Your Service Writer Types Every Estimate Twice — Once Into Fullbay, Once Into the Fleet's Portal

You already bought this once. Somebody sold your shop a connection between your shop management system and the fleet portal, it worked for about five months, then a screen changed on the portal side and it quietly stopped sending. Nobody noticed until a fleet uptime manager called asking why unit 8817 had shown "in diagnosis" for six days. You paid to have it fixed, it broke again the following spring, and you went back to having a person type it twice. That is a completely reasonable reason to be skeptical, so let me be specific about what is different in 2026 and what is not.

Time the double-entry with a stopwatch before you argue about it

Stand behind a service writer on a Tuesday and watch one national-account case from arrival to approval.

He builds the estimate in the shop system — Fullbay, Karmak Fusion, Procede Excede, Mitchell 1 Manager SE, Trimble TMT, whichever you run. Labor lines, parts, the complaint-cause-correction, the estimated repair complete date. Six to ten minutes of real work he should be doing.

Then he opens the fleet's case in the portal — the Volvo or Mack side, Navistar OnCommand Connection, or whatever service platform that customer requires — and types the same lines in again. Same parts, same hours, same estimated completion. Four to eight minutes of work that produced nothing new in the world.

Then, while the unit is down, the fleet expects a status update every four to eight hours. Somebody has to go back into the case and move the estimated repair complete date when the wheel seal is on back order. Three to six of those per case. Then the invoice gets keyed again, and if the job is warrantable, the narrative gets typed a fourth time into the manufacturer's claim system.

Call it fourteen minutes of pure re-typing per fleet case, not counting the status updates. At a shop closing 240 repair orders a month with 60% of them on national accounts, that is roughly 34 hours a month of a service writer's time producing exactly zero new information.

What actually changed in 2026

Two things, and only one of them is technical.

The first is that the industry settled on one common plug between AI assistants and business software — the Model Context Protocol, usually just called MCP. In plain terms: instead of every AI tool needing a custom, brittle connection built by hand to every piece of software you own, there is now one agreed way to plug in, and the software vendors are the ones building the socket. That is why the 2024 version broke and this one is less likely to: the connection is the vendor's product, not a bespoke job somebody did for your shop.

The second is that the big business-software vendors shipped their own AI workers sitting on top of it. Oracle shipped AI Agent Studio for its Fusion applications, Microsoft shipped a Sales Agent and a Service Agent, Salesforce wired the same idea into Slack. None of those are truck shop products. What they establish is the pattern every vendor is now copying: the assistant reads and writes inside the system of record itself, rather than living in a separate tab where a human is the connection between the two.

flowchart TD
  A["Tech finishes diagnosis, writer builds the estimate in the shop system"] --> B["Assistant reads the estimate lines"]
  B --> C["Same lines posted into the fleet case in the service portal"]
  C --> D{"Fleet maintenance director responds"}
  D -->|Asks a question| E["Answer written back onto the repair order, writer notified"]
  E --> C
  D -->|Approves| F["PO number written into the shop system"]
  F --> G["Parts released, bay time scheduled"]

Tuesday, 7:15am, from the service manager's chair

Here is what the morning looks like when the second typing is gone. Your foreman assigned three units overnight. The writer opens his board and sees eleven active cases, and next to each one, the current status in the fleet's portal — not a note saying he needs to go update it.

Unit 8817's wheel seal shipped from the supplier at 6:40am, so the estimated repair complete date moved from Thursday to Wednesday in the shop system, and it moved in the fleet's case too, without anyone opening a second window. The fleet's uptime team saw the change before your writer had coffee, which means they did not call.

The estimate on the Cascadia went into the case at 4:50pm yesterday. The maintenance director approved it at 9:10 last night from his phone and typed a purchase order number into the case. That number is now on your repair order. The parts counter released the parts at 7:05 this morning without waiting for a writer to arrive and read an email.

