ServiceTitan said it isn't built for small shops. The small shops heard them.

A contractor paid ServiceTitan for a full year and never used the software. In December 2024 he filed a Better Business Bureau complaint saying his shop had “NEVER BEEN ONBOARDED” while the subscription kept billing. He is not unusual. He just complained on a public record where anyone could find it.

Most small shop owners I have met have a version of that story. The demo was great. The salesperson was great. The quote arrived and the number did not match the business they actually run. They signed anyway, because what else was there.

About 210 contractors a month type “how much is service titan” into Google. ServiceTitan will not tell them. They end up on Reddit reading other contractors guessing. That is roughly where this whole category lives in 2026.

What the bill actually looks like

The base subscription is $250 to $500 per technician per month. A five-truck shop is staring at $1,250 to $2,500 a month before anyone has dispatched a job. Implementation runs $5,000 at the small end and into $50,000 at the top. Going live takes twelve to sixteen weeks, and most shops report six months before the platform is actually doing real work. The contract minimum is twelve months.

Add-ons are sold separately. Marketing Pro, Dispatch Pro, Fleet Pro, Phones Pro, Pricebook Pro. A loaded setup with the modules a shop actually needs can double the monthly number. Early cancellation has been reported at $39,000.

ServiceTitan has written, on its own materials, that the platform is “not optimized for a company with three or fewer technicians.” That is the company’s own line. They have other customers in mind. The small shops keep showing up to the demos because the alternatives are not advertised the same way.

A reviewer on YouTube in May 2025, comparing ServiceTitan against the smaller competitors, said it cleanly: “ServiceTitan was built for enterprise-level contractors with full departments. If a company with six people is getting the same software as a company with huge staff, it’s really not a good fit.” Six people is a lot of small shops in the US.

The actual stack, line by line

A typical small home-services shop in the US runs three trucks, has one office person (often the owner’s spouse), and books five to twelve jobs a day. The software bill, before anyone changes anything, breaks down roughly like this.

Field service software is the largest line. Jobber publishes its tiers: $49 a month for one user, $139 for five users, $199 for ten, and $699 for the 15-user Plus tier, plus $29 a month per extra user. Housecall Pro lands in a similar range. ServiceTitan is opaque, $250 to $500 per tech. Most three-truck shops we have looked at end up between $200 and $1,500 a month on this one line.

QuickBooks Online is another $30 to $200 a month, depending on which tier Intuit has invented this quarter. Then a separate phone routing tool. A separate review-collection tool. A separate scheduling layer if the field service software does not do it the way the owner wants. A separate estimating tool. Sometimes a separate CRM. The full bill, end to end, lands between $400 and $2,000 a month for a shop grossing somewhere between $50K and $300K a month.

The bill is not actually the worst part. The worst part is that most owners use 20 to 30 percent of any one of those tools. The other 70 percent was built for a shop three times their size.

Why the tools fit wrong

The large platforms were built for shops with twenty to two hundred technicians. The marketing assumes that audience. So does the pricing. The training videos assume a dispatcher whose entire job is dispatch.

A three-truck shop has no dedicated dispatcher. The owner dispatches. Sometimes the owner’s spouse dispatches. Sometimes the senior tech dispatches from the cab on the way to the first job. There is no separate sales department, ops department, accounting department, or marketing department. One person is wearing six hats and the software needs to know that.

ServiceTitan has a sales module and an ops module and a CRM module and a marketing module. Each one assumes a person sitting in a chair somewhere whose whole job is that module. In a four-truck shop, half of those modules sit empty. The owner pays for them anyway.

This is the mismatch. The software is not bad. It was built for a different shape of business. It happens to be the only shape the demos showed.

The honest small-business reviews on G2 and Software Advice keep saying the same thing in different words. October 2025 on Software Advice: the platform “would probably work great for a very large company” but “does not work well for a company that wants to keep things simple with fast one-time projects.” June 2025 on G2: “The product is complicated, which means you need help regularly, but their product support is TERRIBLE.” Those are people who already paid.

