Case study · Healthcare · Invoice Processing
Healthcare Invoice Processing: Days → Hours
The whole build: 3 ai agents, 5 tools, and the before-and-after on metrics the team already tracked.
Invoice Processing Time
Process today's healthcare invoices.
Acme Supplies
PO-4471 · matched
Northwind Co.
PO-4472 · matched
Initech LLC
price +6% vs PO
Meridian Group
no PO on file
3 AI agents · 5 tools connected · live in 2 hours · no code
- Company
- Mid-size healthcare provider
- Team size
- 30-150 employees
- Industry
- Healthcare
- Time to live
- 2 hours
- Agents deployed
- 3 AI agents
- Tools connected
- 5 integrations
The context
Why Invoice Processing is hard in healthcare.
Invoice Processing is not hard in the abstract. It is hard in healthcare, where the work arrives as referral faxes, patient portal messages, phone calls, and EHR work queues — every channel a different shape, none of them waiting their turn. The team runs against the appointment book and the payer clock, so the real cost of a slow invoice processing step is never the step. It is a record touched by the wrong hands, or a detail dropped between a referral and a chart.
Constraints the build had to hold
Protected health information
Agents were scoped to the records the staff account could already see. Throughput widened; access did not.
The EHR is the system of record
Nothing counts as done until it is in the chart. Agent queues are scratch space, never a second truth.
Clinical judgement stays human
Agents move information and prepare decisions. Anything clinical goes to a person with the context attached.
The change
Same job. Two chains.
Every handoff in the left-hand chain is somewhere Invoice Processing used to wait. The right-hand chain has the same steps and none of the waiting.
By hand
- Invoice hits the shared inboxin any format a vendor invents
- Header and lines keyedthen matched to the PO by eye
- Approval walked aroundstalls on one traveller
- Posted, days later
discount window already gone
With agents
- Work arrives on any channelpicked up in seconds
- Capture agenthanded straight on
- Matching agenthanded straight on
- Approval agent
logged, and reviewable
When the work can happen
Before and after
What Invoice Processing cost them, and what replaced it.
The challenge
Accounts payable at this mid-size healthcare provider ran on paper habits inside digital tools. Invoices arrived in a shared inbox in every format a vendor could invent, and a person opened each one, keyed the header and lines, matched them against the purchase order by eye, and walked the approval around by email.
The full background
At their volume that was a full-time job with a permanent backlog behind it. The healthcare finance team measured processing in days, not minutes, and it showed at month end — close slipped because the ledger was still catching up on invoices received three weeks earlier. Early-payment discounts expired unclaimed because nobody could see which invoices were close to the window. And the control that mattered most, catching a price or quantity variance against the PO, depended entirely on whether the person doing the keying happened to notice.
What they built
The finance team put DeskFerry between the AP inbox and Epic. Invoices are captured on arrival whatever format they come in, matched line by line against the purchase order and receipt, and posted when they agree.
How it was wired
The design point was that matching produces a number, not a flag. A variance is quantified — this line is 6% over the PO price, this quantity is short by two — and routed to the approver who can act on it, with the difference already highlighted. Clean invoices post without anyone touching them; the small proportion that genuinely need judgement arrive pre-analysed. Approvals go out with one-tap actions, chase themselves on a schedule, and escalate before the healthcare payment term expires, which is how early-payment discounts stopped quietly lapsing.
The impact
What changed, measured the same way on both sides.
Before and after across the metrics that matter for healthcare Invoice Processing.
Invoice Processing Time
Dramatically faster
Processing Cost per Invoice
Major savings
Error Rate
Major reduction
Early Payment Discounts Captured
Significant increase
Monthly Invoice Volume
Major throughput increase
How these were measured
- Baseline
- The "before" column is the team’s own measurement of their manual invoice processing process, taken over the four weeks before anything was connected.
- Comparison
- The "after" column is the same measurement repeated on the same process once the agents were live, so both sides count the same things in the same way.
- Why no percentages
- These are composite scenarios built from patterns across many deployments, not one audited customer’s books. Directional language is the honest way to report that — your own numbers will depend on your volume, your process, and your starting point.
A day, either side
The same day, before and after.
What Invoice Processing actually looked like for this healthcare team — the version they described in the first call, and the version they run now.
Before DeskFerry
9:00
Invoices arrive in a shared inbox. Print, stamp, key, file.
11:00
Match line items against the PO by eye. Two do not agree.
14:00
Chase an approver who has been in meetings all day.
Day 12
An early-payment discount expires unnoticed.
Month end
Close slips because the healthcare ledger is still catching up.
After DeskFerry
9:00
Invoices were parsed and matched to POs on arrival overnight.
9:05
Clean matches are posted. Variances are queued with the discrepancy highlighted.
11:00
Approvals go to the right approver with one tap, chased automatically.
Day 2
Early-payment discounts are taken because nothing waited in a pile.
Month end
Close runs on a ledger that was never behind.
The build
The 3 agents that run it.
One job each, with an explicit handoff between them. Splitting Invoice Processing this way is what makes a failure legible — you can see which step it went wrong at instead of debugging one agent that does everything.
