Questions, answered.

Growing commercial general contractors — an office running several jobs at once, with project managers, estimators, and someone who closes the month, plus crews and subcontractors in the field. The work it fits is tenant improvements, retail, restaurants, light-industrial, ground-up, and mixed portfolios. It assumes AIA G702/G703 billing, a schedule of values, retainage, lien waivers, and WIP are part of your month, and it gives every seat in the office its own project-level role with a money firewall behind it. Built for the office and the field together, not just one or the other.

Because AI is wrong sometimes, and contractors get sued. Core workflows have a manual path, AI output is advisory, and outward financial or contractual content stays in draft until a human approves it. Each AI call shows its quota cost up front, and the result is saved so you can review it later. The AI is leverage, not a black box you have to trust blindly.

It ingests the artifacts on a job — drawings, daily logs, photos, RFIs, change orders, schedules, bills — into a per-project memory. Agents query it to answer questions, draft RFIs and change orders, summarize the week, assemble pay apps, and flag risks. Answers cite the source artifact so you can verify them.

Procore is enterprise-priced and complex; Buildertrend leans residential. neuroBLDR is built for the growing commercial GC in between — AI-native from day one, shipping AIA-format G702/G703 pay-app billing and conditional/unconditional lien waivers in the core product rather than reserving them for top tiers, and priced on the office side, so a full field crew on free field seats costs nothing.

You control it. Money visibility is role-based per project: owners and admins see everything; project managers and estimators see contract values and margin; superintendents and foremen see no dollar figures unless an owner grants them; free field-crew seats and viewers see none at all. Owners can tune the grid per role, and can put a field seat on the invoice at the office-seat rate — a paid field seat reads the change-order, purchase-order and estimate registers' quantities by default, dollars only if you grant that class dollars, and nothing at all if you switch that class off or the person is a viewer on the project.

We schedule setup with you and run it with you. Your cost codes, subs and drawing set come in through a guided import. Once your drawings are in, takeoff, estimating and AIA billing are ready to use. There's a guided walkthrough in-app.

Yes. The field surface is a mobile-first web app you install from the browser (no app store) — daily logs with voice, photos, punch, and crew time, all role-aware so free field seats never see dollar figures. Subs get their own magic-link portal to return submittals, sign waivers, and log time without an account.

Both are in here. WIP is a cross-project schedule — revised contract, cost to date, percent complete, revenue earned, and the over- and under-billing split — as a report and as a CSV, which is the format a CPA or a surety asks for rather than a dashboard screenshot. Close a month and that schedule is frozen: it's written once and read back from that record, so a cost backdated into a closed month cannot change what that month reports, and a month that was never frozen shows nothing instead of a number recomputed from today. At the other end, the closeout binder is assembled from the record you already kept — current drawings at their latest revision, warranty documents and O&M manuals, the final signed lien waivers, the final punch report, and the last pay application, collected into a cover index with the documents behind it.

It refuses, and says why. There is no default scale anywhere: measure tools stay disabled until you calibrate a page, and a markup pasted onto an uncalibrated sheet keeps its geometry and loses its numbers. Compare two drawing revisions that don't line up and you get a refusal rather than an invented quantity delta. A file our scanner can't read is reported as unscanned, never as clean. And every AI output is a draft a person approves. We would rather hand you a blank than a number you'd put in a bid without knowing where it came from.

You can prove exactly what the record supports, which is the honest half. Issuing a transmittal freezes the list of sheets and the revision each one was at, and creates an acknowledgment slot for every sub you sent it to — acknowledged in their own portal, under the name of the person who did it, with the time. Nobody gets marked as having received anything: a sub who hasn't acknowledged stays on the list as a sub to chase, which is the whole point, because the argument you are trying not to have starts with a transmittal nobody can find. We won't assert a delivery we have no record of.

Both are records here, not attachments. For the meeting, the agenda is assembled from your own live registers — the open RFIs with how long they've been open and who's holding them, the submittals still in review, the change orders pending, the open delay events — and frozen onto that meeting, so the numbers in the room are the numbers in the system. You type the minutes; distributing them is its own timestamp, because that is when they start standing as the record. Items nobody closed carry forward numbered with the meeting that raised them. For submittals, the architect's four stamps stay four — approved, approved as noted, revise and resubmit, rejected — a resubmittal starts its own round pointing at the one it answers so the returned copy survives, and when a sheet gets a new revision, any approval that was reviewed against the superseded sheet is stamped stale with the reason. Clearing that is a person re-reviewing it. Nothing resubmits itself.

Not a self-serve one. We set pilots up with you. An invited trial is a 30-day pilot with drawings, manual takeoff, RFIs, submittals, change orders, pay apps, daily logs, the AI drafting agents and the Blueprint PDF engine, and it needs no credit card. When the window ends, editing pauses until billing is set up: everything stays readable and nothing is deleted. Free field-crew seats and client and sub portals are unlimited.