What changed in 2026

A coding agent now runs on your laptop, writes code while you describe what you want in plain English, and is honest enough about what it did that a person who has never written a line of code can verify the work after the fact. The model underneath has been moving for two years. The thing that changed this year is that the agent got good enough that a shop owner can sit down on a Saturday morning, describe one piece of their workflow, and have a working version of that piece on a real screen by Sunday afternoon.

Just the one piece of the workflow that the platform you already pay for makes harder than it needs to be. A coding agent will not replace ServiceTitan and QuickBooks and the rest of the stack in a weekend. Trying that is how shops give up in week three.

For most small contractors, the one piece is one of three things. It is the job board the dispatcher (the owner) actually looks at five times a day, cut down to the four columns that matter. It is the quote-to-invoice flow that ought to turn a Sunday-evening walkthrough into a sent invoice by Tuesday morning, but takes nine clicks because each step lives in a different module. Or it is the customer text that should go out the day before, the day of, and the day after a job, but instead gets typed three times by hand.

Each of those is a five-weekend project for an owner with average tech literacy, a standing brief, and a willingness to actually press the buttons on a coding agent. The agent itself is $20 a month for Claude or Cursor. The database underneath is zero to $25 a month. Hosting the result is zero to $5 a month.

The math, end to end, shifts from $400 to $2,000 a month of bloated platform plus workarounds, to $50 to $100 a month of focused tool plus the platform you keep for the parts it does well.

Three shops, three weekends

The descriptions below are composites. The shapes are real, drawn from contractors we have watched sit down with Claude Code on a Saturday morning at the workshop. The dollar figures come from their actual bills.

The first is a one-truck HVAC operator who dispatches from his phone. He pays Housecall Pro $79 a month. The dispatch part of Housecall Pro is fine, but the app shows him every customer in the database every time he opens it, when what he actually wants is the three customers booked today and the seven he should be calling about maintenance. He spent two weekends with the agent and built a Saturday-morning page that shows exactly that, plus a one-button “send the reminder text” action. Housecall Pro stays in place because the QuickBooks sync works. The new page sits on his phone’s home screen. He opens it instead of opening Housecall Pro most mornings.

The second is a three-electrician shop with the owner’s wife running the office. They pay ServiceTitan $720 a month (a contracted rate from two years ago) and QuickBooks Online $180 a month. The pain is the quote. Every quote requires the office to open ServiceTitan, build the quote in their editor, save it, export to PDF, attach the PDF to an email, send the email from a separate Gmail account. If the customer says yes, the office manually marks the quote accepted, manually converts it to a job, manually creates the invoice, manually marks the invoice sent, manually marks it paid when QuickBooks confirms the deposit. Roughly nine clicks per quote, twenty quotes a week. The office built a one-page quote-to-invoice flow with Claude Code in three weekends. A customer accepts a quote in their email. The job appears on the schedule. The deposit invoice fires automatically. ServiceTitan still sees the data through its API. The office sees their own page.

The third is a kitchen-remodel firm with one project manager, two carpenters, and an owner who is on every job site by noon. They pay Buildertrend ($459 a month last time we asked) plus QuickBooks. The pain is the Friday client update. Every week the PM writes ten or twelve clients a “here is what happened this week” email, by hand. Half the photos come from his phone, half from the carpenters’ phones, half from his own walkthrough that morning. He built a small uploader where the carpenters text photos with a project tag, and a Friday-morning page that drafts each client email from the photos and a thirty-second voice note. The email still gets read and sent by him because he wants the final read. Drafting time dropped from forty minutes a week to seven. Buildertrend stays for scheduling and budgeting. The Friday update lives in his own tool now.

None of the three replaced the existing platform. All three built the part the existing platform got wrong.

