The build is now the cheap part

A plumbing company owner recently described on X how he automated most of his office’s work-order handling for about $200 a month.

His situation will sound familiar to anyone who runs a trade business with commercial accounts. His biggest customers are facility maintenance companies, warranty companies, and property managers. Every one of them sends work through its own app or portal, with its own login and its own rules. Nobody consolidates them, so his office staff did: watching each portal, re-typing each work order into their scheduling software, calling the homeowner, then typing the appointment back into the portal so the customer could see it.

He decided to automate the whole chain. He is not a software developer. He spent about a week connecting the pieces with one AI tool and got most of the way there. The part he could not crack, driving a web browser inside his customers’ portals, he finished on a newer agent platform in a single evening. By his account, a bot now reads incoming work orders, checks the calendar, confirms the customer, and books the job, and real plumbers work those jobs daily.

Every number in that story is his, and we cannot verify any of it. But we build systems like this, and the shape of the story rings true.

Why the build got cheap

For decades, the expensive part of connecting business software was writing the connection. Each pair of systems needed custom work, so integration projects billed in weeks and months, and most small businesses never bothered.

That cost is collapsing. Modern AI tools already speak to email, chat, and the major field-service platforms through published interfaces. The newest ones can operate a web browser, which means even the client portal with no interface is reachable. A determined owner with no software background can now do in an evening what used to be a consulting project.

That is good news. If you have the itch, build it. We teach owners and teams to build, and we want more builders in the trades, not fewer.

What his post never mentions

Now read the story again and notice what is missing.

• Who notices when it stops? A client portal is not a stable surface. The company that owns it redesigns a page, and the bot that reads it goes quiet. Work orders arrive and nothing happens. His commercial customers grade him on response time, so silence is the most expensive failure he can have. The post names no alarm.
• Where is the record? Sooner or later the bot books a technician to the wrong address on a warranty job. Someone has to reconstruct what it read, what it decided, and why. A booking with no trail behind it becomes an argument with a customer that you cannot win.
• Who holds the passwords? Six systems, including his customers’ own portals, now have their credentials stored inside a third-party AI platform. His agreements with those facility and warranty companies almost certainly have opinions about who may access their systems, and how. That question is worth answering before the customer asks it.
• What does it cost next month? The platform charges by usage, and the work is repetitive and daily. His own post hopes the plan “will be enough.” Hope is not a budget line.
• Who else understands it? He built it himself, and the manual process it replaced is gone. The week he is unreachable, the office is not running the old system or the new one.

None of these questions stops the project. Each one has a known, ordinary answer. But the answers are operating work, not building work, and no platform sells them in the box.

The pitfalls you cannot see from inside the build

The five questions above are about the system. The next four are about the builder, and they are harder to spot because nothing breaks.

• Nobody is asking you the hard questions. Building alone means your plan never gets challenged before you execute it. The expensive gaps are the things you did not think of, and you cannot review your own blind spots.
• The loudest pain is not always the best build. The plumber automated the process that annoyed his office most, which is fair. But was it the highest-return automation in the shop? An owner deep in one build stops surveying the rest of the business, and the survey is where the money is.
• The half-built shop. The owner who automated one process still hand-does a dozen repetitive tasks right next to it, because the next build never reaches the top of the pile. The tool that booked the work order could draft the quote follow-ups too. Knowing how to build and knowing what to build next are different skills.
• The builds pile up. Every automation you keep is a system you now operate. At one, that is a hobby. At five, you are running a small software company on nights and weekends, next to your real one.

Owning the outcome and owning the system

Good software teams settled this a long time ago. Making a thing work once is a demo. Keeping it working while it touches real customers, real money, and real schedules is a separate job with separate habits: monitoring that raises its hand when the system goes quiet, a log that can reconstruct any decision, credentials kept where you control them, a cost ceiling you chose on purpose, and a second person who knows how it all runs.

This is why our method ends with a run phase and not with a launch. The automation you build in an evening is real. The automation your business can lean its weight on has those answers written down, and someone accountable for each.

So build it. Keep both lists from this post taped above your desk, and when one of the pitfalls starts to feel familiar, treat that as information, not failure. Owning the outcome and owning the system are different jobs, and only one of them has to be yours.

Related: What a Company Brain is