The prior auth doesn't move in your EHR. It moves between payer portals, faxes, and desks.
The same chart fields retyped into a fourth payer portal. The referral that went out and nobody confirmed arrived. The denial that resets the clock because the packet has to be rebuilt from scratch. Below is what that work looks like today, what it looks like once Silow has mapped it, and what it gives back — inside your own perimeter, with no clinical decision touched.
The answer to the question your privacy officer is about to ask.
A tool that records how people work is exactly the tool a HIPAA-covered operation is right to be suspicious of. So here are the boundaries first, in writing, before a single claim about prior auth, referrals or billing.
Zero-access mode: Silow runs air-gapped on-prem or inside your own VPC. Anything touching a chart extract, a payer portal or a claim is processed in place. Silow does not receive, access or store your raw captures. This is the mode we point PHI-handling teams toward, and the only mode in which we will ever say that sentence.
These are product constraints, not policy promises. We never record what is typed, nothing is captured from camera or microphone, and private communications are not read.
No productivity ranking, no performance evaluation, no automated decision about anyone's job. The unit of analysis is the process — in our AI-adoption report, not a single personal name appears in the underlying facts at all.
We never publish a blanket "Silow never sees your data." That is only true in zero-access mode — and that is the only place we will ever say it.
How the work looks today. How it looks after.
Seven processes, in the order we would take them. The left-hand column is what teams we recorded were observed doing. Rows tagged "Measured" carry a number that came out of that recorded work; rows tagged "Observed in the trace" are patterns seen in the recording without a standalone metric.
The same clinical facts get retyped into whichever portal that payer happens to use this year, or faxed on a form that hasn't changed since it was printed. A case is "in progress" for days and is actually being worked on for a few minutes of it — the rest is the portal loading, the fax confirming, or the case sitting in a queue nobody is looking at.
The known fields fill themselves from the chart extract, in the shape each payer's form or portal expects. What reaches the reviewer is a packet with the gaps flagged, not a blank form and a stack of source documents.
In a recorded team, one document was manually located and re-attached across two messengers on five separate occasions in three weeks, because no shared link for it existed. In a prior-auth or referral workflow that document is an insurance card, a chart note, an EOB, a referral form — and the request for it is retyped from scratch every time, by someone who already knows what it will say.
Standard requests draft themselves from the language your team already uses, with the right attachment already attached. The human reviews and sends. The file has one retrievable place, so it is never hunted for again.
Work that needs a second person — a signature from the ordering clinician, a callback from a payer, a reply from a specialist's office — waited a median of about seven days in the teams we recorded. Work one person could finish alone waited about one. Roughly two thirds of all waiting sat on the far side of a handoff, and it was not schedules or time zones: the least-overlapping pair still shared a quarter of the working day. The handoff is simply not defined anywhere.
The handoff becomes an object rather than an assumption. Nothing sits silently, because "waiting on someone" is a state the case is in, not a thing you remember to check.
A referral goes out by fax or portal upload and nobody confirms it arrived. Someone calls the specialist's office to check. The patient calls to ask what's happening. The confirmation, when it exists, lives in a note in a different system than the referral did.
The referral's status is tracked as a matter of course: sent, received, scheduled, seen — drafted from what the tools already show, with a human confirming anything that reaches a patient or another office. Nothing auto-sends outside the practice.
A denial arrives with a code and a boilerplate reason. Someone re-assembles the original submission plus whatever additional documentation the payer wants, from scratch, against an appeal window that is already ticking.
The appeal packet pre-assembles itself from the original submission and the payer's stated reason, with the gaps listed. The clinical argument, and the decision to appeal, stay with the person who is supposed to make them.
A task is moving about 3% of the time it exists in the teams we recorded — the other ~97% it sits in someone's queue. In a prior-auth or billing operation, that queue is where your turnaround time and your appeal window actually live, and it is invisible to every tool you own.
The queue becomes visible: what is waiting, on whom, since when, and what would unblock it. You are not making anyone faster. You are attacking the ~97%.
Counting AI by application name undercounted every person in a team we recorded, and one of them by around seven times — their assistant ran inside another tool, so the window reported the host's name. If you bought licences for the billing team and the dashboard says nobody uses them, the dashboard is the thing most likely to be wrong.
Adoption measured from the work itself rather than from seats and logins — and, more usefully, the reason. An assistant nobody opens is almost never a model problem. It is that nobody mapped it onto a procedure anyone actually performs.
The teams behind these rows: a recorded team of under ten people in document-heavy, regulated coordination work over five weeks, and a cross-border operations team we also recorded. Both anonymized. Rows tagged "Measured" carry a number from that work; rows tagged "Observed in the trace" are patterns seen in the recording without a standalone metric. Figures are rounded, and every one was computed before a model was allowed to write a sentence about it.
What an agent may touch, and what it may not.
In healthcare admin this line is not a nicety, it is the entire premise. Silow does admin assembly around prior auth, referrals and billing. It does not touch a clinical decision, and we draw that line before anything is built, not after.
- Assembling the prior-auth packet: chart-extract fields matched to the payer's form, attachments collected, gaps flagged.
- Drafting the request for the missing document or the payer-specific form the team fills out every month.
- Watching the queue: which authorization, referral or claim has been waiting, on whom, for how long.
- Producing the record of what moved, when and in what order — as a by-product, not a task.
- Pre-filling the denial-appeal packet from the original submission and the payer's stated reason.
