Every fax that came in, and what happened to it.

Reading it

Owed, done, remaining.

A ledger is only worth reading if the three words are exact. Most automation reports the middle one and lets you assume the other two: it will tell you how many documents it processed, and leave you to work out what processing meant and what was left over. Abel keeps all three, continuously, because they are the same three the practice already keeps in its head. A number you have to ask for is a number nobody looks at.

Owed

Work your own rules say has to happen because something arrived. Nobody has to notice it or start it — the arrival is what makes it owed.

Done

That work carried to the end state the rule names. Not read. Not classified. Not moved to another queue for somebody else to finish.

Remaining

Owed minus done. The only number that tells you whether the work is actually being handled, and the only one that can be checked against the pile on the desk.

How it works

How Abel knows work is owed before anyone starts it.

Most systems wait to be given work. Abel derives it. That is the whole reason the ledger can be complete rather than a report of whatever somebody happened to click: a job nobody started is not missing from the count, it is on the ledger as owed and not yet done.

1. 01

   Something arrives. That is a fact.

   A fax lands on one of your lines. Abel reads it once and writes down what it is — whose patient, which provider ordered it, what kind of document, whether anything on it is outside the normal range. One read, recorded, never guessed at again later.

2. 02

   A rule says what that fact obliges.

   Your rules, in your words. A normal result gets filed to the chart. An abnormal one gets filed and the ordering provider gets a text. A records request gets answered. The rule is what makes the work owed — not a person noticing the fax and picking it up.

3. 03

   Owed minus done is what is left.

   Abel does the owed work and records that it did. Anything owed and not yet done stays open, and open is a line on the ledger. Nothing has to be queued, assigned, or remembered by anyone for it to be counted.

The 353 Abel handed to a person, and why.

A stop is not a failure — it is the reason this line is worth publishing. Abel will not pick an answer in order to keep moving. When it cannot finish something honestly it stops, writes down why in plain words, and texts the person whose job it is. Each of the 353 carries its own reason in the console. What follows is the kinds of reason, not a count of them — nothing has measured how these 353 split across the four, and a page arguing for real numbers should not invent one here.

• A result a clinician has to see

  A value outside the normal range. Abel files the document and texts the ordering provider; the decision is theirs and was never Abel’s to make.

• No one matching patient

  The name and date of birth on the page did not resolve to exactly one chart. Abel will not choose between two people, so it hands the fax back with what it found.

• A document nobody has walked it through

  A form Abel has not been taught. It goes to the staff worklist unchanged — and stops arriving there once someone shows Abel the job once.

• The next step is a judgement call

  Scheduling, a coverage question, something that needs a phone call. Abel prepares the work and asks the person whose job it is.

Most of these shrink as Abel is taught more of the practice's work — a form nobody has walked it through yet stops landing here the moment someone does. One kind should not shrink. A result a clinician has to look at is not a job to automate away; it is the job, and putting it in front of them quickly is the point.

And the 235 that did not finish.

5.1% of the owed work reached no end state at all. 18 of those ended in an outright error; the rest stopped short of the state their rule names. That line is on this page for the same reason the other two are: a completion figure with its failures lifted out of the denominator is not a measurement, it is a brochure.

What happens to them inside the practice is what happens to any stop — the work falls back to your staff exactly as it would have without Abel, and the run sits in the console with the step it got to and the screen it was looking at. Nothing is retried forever in the background, and nothing that did not finish is written down as though it had.

This is also the number we expect to move, and the one we work on. When it moves you will see it here, with the count underneath it, the same way.

The console

Every line on the ledger opens.

A figure you cannot open is a figure you have to take on trust. Any number on this page opens into the runs behind it, and any run opens into this: the steps in order, the screen Abel was looking at when it took each one, what changed in your systems, and what it cost. Including the runs where nothing went wrong.

Every step carries the screen it acted on, so a run can be checked without taking Abel's word for any part of it. This is also how a stop gets read: the run ends where Abel stopped, with the reason in the same plain words it texted to the person.

What we measure

The number nobody publishes is the one you needed.

The accuracy figures published in this category are, almost without exception, measuring one thing: how often the software read a document correctly. That is a real number and it is worth having. It is not the number that decides whether there is still a pile on the desk at five o'clock. We have both, so here are both.

one practice, last 30 days

Document accuracy is a claim about a model. Task completion is a claim about your Tuesday afternoon. We publish the second one, with the counts underneath it, from the one practice we have.

Pricing

One price. It does not move with the ledger.

You should be able to work out what Abel costs you without a call, and it should be the same number in a heavy month as a light one. A ledger you pay by the line is a ledger you would rather not read.