Custom AI Software for Trades: When It's Worth It, and What It Costs

"We need custom software" is almost never the actual finding. It is the conclusion someone reaches after a specific frustration: the same job details typed into three systems, a report that takes a morning, a platform charging for eleven seats when four people use it. The frustration is real. The conclusion usually skips two cheaper answers.

So before the quotes, work out which layer your problem lives on.

01Layer 1: what the platforms actually cost

Read from each vendor's own pricing page on 5 September 2026. Note that per-user pricing is where the real money is, so the plan price alone will mislead you.

Jobber and Housecall Pro figures are first-party, read from their pricing pages. Jobber also publishes a middle rate for a one-year commitment and charges $29 per extra user; its Plus plan runs $499 monthly or $399 annual. ServiceTitan publishes no pricing at all, so the range shown is third-party reporting and should be treated as an indication, not a quote. Verify before budgeting.

Two things worth pulling out. First, extra users are the real cost driver: Housecall Pro's Essentials plan is $149 a year-round for five people, then $100 a month for each one after. A ten-person crew on that plan costs more in seats than in subscription. Second, ServiceTitan not publishing a price is itself information. It is sold through a demo and a proposal, which is fine, but it means you cannot compare it without entering a sales process.

02Layer 2: the glue, and why it is usually the answer

Here is the pattern behind most "we need custom software" conversations. A job comes in by phone. Someone writes it on a pad. Someone types it into the scheduling platform. Someone re-types the same details into the accounting system. Someone copies the totals into a spreadsheet on Friday.

That is not a missing feature. It is four systems that do not talk, and the fix is plumbing rather than a new product. A general automation platform costs somewhere between $12 and $30 a month, and wiring your specific systems together is a small scoped job rather than a build.

This layer gets skipped constantly, for a reason worth naming: nobody sells it hard. There is no salesperson whose commission depends on you spending $400 instead of $20,000. But the expensive problem in most trades businesses is not an absent capability, it is the same information being entered three times by three people, and that is exactly what this layer removes.

03What custom actually costs, and what drives the number

Published 2026 guides put one-off custom builds at roughly $5,000 to $50,000, driven by how many processes and integrations are in scope. Our own builds start at $1,000, which buys one well defined workflow rather than a system, and we publish the rest of our pricing here.

The comparison that matters is not build price against build price. It is build price against the subscription you would otherwise keep paying. Jobber Grow and Housecall Pro Essentials both land at $149 a month on annual billing, which is $1,788 a year. So a $9,000 custom build is roughly five years of that platform. Custom does not just have to work. It has to beat five years of a product that thousands of other companies are jointly paying to improve.

What actually moves a custom quote, in rough order:

• Number of systems it has to talk to. Each integration is its own small project, and the ones with no modern interface cost multiples of the ones with a clean API.
• Whether it touches money or compliance. Payments, tax and regulated records carry edge cases and audit requirements that ordinary features do not.
• How settled the process is. Automating a workflow you are still redesigning means paying to build it twice.
• Who runs it afterwards. A build handed over runs cheaper than one somebody hosts and monitors for you, and it also fails more often, which is the trade.

04The four tests for when custom is genuinely right

If none of these is true for you, spend the money on layer 1 or layer 2 instead.

1. Your advantage lives in a process no platform models. If how you quote, schedule or inspect is genuinely different from your competitors and that difference is why customers choose you, forcing it into a generic platform sands off the thing you are selling.
2. You are paying per seat for software most of your team barely touches. Ten technicians who need to see today's jobs and upload three photos do not need ten full licences. At $75 to $100 per extra user per month, a narrow app for that one task can pay for itself inside a year.
3. A manual step costs more per year than the build costs once. Put a real number on it: hours a week, times the loaded hourly cost, times fifty. If that annual figure exceeds the quote, the case makes itself. If it does not, no amount of enthusiasm will fix the arithmetic.
4. Two systems must talk and no integration exists. You have tried layer 2, and the connector genuinely is not there. This is the most common legitimate reason, and it is usually a small build rather than a large one.

05The five ways custom builds fail

Almost never because the code was bad. Every one of these is a decision made before anyone writes a line.

• The process was still moving. It got automated mid-redesign, so it was wrong on delivery. Stabilise the process on paper first, and run it manually for a month.
• Nobody inside owned it. Software changes how people work. Without one person in the business who wants it to succeed, it gets quietly worked around and the old spreadsheet reappears.
• It was built around one person. Modelled on how your best estimator works, then that person leaves and the software encodes habits nobody else has.
• Nobody budgeted for maintenance. Dependencies age, integrations change, certificates expire. A build with no maintenance line decays into something everyone routes around by year two.
• It was scoped as a replacement rather than an addition. Rebuilding scheduling, invoicing and dispatch from scratch means competing with products that have thousands of customers funding them. Custom wins at the edges, not the middle.

06What to settle in writing before commissioning

1. Who owns the code? In the contract, not in conversation.
2. Whose accounts hold the hosting and the API keys? If they are in the developer's name, leaving means rebuilding rather than transferring.
3. What does it cost in month thirteen? Hosting, monitoring, provider usage and maintenance, itemised.
4. What happens when it breaks on a Monday morning? Who is called, how fast, and is that contractual.
5. Can another developer pick it up? Ask what documentation ships with it. "It is all in the code" is a no.
6. What is explicitly out of scope? A quote that only lists inclusions is not a quote yet.

07Where we fit, and where we do not

We build layer 3, and some layer 2. Builds start at $1,000 for a single well defined workflow, managed operations are $750 a month if you want us running it, and the audit that scopes it is free. Full pricing is in the agency comparison.

We are the wrong call if you do not yet have a field service platform, because a custom build is not a substitute for one and we would be selling you a worse version of Jobber. We are also the wrong call if you have not written down how many times a job's details get retyped, because that list frequently turns a $9,000 conversation into a $400 one. And if your procurement needs a completed third-party security certification today, we do not have one.

The audit is free specifically so that answering "you do not need us" costs you nothing.

08Common questions

---

Sources: Jobber and Housecall Pro pricing pages, read 5 September 2026, first-party. ServiceTitan publishes no pricing, so figures attributed to it are third-party reporting and are labelled as such. Custom build ranges come from published 2026 industry pricing guides, not from our own book of business. Prices change without notice, so verify before you budget. Related: Best AI Automation Agencies (2026), Best AI Website Builders (2026) and What an AI Receptionist Does When It Doesn't Know the Answer.