Domestic water analysis is demand math you’ve done a hundred times, a model you build the same way every time, and a pressure check with a floor and a ceiling. None of it is hard. The grind is the assembly and the writeup, not the judgment, which is exactly the kind of work I didn’t want the professional engineer in charge spending a day on.
So I handed the production part to an agent, the same way I did with fire flow. The two workflows are closer than they look, and that turned out to matter.
What a domestic water analysis actually involves
Start with demand. Count the fixtures or the units, run them through Hunter’s curve (UPC Table 610.3, or IPC Appendix E depending on the adopted code), and get a peak GPM. Then distribute that demand across the service nodes in proportion to what each one actually serves. Get the demand basis wrong and every pressure number downstream is wrong with it.
Then pull the flow test from the water purveyor and read the static, residual, and pitot. Here’s the first place domestic differs from fire: you use the raw numbers. Some fire districts apply a ten percent reduction to a flow test for sprinkler calcs. A domestic model is not a sprinkler calc, so that reduction does not belong in it, and reaching for the reduced number is a quiet way to model a system that’s weaker than the one you’ll actually build.
Then you build the model. In EPANET that’s the reservoir or the pump curve off the flow test, the public main, the on-site loop, the services, the meters, the backflow assemblies. You set the C-factor by material and age, not by habit. New C900 PVC runs at 150. A fifty-year-old asbestos-cement main is closer to 110, and modeling it at 150 hands you pressure you don’t have. The meter and assembly losses come from the serving utility’s approved detail and its certified test data, not a generic default and never a K-value carried over from a different district’s project.
Then you run the scenarios. Peak hour with everything drawing at once. Peak day sustained. Peak hour plus fire flow at the most remote hydrant. And dead-end flushing, to confirm you can actually move water fast enough at the ends of the system to keep it from going stale.
Then you check the window, and this is the real difference from fire flow. Domestic is two-sided. Every service has to hold its minimum at peak (15 psi under the plumbing code), and nothing is allowed to sit above 80 psi without a pressure-reducing valve. Fire flow only cares about the floor. Domestic cares about the floor and the ceiling, which means a flat site with strong supply can fail the analysis just as fast as a starved one on a hill.
If a node fails, you size up. Trace the critical path to the failing node, rank the pipes by how much head they’re each burning, upsize the worst offender to the next standard diameter, and re-run. Repeat until it holds.
Then you write it up. Demand tables, scenario results, pressure maps, the PRV recommendations, the optimization log, the code references, the appendices. Format to firm standards. QC it. And then the part that’s actually the job: the professional engineer in charge reviews the analysis, owns the judgment calls, and stamps it.
If you’ve sized a domestic system, none of that is news. If you manage the people who do them, you already know it’s the better part of a day, every time.
What it used to take us
A full domestic water analysis at Calichi (the EPANET model, the demand calc, the scenario runs, the written report with pressure maps and appendices) ran the better part of an engineer’s day, same as any full hydraulic report.
I can give you a hard, measured number on the workflow it’s built on. Fire flow, which runs on the same EPANET pipeline and the same flow-test parser this agent uses, went from 16 to 24 hours of engineer time down to about 2. That’s our own production-measured number, on our own projects. Domestic water rides that same machinery. I’m not going to hand you a separate stopwatch figure for it and pretend I measured each one the same way, because the honest way to find your firm’s number is to measure it on your real projects, not to read it in a post.
The frustrating part was never the judgment. The demand basis, the read on whether the purveyor’s main can actually serve the project, the PRV strategy, that’s real engineering and it deserves an engineer. The frustrating part was the other ninety percent: the model assembly, the scenario runs, the table-building, the formatting. Same shape every time, done by the most expensive person in the room.
How the agent does it now
The engineer emails the agent the way they’d email a junior: domestic water for this project, flow test attached, unit count or fixture schedule attached, water plan attached. From there:
- It computes demand from the fixture units or the per-unit basis (Hunter’s curve) and distributes it across the service nodes.
- It parses the flow test and reads out the static, residual, and pitot, using the raw values.
- It builds the EPANET model from the project files: the public main, the on-site loop, services, meters, and assemblies, with C-factors set by the as-built material and age, and pipe geometry pulled from the CAD water layer so the model looks like a plan view of the site instead of a random graph.
- It runs the scenarios (peak hour, peak day, peak-hour-plus-fire, dead-end flushing) and checks every node against both the minimum and the maximum pressure.
- Where a node fails, it traces the critical path, upsizes the governing pipe to the next standard size, and re-runs until the system holds.
- It drafts the report in firm format: demand summary, scenario pressure tables and maps, PRV recommendations, the optimization log, and the appendices.
- It routes the whole package through a separate QC agent before a person sees it.
Then the finished package lands in the project folder, and the engineer picks it up the same morning instead of clearing a day for it.
The QC gate
Same architecture as everywhere else in the system, and I wrote a whole piece on why it exists, so here I’ll just say what it catches on this specific workflow. The independent QC pass flags a flow test that got used with a reduction it shouldn’t have, a C-factor that’s too generous for an old main, a node sitting above 80 psi with no PRV called out, and a dead-end that won’t hit flushing velocity. It checks the arithmetic and confirms the scenarios are complete before any of it reaches the engineer.
The point isn’t that the first agent is perfect. It isn’t. The point is that a deterministic check that runs the same way every time catches the boring mistakes, so the engineer spends attention on the ones that need judgment.
What stays human
Everything that needs the professional engineer in charge. The engineer decides the demand basis, makes the call on whether the purveyor’s supply is realistic or needs a will-serve conversation, sets the PRV strategy, weighs the unusual hydraulic condition, and applies the stamp. The agent does the production pattern-work. The engineer does the engineering. That division is the whole design, and it’s why this supplements your engineers instead of replacing them.
Where it applies
Domestic water modeling shows up on most projects that pull potable service off a public main: subdivisions, campuses, multifamily, mixed-use, commercial. It crosses jurisdictions, which is where the manual version gets slow, because every purveyor has its own approved assemblies, its own standard details, and its own flow test format. An agent that resolves the jurisdiction from the project location and applies the right standards turns the slowest part into the fastest.
We run it everywhere we practice.
What this means for your firm
I’m not going to promise you a number on this one. I’ll show you the workflow it’s built on and the number we measured there, and then the honest way to find yours is to measure it on your real projects, with your templates and your purveyors and the way your engineers work today.
That’s what the first two gates of how I deploy are for. An AI Readiness Audit to find where the highest-value automation sits across your divisions, then a Strategic AI Discovery that puts my engineers and yours on your actual workflows and produces the hours-saved picture by workflow, so you see exactly where the time goes back. Every gate is a stop or go. You only go further once you’ve watched the last step work on your own projects.
Domestic water was an early one for us, because it’s the same family of work as fire flow and it rides the same pipeline. Once one hydraulic workflow proved out, the next one was mostly a matter of pointing the same machine at a different question.
If you run a multi-office firm and your engineers are building these models by hand, the first step is the AI Readiness Audit.