When people talk about artificial intelligence (AI) in construction, they jump straight to the shiny stuff: the robots, the automation, the blueprint-reading software, and the latest pitch that makes it sound like the trades have been standing still while everyone else has discovered the future.

Anyone who has actually worked in construction knows better. The trades have always been innovative, but usually in ways that are practical, quiet and tied directly to survival. We adopt tools to help us bid better, build faster, protect our margins, reduce mistakes and make sense of a project before it turns into a jobsite problem wearing a hard hat.

This is where AI gets interesting for plumbing contractors; not as a replacement for plumbers, estimators, project managers or foremen, but as a way to capture the knowledge that already exists inside our companies before it disappears into old folders, forgotten spreadsheets, archived emails, jobsite conversations, and the memory of the one person who knew exactly why the last job made money while the next one nearly ate the company alive. 

Contractors have always had valuable data moving through their businesses, but too often it is scattered across estimates, timecards, invoices, purchase orders, change orders, payment requisitions, inspection notes, closeout reports and the lived experience of the people doing the work.

The opportunity is not to use AI because it is new. The opportunity is to use AI the same way the best contractors have always used better tools: to organize, optimize and scale.

Organize: Capture the knowledge before it disappears

When I first started working around estimating, I learned from estimators who prepared bids in completely different ways. Some leaned on unit pricing, others built detailed estimates from the ground up, and many had personal systems developed over decades of field experience, supplier relationships, old bid tabs, handwritten notes and hard lessons learned on projects that either made money or taught everyone exactly where the money went. 

What stood out early on was that estimating was never just math. It was construction knowledge, company history, labor awareness, material timing, cash flow strategy, and the ability to look at a set of drawings and understand how the job would actually be built long before anyone in the field touched a piece of pipe.

Before on-screen takeoff became common, there were digitizers, those large boards connected to estimating programs where an estimator could trace drawings and turn paper plans into measurable quantities. It was a major step forward because the computer was not doing the thinking; it was helping capture the measurement in a more consistent and usable way. 

On-screen takeoff pushed the process further. Plans became digital, counts and lengths could be stored with markups and notes, assemblies could be organized more cleanly, and estimators could create a better connection between the drawings and the final number.

This was when I started seeing where this was headed. The future was never only about counting faster. The future was about creating a more streamlined process in which plans, specifications, fixture schedules, labor categories, material lists, alternates, payment applications and actual job performance could begin to talk to each other. AI is the next step in that same evolution.

The first job for a contractor is not to build some giant AI system. The first job is to organize the existing data. That means gathering completed estimates, fixture counts, labor breakdowns, material buyouts, timecard summaries, change orders, inspection notes, closeout reports and lessons learned from finished jobs. It means taking the information that currently lives across five people, seven folders, three accounting exports and one estimator’s memory, then turning it into something the company can actually reuse.

A contractor can start simply by asking AI to clean up what is already in front of them.

Starter prompt:  

“Act as a plumbing contractor’s estimating assistant. I am going to paste notes from a completed project. Organize the information into categories for labor, material, equipment, change orders, payment requisitions, inspections, delays, lessons learned and items that should be reviewed before bidding on a similar project again. Do not invent information. If something is missing, create a list of follow-up questions.”

This type of prompt is not glamorous, but it is useful because it turns scattered job knowledge into a repeatable record. Use the same approach at the end of every project, and you slowly build a library of lessons learned that can be searched, compared and brought back into the next estimate before the same mistake shows up wearing a different project name.

Custom GPTs take this further. In plain contractor language, it means a plumbing company can begin building an assistant or agent around how it already estimates, tracks, communicates and learns. Not generic national averages. Not fantasy labor units from someone who never had to make payroll. A contractor’s own labor categories, fixture assemblies, preferred vendors, common exclusions, markup structure, alternates, payment requisition categories, closeout lessons and scars.

Custom GPT prompt: 

“Help me create instructions for a custom GPT that acts as my company’s internal plumbing estimating assistant. It should review project notes, estimates, labor summaries, fixture counts and closeout lessons. It should organize information using our company categories, flag missing scope, suggest add/deduct alternates, identify potential labor risks and produce a clean bid review checklist. It should never provide a final price unless I enter the labor rates, material costs, overhead, profit and markup assumptions.”

The point is not to let the machine own the estimate. The point is to teach the machine how the company organizes the thinking behind the estimate.

Optimize: Make the work sharper

When I was a chief estimator, estimating multimillion-dollar developments from plumbing, mechanical and civil drawings and related specifications, the drawings were sometimes detailed and well-coordinated and sometimes conceptual enough to make an estimator raise an eyebrow and start asking better questions (RFIs). 

The first responsibility was always to bid the project as drawn and specified because the base bid matters and the documents still define the scope. The competitive edge often came from what happened after that base number was prepared.

Once the project was priced by plan and specification, I would look for add alternates, deduct alternates and other options that could make the project better, more competitive, more buildable or less risky. 

Sometimes that meant helping the owner save money without undermining the design by offering a menu of value-engineering options. Sometimes it meant identifying a better fixture package, a simpler routing option or a constructability issue that would become much more expensive if nobody caught it before the job started. 

This kind of thinking does not always show up neatly in a spreadsheet because it comes from experience, and experience is data. This is where AI becomes useful, but only if contractors teach it how the work actually happens. 

