.

What the math actually looks like

Three clerks: $210,000 a year. The AI Super Agent handling the same function: $84,000 to $108,000 a year. Real savings, but the smaller half of the story. The bigger half is time: firms running this build compress records timelines from an average of 142 days to 89. That's 53 days across roughly 400 cases a year, over 21,000 case-days freed. For a firm fronting case costs, faster records means faster demands, faster settlements, and working capital that stops being a constraint. Plus one quiet benefit that partners feel in their sleep: the statute-of-limitations risk from a human forgetting to follow up goes to zero, because the machine doesn't forget.

What to do with the people

Don't lay off the clerks. I mean that practically, not sentimentally. Your records team holds the provider network knowledge the AI needs to learn: which hospital's records department actually answers, which one needs the fax resent twice. During the 30 to 60 day draft period, they're the trainers, correcting routing decisions and teaching preferences. Then your strongest clerk becomes the records operations lead and the others move to higher-leverage work in case management or settlement coordination. The realistic version of the savings is this: you stop hiring the fourth and fifth clerk you were about to add.

The case for moving first

A 30 to 60 day timeline advantage on records translates directly into faster demand packages, which referral partners and lien holders notice. The market will eventually normalize this speed and the advantage will evaporate. Right now there's an 18 to 24 month window where it differentiates you. Better to build it with your own records team teaching the system than to read about it in a competitor's case study.

Start with the chase, not the chatbot

If you're deciding where AI goes first in your firm, skip the intake widget. The records chase is the highest-leverage, lowest-judgment, most-measurable function you have. Schedule a records workflow audit and we'll map your current chase, your provider mix, and what the 53-day compression would mean for your caseload. Or see how the voice side of the build works.