- 01
Capture agent
Trigger
An invoice arrives by email or portal
Extracts header and line items, identifies the vendor, and normalises the format regardless of who sent it.
Agent 1 of 3 in the Healthcare workflow.
- Passes a structured invoice to matching.
- 02
Matching agent
Trigger
An invoice is captured
Matches lines against the purchase order and receipt in the ledger, and quantifies any variance rather than just flagging it.
Agent 2 of 3 in the Healthcare workflow.
- Clean matches go to posting; variances go to a named approver with the difference highlighted.
- 03
Approval agent
Trigger
An invoice needs a human decision
Routes to the right approver by amount and cost centre, chases on a schedule, and escalates before the healthcare payment term expires.
Agent 3 of 3 in the Healthcare workflow.
How they did it
From nothing to production in 2 hours.
No code, no IT ticket, no vendor implementation team. These are the steps in the order this team took them.
Step 01
Connected the healthcare stack
Epic, Cerner, and Athenahealth via pre-built connectors. No API keys, no custom code.
Step 02
Wrote the business rules
Scoring, routing, escalation thresholds, and exception handling for healthcare invoice processing — in the visual builder.
Step 03
Tested on real history
Replayed a week of past invoice processing to check accuracy and surface edge cases, then adjusted the weights.
Step 04
Launched and watched
Live with close oversight for 48 hours, then down to a weekly review.
The stack
Nothing was replaced. Everything was connected.
The healthcare team kept the tools they already ran — DeskFerry sits between them.
Epic
The chart — read for context, and written back so it stays the record
Cerner
Second system of record; reconciled against the chart before any write
Athenahealth
Scheduling, billing, and clinical work queues
Kareo
Practice management, scheduling, and claim state
Google Forms
Structured intake so the first agent starts from clean fields
Invoice Processing handled end to end · hours, every time
What stayed human
The parts they deliberately did not automate.
Automating Invoice Processing end to end was never the goal. Removing the volume so the judgement calls got proper attention was.
Every variance
Price and quantity differences against the PO are quantified automatically and approved by a person. Nothing that costs more than expected posts without someone agreeing to it.
New vendors
A first invoice from an unknown vendor is always reviewed, which is the cheapest control against invoice fraud that exists.
The HIPAA compliance question
Asked first by every healthcare team. Agents run on the access the staff account already had, every action is logged, and any step can be stopped without unwinding what ran.
Takeaways
What transfers to your team.
The parts of this that are not specific to one company's tooling or volume.
- 01
Connecting the existing healthcare stack beat replacing it.
- 02
Backlogs went away because the work no longer waits for office hours.
- 03
Quality stopped varying by whoever picked the task up.
- 04
Starting from a template and tightening the rules weekly beat designing it upfront.
In their words
“Before DeskFerry, our invoice processing process was the bottleneck that every healthcare team complained about. Now it's our competitive advantage. We process faster, more accurately, and at a fraction of the cost. Our competitors are still doing this manually.”
Composite — written from what teams running this workflow report, not a single named customer.
FAQ
Questions people ask about this build.
Automating Invoice Processing in healthcare — what it takes, and where it stops.
How long does it take to set up invoice processing automation for a healthcare business?
This team was live in 2 hours. Pre-built healthcare templates cover the wiring, so most of that time goes on your business rules rather than on connecting things. No code.
How many AI agents does invoice processing automation actually need?
3 here: capture agent, matching agent, approval agent. The split matters more than the count — one job and one handoff each means a failure tells you which step broke. One agent doing everything does not.
What results can a healthcare business expect?
The figures here are directional, not audited — composite scenarios, not one customer's books. What transfers is the shape: routine volume stops needing a person, exceptions surface instead of sinking, and nothing waits for office hours. Your numbers depend on your volume and starting point.
How does DeskFerry handle healthcare data and access?
Agents run on the same access the staff account already had — throughput widens, permissions do not. Every action is logged with what it read and changed, and any step can be stopped without unwinding what ran. DeskFerry holds no formal healthcare certification, so scope it as you would any other system in your control environment.
What still needs a person?
More than most automation pages admit. Anything outside the rules stops and goes to a named owner with context attached, rather than being guessed at. The rules themselves are changed by people — agents never widen their own tolerances. The carve-outs this team kept are named above.
What tools does this connect to?
1,500+ integrations. This build used Epic, Cerner, Athenahealth, Kareo and Google Forms; most healthcare stacks are a variation on that. CRM, email, chat, databases, and industry-specific software all connect without code.
Run this in your own stack.
Describe how Invoice Processing should work at your healthcare business, in a sentence. DeskFerry builds the agents, wires your tools, and takes the routine volume from there. Start free — no credit card.
Or start from a template — Invoice Processing agent for Healthcare.
Keep exploring
Related case studies.
The same job in another industry, or another job in this one.
More Healthcare case studies
Invoice Processing in other industries
Composite scenario — built from patterns across many healthcare Invoice Processing deployments rather than one customer's books. Figures are directional; your own depend on your volume, process, and starting point.