Take one plumbing fixture. On paper it may look simple: one fixture, one count, one line item. In the field, that fixture may require multiple visits, labor phases, inspections and opportunities for something to go wrong. A proper estimate needs to account for rough-in, pipe, fittings, valves, supports, carriers, hangers, testing, rough inspection, fixture delivery, staging, setting, trim, final inspection, punch list, cleanup, supervision, overhead, profit and whatever markup structure the company uses.

A good estimator is not only counting beans; he is building the job in his head before the field ever builds it with their hands.

Fixture assembly prompt: 

“Act as a plumbing estimator. Help me build a fixture assembly checklist for a commercial water closet. Break the scope into rough-in labor, carrier installation, water piping, waste and vent piping, valves, supports, testing, rough inspection, fixture setting, trim, final inspection, punch list, cleanup and supervision. Do not assign labor hours yet. First create the checklist and identify what information I need before pricing it.”

Once that checklist exists, the contractor adds the company’s actual labor units and material assumptions. AI helps structure the assembly, but the contractor still owns the labor, the markup, the risk and the number.

The same logic applies to plan review. A plumbing takeoff is not only counting symbols and determining scale for linear pipe. It needs to compare what is drawn against what is scheduled, what is specified against what is submitted, and what is being carried in the estimate against what the company will actually be expected to furnish and install.

Plan review prompt: 

“Act as a plumbing estimating assistant reviewing project documents. I will upload or paste information from plumbing plans, specifications, fixture schedules and addenda. Create a conflict log that compares the drawings with the fixture schedule and specifications. Organize the results into missing information, conflicts, possible alternates, scope clarifications, exclusions and questions for the general contractor or engineer of record.”

AI can help read plans, compare sheets, summarize specifications, flag conflicts and organize alternates, but the contractor still has to know whether the quantities make sense, whether the labor is realistic, and whether the number being carried is one the company can stand behind when the job gets real. AI does not know your crew, your foreman, your local inspector or whether a particular job has all the early warning signs of becoming a coordination war. That information still belongs to the contractor.

Scale: Build tools that make you different

The piece many contractors are still missing is that AI is not limited to answering questions inside a chat window. Once a contractor understands the workflow he wants, he can begin using AI to help build small internal applications, calculators, checklists and private website tools that turn company knowledge into repeatable systems.

This is something I have already been doing. One result is Smooth Estimator, a free, mobile-friendly plumbing takeoff tool that turns field notes or voice input into a clean materials-and-labor table, applies your markup and generates a customer-ready proposal you can edit live and save as a PDF. 

I described the workflow the same way any estimator would describe a fixture assembly, and AI wrote the first version. I tested it, pointed out what broke in the field, refined the catalog and pricing logic and kept iterating until it felt like a tool a plumber would actually open on a phone in a truck.

Progress should not be held up by gatekeepers. If a contractor can describe the problem clearly, AI can help turn that description into working code.

How to actually use these prompts

Open your favorite large language model: ChatGPT, Grok, Claude or Gemini. Copy the entire prompt and paste it into the chat. Most of the time, ChatGPT will hand you a downloadable HTML file plus a temporary hosted link so you can click and try the tool live right away. Claude often shows an interactive preview in the same chat. 

If you want full control, as I did with Smooth Estimator, ask the AI for the complete single-file HTML, CSS and JavaScript version. Embed it on your own site so the tool lives permanently on your domain and works both online and offline.

Test it like a real job. Enter a couple of fixtures, change the labor hours, generate the proposal and see what breaks. Then stay in the same chat and tell it exactly what to fix in plain English. Every time you describe a change, the AI rewrites the code or updates the live preview. That back-and-forth is how Smooth Estimator went from a rough idea to a working field tool. The same loop lets you keep adding features until it matches the way your company actually works.

Simple internal tool prompt: 

“Act as a software developer and plumbing estimating assistant. Help me build a private internal web tool for my company. The tool should let our estimator enter fixture type, quantity, material cost, rough-in labor hours, trim-out labor hours, inspection time, overhead percentage, profit percentage and notes. 

“It should calculate total labor hours, total material cost, markup and final price, and export a clean summary that can be copied into an estimate or payment requisition backup. 

“Write the code in HTML, CSS and JavaScript so I can embed it on a private company webpage. After writing the first version, explain how I can edit the fixture types, labor categories and markup percentages.”

The first version does not need to run the company. It needs to solve one problem. Then you test it, improve it, brand it, add features and slowly build a library of internal tools that reflect how the company actually works.

Every completed job creates valuable information: estimated labor versus actual labor, material buyout versus estimated material, change orders that should have been caught earlier, fixtures that took longer than expected, inspections that delayed progress and mistakes that should never happen twice. This information is gold, but too often it disappears. AI gives contractors a chance to stop losing that information and start feeding it back into the next estimate.

The future of AI in plumbing estimating is not asking a computer to magically know the work. It teaches the computer how a real contractor sees the work, then uses that intelligence to organize, optimize and scale. 

The contractor still determines the labor and markup, still understands the risk and still owns the number. With the right systems in place, AI can capture the thinking behind that number, compare it to past performance and make the next bid cleaner, faster and smarter.

That is the edge: not AI replacing the trades, but trade intelligence finally getting captured and shared. All while the contractors who figure it out first are going to be the most dangerous, in the best possible way.