The Button Moved

Taped beside the monitor in your scale house there's a sheet of paper in a plastic sleeve. Handwritten, three columns, a thumb smudge in one corner. It says which product code means washed sand, which trick makes the customer lookup find the county, and what order the screens want so the ticket will print. Lift the sleeve and you'll find it's the third version. The first two are still underneath.

That sheet is your operation translating the software to itself. The layers are the tell: the translation keeps expiring.

The spring release ships in spring

The email lands on a Sunday night in May, subject line cheerful: we've refreshed the ticketing experience. Monday at 6:50 the first truck is on the scale, and your operator, who could run the old screen without looking at it, is looking at it. The customer field is now a type-ahead. The product list is sorted by something new. The print button lives under a menu it didn't live under on Friday. Nothing is down. Nothing is broken. By the vendor's definition this is the system working better than ever, and there are seven trucks in line to watch her relearn it.

Notice when this happened. A software company's release calendar follows its own seasons: ship the spring release in spring, the big rewrite when the roadmap says so. Your calendar peaks at the exact same time, because spring is when the ground opens and the trucks run. Nobody at the vendor chose to retrain your scale house on the busiest Monday of the year. Nobody chose not to, either. Your season simply isn't an input to their schedule, and it never was.

And the roadmap itself isn't yours. What gets rebuilt, and when, is steered by the vendor's growth plans and its largest accounts. A 12-truck outfit's vote on whether the button should move rounds to zero.

What a moved button costs

The outage posts get written about the day the system dies. The quieter bill arrives the week after any release, while everything is technically working.

Start with the miskeys. Say the refreshed lookup now grabs the first match, so for three days "Johnson" pulls Johnson Farms instead of Johnson Paving. Fourteen tickets land on the wrong account at the wrong rate before anyone notices. Half underbill quietly. The other half overbill a customer who calls, and now you're writing a credit memo with a straight face about an improvement. A leak, minted fresh by an upgrade you didn't ask for and couldn't refuse.

Then the slowdown. Your nineteen-year operator ran the old screen at a truck a minute from muscle memory. For two weeks she works at the new hire's pace, at the gate, where seconds are queue length. The cheat sheet gets rewritten, again, on her time. None of this appears on any invoice from the vendor, which is why the subscription looks cheap.

You can't opt out. Cloud software updates everyone on the vendor's date, and the classic-view toggle they offer as a mercy lives on borrowed time. The on-premise era was no kinder; it just moved slower. Support for version 7 ends, the upgrade is mandatory, and the upgrade is a small implementation project with the same consultants as the first one.

Underneath all of it is the trade made at the original rollout. When the operation rebuilt its process around the software's screens, which is what that generation of software required, ownership of the process moved. The sequence of motions your operator performs forty times a morning became the vendor's to redesign, and redesign it they do, on their schedule, for their reasons. You rent your own workflow back, and the rent is collected in retraining.

Paper has never shipped an update

Here's the claim your software-minded friends will fight: the point of work deserves an interface that never changes, and the only interface that has never shipped a release is paper.

The objection writes itself. Updates carry improvements, security patches, features people begged for. Freezing an interface means freezing its flaws. All true in the office, where a rearranged report costs nobody anything and a better search is purely a gift. The window is different. At the window, at forty trucks a day, every change is a cost, including the good ones, because the operator's speed is muscle memory and every release resets some of it. An improvement nobody at the gate asked for is indistinguishable from damage for the two weeks it takes to relearn the screen.

The carbon book's interface was finalized before your operator was hired. Ticket number, customer, product, quantity, signature. Her hands have been compounding on that layout for nineteen years, and no release will ever take it back to zero. The owners who kept the book weren't refusing the future. They were declining to let a third party hold the pen on how their own scale window works, and that was a judgment about control. Computers had little to do with it.

flowchart TD
    A["Vendor ships the spring release"] --> B{"Where does the change land?"}
    B -- "the process lives in the software" --> C["Scale window screen changes"]
    C --> D["Retraining at 40 trucks a day"]
    D --> E["Days of miskeyed tickets"]
    B -- "the software reads the paper" --> F["Reading layer updates in the office"]
    F --> G["Window keeps the same book and pen"]

Change should land where the trucks aren't

Software stopped needing to own the process in order to help it. AI can read the paperwork the operation already produces, as written, which means the process can stay the operation's property. That's the arrangement Termco is built on: the scale house keeps its book and the drivers install nothing. The paper gets digitized as it is. A ticket that breaks a rule gets flagged. Billing rules and reference data live in one structured place, and invoice totals assemble from the tickets that cleared. When the software side improves, and it does, the change lands in that reading layer, behind the office door, so the window doesn't hear about it. There's no release email your scale operator has to read, which also means a shop that never digitized at all starts from its own ticket books instead of a retraining project.

Date the cheat sheet

The check takes five minutes and no login. Walk to the scale house and lift the plastic sleeve. Count the versions and date them; each layer is a retraining your busiest person absorbed in silence. Then find the vendor's last few release notes and sort the changes into two piles, ones your operator asked for and ones she just got. The second pile is the rent.

If you kept the paper, your window's interface has run thirty years without a release, and whatever needs fixing lives downstream of the drawer. Send 10-20 sample tickets to hi@termco.ai and findings come back within 48 hours.

Either way, the button will move again. The only question your operation gets to decide is whether it moves at the window, in front of seven idling trucks, or behind the office door, where it costs minutes instead of a truck line.