The 100% Trap
Full automation is a beautiful demo and a terrible target. The first 80% of almost any workflow automates fast and cheap, because it’s the repeatable core: same inputs, same rules, same outcome. The last 20% is where the weird lives: the payer with the fax-only policy, the patient with three insurances, the note that needs a human read. Chasing that tail costs more than the whole first 80 did, takes longer, and produces brittle automation that breaks on every edge case it meets. Teams that aim for 100 ship late, overspend, and often ship nothing, because perfect became the enemy of Tuesday.
80 Out of the Gate Is the Standard
Our operating rule is simple: if 80% of a process can be automated now, automate it now, and bank the return while it’s available. The remaining 20% doesn’t get ignored. It gets triaged, on purpose, into three buckets, and every item in the tail gets a bucket and an owner instead of a shrug.
Triaging the Last 20
Bucket one: human by design. Some of the tail should stay manual forever, because it’s judgment: the coding call on an unusual visit, the conversation with the upset patient, the exception that doesn’t match any rule. This work belongs to people. Marking it as permanently human is a decision, and making it explicitly protects it from bad automation later.
Bucket two: exceptions, routed. Most of the tail is items the automation recognized it couldn’t handle. The build isn’t “automate them anyway”; it’s a clean exception queue where the machine hands each one to a named person with the context attached. The machine does the sorting and the fetching, the human does the deciding, and nothing falls between them.
Bucket three: the bridge. Whatever is repetitive but still stubbornly manual runs on the cheapest reliable labor you can supervise well, often offshore, until the automation catches up to it. The bridge is honest about being temporary: it’s a holding pattern with a price tag, revisited quarterly, and each quarter some of it graduates into the automated 80. The 80% line isn’t a ceiling. It’s a starting standard that creeps upward every review.
The Sentence Every Candidate Must Finish
Before any automation gets a dollar: this change deletes these hours, or moves this number, by this date. The 100% chasers can never finish that sentence for the tail, because the honest answer is “years and maybe.” The 80% standard finishes it every time, which is why it ships, returns, and compounds while the perfectionists are still in requirements meetings.
Check Yours This Week
Pick one workflow you’ve been meaning to automate and split its monthly volume honestly: what share is repeatable core, what share is genuine judgment, what share is exceptions, what share is repetitive-but-hard. If the core is anywhere near 80, you have your green light and your scope in one look. If someone on the team argues for waiting until the whole thing can be automated, show them the tail’s price and ask who’s funding the wait.
Where to Start
Start with the workflow whose 80% sits closest to cash, and run the buckets on the rest. The automation map covers which territory to enter first, the robots guide covers what your people do with the hours that come back, and the variability guide covers why the automated 80 gets more consistent along with faster. Or grab 30 minutes with us. Prep nothing. We’ll walk a real workflow through the 80-and-the-rest triage on live operations, and you’ll see where your own 80 sits.