Engineering
Civil site plan sets against agency design standards and submittal checklists: stormwater, erosion control, grading, utilities, encroachments, and whether the sheets agree with each other.
Review Assistants are AI-assisted QC for plan sets, plats, submittals and closeout packages, built on your standards and the agency’s own checklists. They read every sheet, check the set against itself and against the rules it will be judged by, and hand your reviewer a written comment list.
Your staff runs them. Your engineer or surveyor signs off on them.
We run the review and send back a written comment list you can hand straight to your designer. No setup and no commitment. It is also the quickest way to see what the assistants find on one of your own projects.
$30 covers a set up to 6 sheets. Every sheet after that is $5.
Agency requirements are checked only for the agencies already in our library. For any other agency, your set gets the General Plan Review: everything that is wrong no matter who reviews it.
We set the Review Assistants up inside your firm’s own Claude account, so your staff can run a review on any set, any day, without sending it anywhere.
$9,500 installed.
Travel is quoted with the install. Need more than three new agencies? Ask.
Some things are wrong everywhere. A misspelled word. A detail callout that points at nothing. A pipe slope that disagrees between the profile and the table. A plant count that does not match the schedule. The General Plan Review checks civil plans, plats and landscape plans for all of it, without an agency checklist, so it works for any firm in any state from the first day.
Spelling, notes left over from another project, notes that contradict each other, abbreviations nobody defined.
Legends, leaders, callouts and sheet references that resolve, scales that match, matchlines that agree.
Plan against profile against table against detail, for everything the set contains.
Slopes, inverts, stations, areas, counts and closure, recomputed instead of trusted.
Add your firm’s standards and it checks those too. Add an agency and the assistants below take over, each one tuned to your firm.
A project from the first submittal to the day the bond is released.
Civil site plan sets against agency design standards and submittal checklists: stormwater, erosion control, grading, utilities, encroachments, and whether the sheets agree with each other.
Plats and surveys against state standards of practice, the recording jurisdiction’s plat checklist, and ALTA/NSPS requirements: certificates, notes, references, closure and title block.
Planting, tree protection and irrigation plans: buffer yards, parking lot landscaping, tree preservation, plant schedules that match the plan. Your local landscape ordinance is loaded as part of the install.
Whether the set can be built from as drawn: missing dimensions and elevations, unlabeled slopes, callouts with no detail behind them, and plan against profile against schedule.
Shop drawings, precast structure cut sheets and product submittals checked against the plans before anyone approves or orders: sizes, inverts, openings, materials and counts.
As-builts against the approved plans and against each agency’s closeout checklist: certifications, stormwater as-built data, utility record requirements, and what is still missing from the package.
Every install includes these at no extra charge. Send us a set from one of these jurisdictions and it is checked against that agency’s own requirements.
Plus the standards that travel: ALTA/NSPS land title survey requirements, the South Carolina standards of practice for surveying, and North Carolina plat requirements. Coverage varies by discipline, so ask whether your agency is in the library for the review you need, or request it. An agency that is not there yet is built from its published ordinance, design manual and checklists.
Tell us which county, city, utility or state agency you submit to. Requests go to the front of the queue, and we email you when it is in the library.
Your agency request is logged and a person reads it. We will email you when it is in the library.
Yes. Review Assistants run on Claude, the AI model from Anthropic. What New Roots builds and maintains is everything that makes it a reviewer for your work instead of a general chatbot: the review method, the agency checklists and ordinances, your firm’s standards, and the report format.
With an in-house install, nowhere new. Your staff run reviews inside your firm’s own Claude account, and New Roots never receives or sees the sets. The account is yours, on a Claude Team or Enterprise plan billed by Anthropic. Anthropic states that by default it does not use inputs or outputs from its commercial products to train its models; read their policy at the source.
When you send us a set to review, we receive that set and handle it under our privacy policy.
No. It is a QC aid. It will miss things, and it will flag things that do not apply. What it does well is the tedious part: reading every sheet, every time, against a long checklist, without getting tired on sheet forty. That leaves your reviewer’s time for judgment.
The list above is what is in the library today: coastal South Carolina and southeastern North Carolina, county, municipal, utility and state. An install includes all of them and up to three new agencies built from their published ordinances, design manuals and checklists. No agency on the list? The General Plan Review needs none.
A Claude plan for the staff who will run reviews, your CAD standards and templates, a few recent sets with the agency comments they received, and the list of agencies you submit to.
The assistants are built from your own ordinance and checklist. Every submittal gets the same complete first pass, so staff time goes to the judgment calls instead of the missing north arrow. It does not approve or deny anything; your reviewers do.
Run the constructability and submittal reviews on the set you are bidding or building: the missing invert, the detail nobody drew, the schedule that disagrees with the profile. Your questions go to the engineer as one list instead of one RFI at a time.