What the writer actually does at 7:15 is the part that needed a human all along: he calls the one fleet contact who asked a question about why the aftertreatment cleaning is not covered, and he decides which of three units gets the last open bay.

The number that matters is not keystrokes

Most vendors will sell you this on time saved, and the time is real — 34 hours a month is a third of a service writer. But the number that moves your business is when the invoice goes out, because a repair order that sits waiting for an approval that is sitting in somebody's email is cash you have already spent on parts and labor.

Stated assumptions, illustrative figures. The honest test is the one you can run yourself: pull thirty closed national-account repair orders and measure the hours between "estimate ready in the shop system" and "estimate visible in the fleet's case." If that gap averages under twenty minutes, you do not have this problem and you should stop reading. In most shops I have watched, it is hours, and on a Friday afternoon it is Monday.

The question to ask your vendor, in these words

Do not ask whether your shop system "has AI." Every vendor says yes and it means nothing. Ask this instead: does your product support the Model Context Protocol, and if so, what can an assistant read and what can it write? Then ask the same question of whoever runs the service portal your biggest fleet customers require. If both answer yes, you can remove the second typing this year. If one answers no, you have found the thing to raise at their user conference, and you have found a real reason to prefer a competitor at renewal.

As of July 2026 the answers vary widely across heavy-duty shop systems and service platforms. Anyone who tells you it is universal is selling. The point is that the question now has a specific technical answer instead of "we could build you something custom."

Where you keep a human, permanently

Give it read access broadly and write access narrowly. Start with status updates and estimated repair complete dates — nobody has ever been sued over a completion date being posted automatically, and status updates are the highest-volume, lowest-risk keystrokes in the building.

Do not let it approve an estimate, change a purchase order amount, or close a repair order. Those three actions change what a customer owes you, and every one of them should have a name attached. Do not let it file the warranty claim either — the narrative is an argument, and arguments get written by the person who will have to defend them when the claim comes back denied.

And keep the phone call. When a fleet maintenance director declines an estimate, the reason he gives the portal is never the real reason. The real reason comes out when your service manager calls him and asks what the truck is actually worth to him this quarter. No amount of connection between systems replaces that call.

Frequently asked questions

We run Karmak Fusion, not Fullbay. Does any of this apply?

The workflow is identical regardless of which shop system you run — the double-entry is between your system and the fleet's portal, and every heavy-duty shop system has that problem. What differs is the answer your vendor gives to the question above. Ask it in writing, keep the answer, and ask again at renewal.

If the assistant posts a wrong estimate into a fleet's case, who is on the hook?

You are, exactly as you would be if a writer keyed it wrong. That is why the write access starts narrow. Posting a status change is reversible and low stakes; posting an estimate the customer then approves is a commitment. Move estimates into automatic posting only after you have run sixty days of status updates without a bad one, and even then have the writer press the button.

Will fleet customers accept updates that were not typed by a person?

They already prefer them. A fleet uptime manager watching forty units in twelve shops cares about one thing: whether the estimated repair complete date on the screen is true right now. An update that is accurate at 6:40am beats one a writer typed at 4pm yesterday from memory. The complaint you will get is about updates being wrong, never about them being automatic.

What do we do first, this month?

Measure the gap. Thirty repair orders, two columns, estimate ready and estimate visible to the fleet. Take that number to your next vendor call. You cannot negotiate for a connection you have not priced, and "our writers spend 34 hours a month re-typing" is a sentence that gets a vendor's attention in a way that "we would like better integration" never has.

The calls that survive all this

CallSphere builds AI voice and chat agents that answer business phone lines and web chat, book appointments and capture leads around the clock. Even when your systems stop making people re-type, the phone keeps ringing — owner-operators who are not on any portal, drivers asking whether the truck is ready, first-time callers with a check engine light on the shoulder of the interstate. Those calls do not arrive through a fleet case, and they are the ones most likely to go to voicemail while your writers are heads-down on the board.