Rosa arrived at a quarter to eight. Always. The janitor opened the door for her before the official time because he liked Rosa and because, in truth, everyone liked Rosa. She put her bag in the left drawer, turned on the computer, put on her reading glasses, and started.
Four hundred applications a month. A few more in September, when deadlines opened. Each application was a PDF, sometimes well made and sometimes a crooked scan of a document filled out by hand in handwriting that Rosa had learned to decipher over the years. The first hour of every morning, Rosa read them. All of them. One by one. She decided if they were type A, B, or C. She decided if they were complete or missing documentation. She decided which department each one went to. And she moved them.
She had been doing this for eight years.
The director called it "the triage process." Rosa called it "reading the papers."
What Rosa knew without knowing she knew it
Rosa wasn't an IT expert. She wasn't an analyst. She was an administrative worker, having passed a public exam in 2011 and taken an Excel course in 2014 from which she remembered just enough to make simple tables. But Rosa had something you don't learn in a course: judgment.
She knew that when an applicant wrote "renewal" in the subject but used a new application form, it was almost always a renewal and the applicant had used the wrong form. She knew that documents arriving from Cartagena usually had the ID number in a different place than those from the capital, Murcia. She knew that when the certificate of residency was missing, it wasn't worth asking for a correction if the application was type C, because type C didn't require it even though the form said it did, because the form hadn't been updated since 2019.
All of that was in Rosa's head. In no manual. In no procedure document. The official procedure said one thing. Rosa did another. And Rosa was right.
The day Rosa wasn't there
Rosa retired in February. She announced it two months in advance, as appropriate. The director organized a farewell with a cake and a speech. Someone gave her a travel book. Rosa thanked everyone with a brief smile and went home thinking that finally she would be able to water the plants in the mornings without rushing.
Her replacement, Elena, arrived the following Monday. Elena was competent, organized, and fast on the computer. Much more so than Rosa. They showed her the process. They gave her the manual. They told her to ask if she had any doubts.
Elena had doubts from day one.
An application from Cartagena. Type A or type B? The manual said it depended on the form submitted. But the form was ambiguous. Rosa would have known it was type B because that applicant had submitted a type A the previous year and this was the continuation. Elena didn't know. How could she?
A crooked scan. Elena couldn't read the ID number. She called the applicant. It took a day to get a call back. Rosa would have looked at the back of the document, where that specific city council always put a stamp with a legible ID number.
Three weeks later, Elena was behind by fifty-two applications. Not because she was bad. Because she wasn't Rosa. And the system needed Rosa to work.
The machine that learned from Rosa
A smart classification system doesn't replace Rosa. It learns from Rosa. It feeds on the eight years of decisions Rosa made: which applications she classified as A, which as B, which as C. What patterns the complete ones followed and which ones needed additional documentation. Which form errors were really errors and which were applicants simply using the wrong form but knowing what they were asking for.
Then, when a new application arrives, the system reads it. Classifies it. Detects if documentation is missing. Sends it to the corresponding department. Not with absolute certainty. With a ninety-something percent confidence that a person can review in seconds, instead of reading the entire document.
Elena is still there. But now her job isn't reading four hundred applications a month. It's reviewing the ones the system marks as doubtful. Which are forty. And dedicating the rest of her time to solving real problems instead of sorting papers.
What nobody wants to hear
There is a conversation that organizations avoid. The conversation about what happens when the person holding a process together leaves. Not when the director leaves. When Rosa leaves. The administrator, the technician, the person at reception who knows who to call when something gets stuck. The people who aren't on the organizational chart but without whom the organizational chart doesn't work.
The knowledge of those people is an asset. An asset that doesn't appear on any balance sheet and that disappears when they sign their retirement. Unless someone captures it first.
Not in a forty-page manual that no one is going to read. In a system that does what Rosa did, with what Rosa knew, without needing Rosa to be there.
A gesture
Elena called her on the phone a month after she retired. Rosa answered from the balcony, watering a geranium.
"Rosa, quick question. Do applications from Águilas go straight to urban planning or do they go through licensing first?"
"It depends. If it's rural land, licensing. If it's urban, straight to urban planning."
"And how do I know if it's rural?"
"Look at the cadastral reference. If it starts with thirty, it's urban. If it starts with zero, rural."
Elena took a note. Rosa kept watering. The water fell on the geranium leaves with a sound Rosa hadn't heard at that time of morning for eight years.
Don't leave your knowledge to chance
Automating isn't just about saving time. It's about protecting your company's knowledge. Let's evaluate your processes to see which repetitive decisions we can delegate to AI before your key person retires.
Let's talk about your processes