What the work actually looks like

The shape is the same every time. Weekend one is the standing brief and a working first screen with hardcoded data. Weekend two is real data wired up to either an Airtable or a Supabase table. Weekend three is the second user (the office person, the spouse, the senior tech) being able to use it too, if it needs to be shared. Weekend four is polish, plus the verification ladder passing clean on both the workflow you built and one unrelated workflow you did not touch. Weekend five is putting it on a phone home screen, a tablet on the truck, or wherever the actual work happens.

The single most useful move at hour one is the one the scoping page covers in detail: name one user, one workflow, one screen. For a contractor, that is one person on your team, one moment in their day, one page they open to handle that moment. If you can write the three lines on the back of an estimate form in under two minutes, the scope is workable. If it takes ten minutes, cut.

The objections that come up every Saturday

“I don’t have time.” The most honest objection. The contractor is right that they do not have a free weekend open for the next six weeks. The realistic version is one Saturday morning and one Sunday evening for five weeks, not whole weekends. Work compounds because the brief written in week one gets reused in every later prompt. The first weekend is the hardest. After that the agent already knows the business.

“What if it breaks on a job.” Do not put it on the truck until it has worked in the office for a week. The new tool runs in parallel to the existing platform for two to four weeks. ServiceTitan or Jobber or Buildertrend stays paid for. The new tool is used only for the workflow you wanted to fix. If the new tool breaks, the existing platform is still there. The new tool only takes over a workflow when the owner actually trusts it.

“I don’t know how to code.” Neither did the three contractors above. The agent writes the code. The owner describes the business in plain English, reads the agent’s plain-English summary of what it changed, and clicks around their own product to confirm it actually does what was claimed. Product judgment, not programming.

“I tried no-code and it didn’t work.” No-code platforms (Bubble, Glide, Softr) are a different category. They make you assemble the workflow out of pre-built pieces, which means the ceiling is the menu of pieces the platform happens to ship. A coding agent is not no-code. It writes actual software, which means there is no menu. The trade-off is that the agent’s output has to be verified (the verification ladder essay covers how), but the ceiling is the actual ceiling of software.

“What if I want to sell my company later.” A custom tool built with an agent does not hurt the sale. The buyer either keeps using it, or exports the data and migrates to whatever platform they prefer. The data is the shop’s. The code is the shop’s. Neither lives behind anyone else’s API rate limits.

“What if the agent company goes away tomorrow.” The code on your laptop does not go away. It runs without the agent. The agent is for building and changing it. Once the tool is running, it is just software.

“I tried ServiceTitan, it was supposed to fix this, it didn’t.” This one usually comes from someone six months into a contract. The advice is the same as for any other platform: keep paying for what it actually does well, build the part it does badly. Cancelling a platform mid-contract is its own fight with its own paperwork. Most shops that try this start with the build and stop paying for the platform later, once the replacement workflow is genuinely better.

The truck comparison

A new one-ton service truck with the build-out for a small HVAC or electrical shop runs $65,000 to $95,000 in 2026. Most contractors finance it. The monthly note is somewhere between $1,100 and $1,600.

The software stack we were talking about earlier, at the four-truck-shop scale, runs $1,500 to $4,000 a month. The high end of the stack costs more than the truck. The truck does a job every day. Most of the stack does maybe a third of what it bills for.

That is the comparison the contractors who have already built their own tools ran in their heads. They are not building because they love technology. They are building because the math finally allows them to.

Where to start

If a specific workflow inside your business is the one that wakes you up at night, that workflow is your first project. Not the rebuild of the whole platform. Not a clone of Jobber. The one workflow.

The scoping reference is the page on how to cut that workflow down. The CLAUDE.md essay is the page on the brief the agent needs in front of it. The verification ladder essay is the page on what to do after the agent says it is done. Those three pages are the whole system.

The Saturday workshop is where contractors bring the workflow they would build first and walk out with a brief and a first screen by lunch. There is no slide deck. There is a room with twelve seats, a coding agent on a screen, and a working first version of the thing you came in with.

The contractors who are quietly doing this are not posting about it. They are looking at a $1,500 software bill, then at a $25-a-month bill for the slice they built themselves, and getting on with the day.