Let's say her name is Molly. She's the office coordinator for a 30-truck service company. She schedules jobs. She dispatches techs. She assembles invoices from the field tickets the crew texts her. She files compliance paperwork, manages subcontractor certificates, follows up on outstanding invoices, and answers the phone when a customer calls to ask where their tech is.
Molly is good at her job. She knows the crew, knows the customers, knows the rhythms of the operation. She works fast. The problem isn't Molly.
The problem is that every single one of those tasks she does exceptionally well is a task that a machine should be doing. And because the machine doesn't exist, you hired four more Mollys to keep up with volume. And now you have five Mollys, none of whom have time to be exceptional, all of whom are spending their best hours on mechanical work that shouldn't require a human at all.
What Molly actually does all day
Walk through the breakdown. Scheduling and dispatch: at any given hour, Molly is on the phone or in the system moving jobs around based on who called out, which job ran long, which customer upgraded from a service call to a full install. This is not judgment work. It's logistics with a lot of variables. A well-configured dispatch AI does this faster, doesn't get tired, doesn't make the call go sideways when a customer gets difficult.
Invoice assembly: the tech closes the job on their tablet or texts Molly the details. Molly pulls the field ticket, matches it against the work order, checks the parts used, types it into the billing system, and sends it. That process, end to end, takes 20-40 minutes per job at a busy company. An automated billing system does it in seconds when the job is marked complete. No rekeying. No lag. No invoices stuck in the queue over a three-day weekend.
"I was sending invoices for jobs from three weeks ago. By then, customers don't remember the job, they push back on the price, and I spend an hour defending work that took fifteen minutes to do."
Compliance paperwork. Subcontractor certificate tracking. Customer follow-up calls. Each of these, individually, seems like a small task. Together they fill a 40-hour week for someone who is genuinely trying to keep up. When volume grows, you hire another Molly. When that one is full, you hire another. The ceiling on this operation is how many Mollys you can find and afford.
The one-Molly model
Here's what changes when you replace the mechanical work with systems. The dispatch AI handles scheduling, rerouting, and exception flagging. The billing system invoices automatically when a job closes. The compliance agent files from operational data. The certificate tracker sends renewal alerts without anyone checking a spreadsheet.
What's left for Molly? The things that actually required her in the first place: the customer who needs to be talked off a ledge, the tech situation that doesn't fit any rule, the relationship with the contractor who gives you 40% of your commercial volume, the vendor negotiation. The work where judgment and history and human presence are the actual product.
The one-Molly operation isn't lean in the painful sense, it's lean in the precise sense. Every function has a system behind it. The system doesn't call in sick. It doesn't get frustrated. It doesn't make the same data entry error on the fourteenth invoice of the day because it's running on two hours of sleep. It just runs.
The person you actually need
Here's the thing about the Molly who stays: she's not the same Molly. Or rather, she's more of what made Molly good in the first place, now that the mechanical load is off her.
The role that emerges in an AI-run operation isn't "office coordinator." It's closer to operations lead, or systems operator. Someone who understands how the AI systems work, not how to build them, but how to read their outputs, configure their rules, catch their exceptions, and communicate with the team about what they're doing. Someone who can spot when the dispatch algorithm is making a decision that technically matches the rules but misses something a good human would know.
This is a more interesting job than assembling invoices. It's also a harder one to hire for, which is why the companies that build this kind of operation early develop a structural advantage. The person who knows how to run an AI-enabled operation is increasingly valuable and increasingly rare. Finding them, training them, keeping them, that becomes the HR strategy.
This is not about cutting jobs. It's about stopping the wrong hires.
Most trades companies are not trying to reduce headcount. They're trying to grow. They're adding trucks, taking on more commercial accounts, expanding into adjacent markets. The instinct, when volume grows, is to hire support staff to handle the administrative load that comes with it.
The better move is to build the system that handles that load, and then grow into the capacity of the system instead of the capacity of the headcount. A dispatch AI doesn't need a bigger desk when you add ten trucks. A billing system doesn't need a raise when invoice volume doubles. The marginal cost of growth drops toward zero for the functions the system runs, and every dollar of growth goes further.
The companies winning in the trades right now are not the ones with the most Mollys. They're the ones who figured out that five Mollys running on manual systems couldn't outperform one sharp operator running good ones.