Pricing is quoted to your office after you apply, rather than published. Apply at neurobldr.com/enterprise — one work email, then six questions. Crews on free field seats, clients, and subs cost nothing. You pay for the people who run the office side: project managers, estimators, accounting, and any superintendent or foreman you give office tools. If your workspace already runs, its plan and price are in Settings → Billing.

Cancel anytime from Settings → Billing through the Stripe billing portal — auto-renew turns off and your subscription stays active through the end of the period you've paid for. We don't offer refunds on subscriptions. Export your data before or after (see the data question below).

Judge it on your own numbers. Crews on free field seats, clients, and subs cost nothing, so what you pay for is the people who run the money and the paperwork, and the return shows up in the takeoff, change-order drafting, pay-app assembly and signature chasing that stop eating their week, plus the margin you protect by catching slips early. We don't publish a one-size ROI figure because it depends on your volume — run it on a live job during your pilot and look at it there.

Not on monthly. Office seats billed monthly are month-to-month — cancel any time from the billing portal, no term contract, and auto-renew simply stops at the end of the period you have already paid for. An annual prepay differs in one way only: you have bought a 12-month service period up front, so turning off renewal ends it at the end of that period rather than refunding the unused part (we don't refund subscriptions either way). Whichever you pick, your data exports out — per-project document ZIPs, CSV registers for RFIs/COs/submittals/pay-apps/WIP, and a full account export. We don't hold your project records hostage to keep you subscribed.

Office seats, plus any field seat you tick as paid. An office seat is anyone who authors the work — project managers, estimators, accounting, the owner — and the number we send Stripe is your office-seat count (office memberships plus the owner) plus any paid field seats. Free field-crew seats are unlimited: they get daily logs, photos, punch, crew time, and view-only drawings, they never see a dollar figure, and they are never counted. A field seat is only on the invoice if you tick it as paid, person by person, at the office-seat rate — it stays a field seat and opens the change-order, purchase-order and estimate registers, reading their quantities while you keep quantities on for paid field seats, dollars only if you grant that class dollars, and neither if you switch the class to nothing or the person is a viewer on that project. Client, sub, and architect portals are magic-link views scoped to one project, not accounts, so they cannot reach the invoice either. Move somebody between field and office whenever you like — the seat count follows the roster.

The product is live and supported, and we are finalizing it around our first customers. Expect changes to land often, and expect us to answer when you tell us something is wrong. Concretely: features get added and changed often, screens move, and we don't commit to an uptime service level — that's stated in the Terms rather than buried. What it does not change: it's the same production database and tenant isolation every workspace gets, your data stays exportable at any time, and nothing AI-drafted leaves your workspace without a person pressing send.

No. The agents are part of the seat — there is no separate AI licence to buy and no per-draft charge. Each Pro seat you are billed for — an office seat, or a field seat you ticked as paid — brings 1,000 AI actions per seat each month, refillable, pooled across the workspace so a heavy estimator doesn't starve a light project manager, and Ultra seats bring 4,000 AI actions per seat each month, refillable. Every call shows what it will cost before it runs, and the result is saved so you can look at it again without spending twice. If an office does run out mid-month, your owner or an admin can add a refill from Settings → Billing — one seat's monthly allowance into the shared pool, priced there before anything is bought — and nobody has to predict their usage in advance to buy one. Free field-crew seats don't include AI and aren't billed for it, unless an admin unlocks AI on that seat; a paid field seat brings its own allowance into the same pool.

You can connect QuickBooks Online to push approved vendor bills and pay-app invoices across, and to pull payment status back into neuroBLDR. Map your cost codes to GL accounts, your subs to QuickBooks vendors, and your projects to customers on the settings screen — that mapping is what the push uses, and a record without one is skipped and reported rather than guessed at. When the two sides disagree on an amount we surface the discrepancy instead of auto-resolving it. Xero and Sage 100/300 Construction are on the roadmap.

Magic-link views that let clients review and approve pay apps and change orders, and let subs sign lien waivers and pull drawings — with no account and no seat consumed. Links are scoped to a single project and expire on a schedule, and the access tokens are hashed at rest.

Through a keyed accounting export: a file your bookkeeper can post, at the grain an accounting system keys on — one row per job, cost code, cost type, period, vendor and reference. Amounts are integer cents rather than formatted dollars, dates are ISO, every row carries a stable key, and running the same period twice produces byte-identical output, so a re-import matches the rows already there instead of doubling them. Anything left out — unpriced labour, uncoded rows, a cost code with no cost type in your own master — is counted on the file rather than dropped quietly. A person still does the import; we don't reach into anybody's general ledger.

