Thomas has been in the trades for 25 years. He didn't set out to build software. He set out to fix a specific problem: the construction industry's assumption that extreme overtime is just how the business works. What he built in the process is worth understanding, because it changes what's possible for contractors who have no coding background but a clear picture of what they need.

Using Claude as a coding partner, Thomas built a fully custom project management and accounting system that has replaced every third-party tool in his business. His crew works 7am to 3:30pm. Onboarding a new hire takes one button. AI agents handle 80 to 90 percent of his accounting by monitoring his email and classifying receipts automatically. Field crews clock in with one tap because the system already knows their schedule and location. This issue covers how he built it, what it actually does, and what contractors without a coding background can take from his approach.

Thomas looked at the tools available to him — BuilderTrend, QuickBooks, the standard stack — and found that most of them were built by people who had never swung a hammer. They solved problems in the abstract but created friction in practice, and the friction added up to administrative overhead that was eating into the time he wanted to spend on production and with his family.

So he built his own.

Not because he wanted to become a software developer, but because Claude could write the code if he could describe what he needed. He used Convex for the database and VS Code for development, and he talked to Claude the same way he'd talk to a subcontractor: here's the problem, here's what I need, here's what's wrong with this piece of the screen.

The system now runs 100 percent of his business. No third-party tools. No software he didn't build for his specific situation. And his crew goes home at 3:30.

Automated accounting

AI agents monitor his accounting email in the background. When a receipt arrives, the agent reads it, classifies it, and handles the entry automatically. Thomas estimates this covers 80 to 90 percent of his accounting work without any manual input. The system also generates Purchase Orders automatically when a quote is created, sends them to the vendor by email, and tracks the vendor's response to update the accounting side when the reply comes in.

One-button onboarding

New hire onboarding sends I9s, W4s, and background check links with a single action. The system auto-fills redundant data across all forms so the new employee doesn't have to type their name and address multiple times. What used to take an afternoon of paperwork takes minutes.

Field-first time cards

Field crews ranging from their 20s to their 50s use a mobile-native version of the app for time cards, schedules, and employee manuals. Clocking in is a single tap because the system already knows each person's schedule and location. No login required, no navigation, no opportunity for the process to fall apart because a 55-year-old carpenter doesn't want to use software.

Role-specific dashboards

Each role in the company sees only the data relevant to their work. A carpenter sees what they need. An HR admin sees what they need. The Super Admin sees everything. No one is navigating a system built for someone else's job.

Data-driven estimating

Thomas integrated data from Craftsman books to build initial Gantt charts and production timelines, then refined them using his own crew's actual performance data over time. The estimates get more accurate with every job because they're built from real performance history rather than static industry averages.

He also uses video and photos to document site conditions before his crew leaves — what's behind a wall, the width of a gate, the exact state of a surface — so no one has to guess or make unnecessary return trips during trim-out.

Thomas's approach to building with Claude is worth understanding on its own terms because it's replicable by anyone who can describe a problem clearly.

He treated Claude like a trade subcontractor rather than a search engine. Instead of asking general questions, he brought specific problems: here's what this screen looks like, here's what it's supposed to do, here's what's wrong. When a UI element was behaving incorrectly, he screenshotted the issue and gave Claude the visual context rather than trying to describe it in words. That approach, using screenshots to give Claude what it needs to see rather than what's hard to describe, is one of the most practical tips in the episode.

His advice for contractors who want to try building their own tools: go break it. Talk to the system. Describe what's wrong and ask for the fix. The willingness to iterate without fear of breaking something is what separates contractors who build useful tools from ones who stop at the first error message.

1. You don't need a coding background to build with Claude. Thomas has 25 years in the trades and no formal development experience. What he has is a clear picture of what friction looks like in his business and the willingness to describe it to an AI that can write the code. That's the skill. The coding is Claude's job.

2. Screenshots are more useful than descriptions for UI problems. When a button is in the wrong place or text is wrapping incorrectly on an invoice, showing Claude a screenshot of the issue produces a faster and more accurate fix than trying to describe it in words. This applies to any visual problem you're trying to solve with AI assistance.

3. The field-to-back-office ratio is a profitability lever. Thomas's framing of the field-to-back-office ratio is worth sitting with. Every hour of administrative work that gets automated is an hour that doesn't require a back-office hire. A system that handles 80 to 90 percent of accounting automatically is a system that changes the staffing math of the whole operation.

4. Start with your biggest friction point, not the full system. Thomas built toward a complete system over time. He didn't start there. Identify the one administrative task that costs you the most time or the most money and build a tool for that specific problem first. A working solution to one problem is more valuable than an unfinished solution to all of them.

5. The goal is reclaiming time, not building software. Thomas's crew goes home at 3:30. That's the metric. The software exists to serve that outcome, not as an end in itself. Before you build anything, be clear about what you're trying to get back: evenings, weekends, the ability to leave a job site without your phone lighting up. That clarity is what keeps the project focused.

Beyond the Bid Circle is a peer community for builders scaling with systems and AI—the same builders who've escaped the 60-hour grind. Swap war stories, shortcuts, and real strategies with people who get it.