- The reporting ritual: the recurring pull from the practice-management system and the payer portals into one sheet.
- Every clinical decision: diagnosis, medical necessity, treatment plan. Silow never makes one and never suggests one.
- The prior-auth submission itself. A human reviews and submits; nothing auto-submits to a payer.
- The denial appeal's clinical argument. Silow can assemble the packet; it does not write the argument.
- The exception nobody has seen before — which is most of what you hired a coordinator for.
- Any decision about a patient, a case or an employee. No ranking, no scoring, no automated judgement about a person.
No clinical decisions, no diagnosis, no treatment recommendations. Silow removes the assembly work around a submission and leaves the clinical judgment, and the human gate on what actually goes to a payer, exactly where they were.
Time, money, and the people you already have.
The rule for this section: not one invented percentage. Every figure is either measured in a recording or supplied by you — hours multiplied by your own loaded cost. A saving you cannot defend to a compliance officer is not a saving, it is a slide.
We do not quote you an industry percentage. Every candidate above gets scored in hours against your own recording and ranked by payback. The anchors we can already point at: a task moving about 3% of the time it exists, a median wait of around seven days each time work crosses a desk, and one document re-found and re-attached five times in three weeks.
The first is hours × your loaded cost — your rate, not a benchmark we invented. The second is the one nobody counts: the AI you already pay for and nobody opens. In a team we recorded, app-name telemetry undercounted AI use for every single person, so the licence decision was being made against a number that was simply wrong.
No new system to adopt, no data to move, no process change while we look — Silow runs on the tools the operation already uses, inside the perimeter you choose. The capacity comes back into the same team, doing the same job with the assembly work removed.
A task is moving about 3% of the time it exists. Making the coordinator faster optimises that 3%. Your turnaround time lives in the other 97%.
From a paperwork path nobody has seen to a layer you own.
The agents are the visible part. The thing that makes them possible — and that keeps paying after they ship — is the record of how the operation actually gets a case through a payer.
It is written down somewhere as an SOP, and the SOP skips the fax numbers, the portal quirks and the payer-specific forks. The real path — the one taken at 09:00 on a Tuesday when the portal is down — has never been observed, which is why every automation attempt starts with a workshop and a guess.
The real steps, from the recording, with the evidence attached. In one of our own recordings all thirty supporting quotes were checked back against the raw scenes they came from, not against a summary of them. Ranked by impact, effort, risk and payback.
Prior-auth packet assembly, the document chase, the referral-status draft. Each arrives as a specification an engineer can build from: the observed steps, the evidence, and the lines it must not cross. Human review — and human submission — stay in the loop by design.
The agents are consumers of something more valuable: a private, structured record of how your operation actually gets a case through, living inside your perimeter. Onboarding, search, audit, and whatever you build next, on data only you own.
The admin work nobody has a line item for.
These six categories are not our taxonomy — they came out of the report on an operations team we recorded, ranked from what that team was observed doing all day. They are not healthcare specific, which is rather the point: this is where the day goes in any document-and-approval operation.
The recurring pull-assemble-format cycle across the PM system, portals and sheets.
The chase: who owes what, who is blocked, what has gone quiet.
Prior-auth, referral and appeal cases assembled into one reviewable object.
The pre-read and the post-meeting actions, drafted from what was actually said.
The submission and appeal packs, rebuilt from scratch for every payer and every case.
Onboarding, access requests, and the tool blockers that stall a whole shift.
The part other vendors leave out.
Two of the hypotheses we most wanted to be true were killed by our own data, and one item below is a legal gap, not a marketing choice. The vendor who tells you what their product cannot do — and cannot yet commit to — is the one worth the second meeting.
Zero-access mode is the reason a BAA may not need to cover the raw capture itself — the data never reaches us to begin with. But we are not going to claim a business associate agreement exists as a standard, signed document before it does. We work through the shape of it for your deployment before you sign anything, and we say so here instead of leaving it implied.
The recording sees activity, never outcome. Whether the authorization was approved, whether the claim paid — that is not in the data. Join one column of your outcomes to our trace and it becomes provable. Until then, any ROI figure quoted at you by anyone is a guess.
A duplicate-work detector: about 44 recoverable minutes across nine people over five weeks, so there is no product. A meeting-ROI report: three quarters of what the metric called "meeting time" had no room attached. Both were cut before they ever reached a customer.
Numbers are computed before the model is allowed to write a sentence about them, so no figure gets invented. But a true number can still be attached to the wrong claim. We have caught exactly that twice — both times with human eyes, not with a check.
Days for legal. Weeks for the roadmap.
Before any recording, we agree the mode. For PHI-handling teams that is usually zero-access — on-prem or your own VPC — with the DPA, subprocessor list and boundaries on the table at the first security call.
Silow runs on the team's existing machines and tools. No migration, no new system, no process change while we look — the point is to see the work as it actually is.
Scenes become tasks; tasks become the real procedure. Every opportunity scored by impact, effort, risk and payback against your own hours.
A specification, not a suggestion: the observed procedure, the evidence, the scenes behind it, and the boundaries it must not cross.
What you do with the capacity is yours.
Silow maps work, not workers, and it does not touch a clinical decision. If the honest answer for your operation turns out to be "there is not enough here to be worth it," we would rather find that in a two-week recording than sell you a year of it. Book a call and we will show you what the map looks like on your own tools, and what it would rank first.