Yes — they can just reply to the email. When the reply came from the address the RFI was routed to and the classifier is confident about what it's reading, the answer is matched to the open RFI and waits in the project's Approvals queue, so nobody has to re-key it or chase a portal login. Nothing from the email reaches the RFI until someone in the office who can see the project's full financials approves it; then it is recorded under their name and the status moves to answered. The proposal says whether the email carried a valid DKIM signature from the sender's own mail domain, which we verify ourselves — anyone can type a From line. Anything else waits in the project inbox for a person, which is what you want for a forwarded thread or a message from an address nobody recognizes. Owners, architects, and subs never need an account for any of it.

Yes, through the same magic-link portal they already use — one daily report per sub per day, with crew by trade, and a change request they price themselves rather than one you price for them. The daily report is where the useful part happens: their stated headcount for the day is compared against your own daily log's manpower for that same day, and where the two disagree, the disagreement is what you're shown. It isn't averaged, merged, or quietly corrected — the mismatch is the finding, and you get it the same week rather than in a claim eighteen months later. Who owes a report is worked out from who has an active subcontract and a live portal link, so it's a list nobody has to keep. Accept a sub's change request and the change order opens on your side already linked to it; once the owner's change order is executed, you flow it down to each affected sub as its own subcontract change order — at the price you set for that sub, never a copy of the owner's number — for them to acknowledge or dispute. That flow-down refuses while the owner's change order is unexecuted, because a sub building work the owner hasn't bought is how the money gets stranded.

No. Once a project has a real substantial-completion date on the record, the owner's portal grows a warranty section, and the owner can log a warranty callback themselves — what's wrong, where in the building, in their own words. It lands on your warranty register, you assign it to the responsible sub, schedule it, and close it with a closure note — and closing without a note is refused, because the note is your warranty defense a year later. There are no dollars anywhere on that surface: a callback is either warranty work or it becomes a priced change order through the normal chain. Substitutions run through the same portal discipline: a sub can file a substitution request — an or-equal request with the spec section, the proposed product, and their own cost and schedule delta — and you decide it on the record, with notes the sub reads if it's rejected, and one click to open a submittal to your architect if it needs their stamp.

Yes — the budget page carries an allowance ledger: each named allowance or owner-contingency line shows what was carried, what's been drawn, and what remains, and the remaining figure is always computed from the draws rather than stored anywhere it could drift. A draw can point at the change order that priced it, and when an allowance goes over, the page says so in plain words — an owner conversation to have now, not a surprise at closeout. On the sub side, when you release retainage to a sub, that's a retainage release record: the percentage and the dollar amount captured at release time, refusing to go past 100% of what was ever held, kept per subcontract. It records the release — your accounting still moves the money — so the trail survives an audit without pretending to be a bank.

Their walk lands as a field observation report — its own record with the date, the observer, and what they wrote, kept separate from your own daily log so your contemporaneous record stays yours. Lines from the report can become punch items with one step, and each one is tagged as coming from that observation. The punch list itself knows whose list an item came from: paste the architect's punch list in — one item per line, straight from their email — and every imported row carries its source, so 'your list' and 'their list' stop being one undifferentiated pile when the closeout argument starts.

You give a submittal the two facts only you know — the required-on-site date and the vendor's lead time — and the register works backwards to a must-approve-by date, which is the number that turns a submittal log into a schedule instrument. If either input is missing, the register shows a dash rather than inventing a deadline, and it tells you how many submittals are missing their dates instead of letting the blanks hide. When a submittal comes back revise-and-resubmit, the next revision carries the dates forward automatically — the rejection didn't move the day the steel has to be on site.

Two independent layers. Postgres row-level security is enforced in production and the app connects as a non-superuser role those policies apply to, so a query that loses its tenant context returns nothing rather than another customer's rows. On top of that, application guards derive workspace and project access from the signed-in user, revalidate child record IDs, and deny unassigned project access. External portal tokens are project-scoped, time-bounded, revocable, and hashed at rest. neuroBLDR does not use customer project data to train its own models, and AI features — the ones you run yourself and the drafting and classification agents that run on records as they're created — send only the context needed, through configured providers under their applicable data terms.

No. Our controls are built to be SOC 2-aligned, and our hosting and database providers, Vercel and Neon, publish SOC 2 Type 2 reports, but neuroBLDR is not itself SOC 2 certified and we won't describe it as anything else until it is. Implemented today: database-level tenant isolation (Postgres row-level security, enforced in production), application-layer access scoping, project-bounded external tokens, encryption in transit and at rest, and audit logging. Export of the audit log comes with the plans that include it. A SOC 2 Type II audit is on the roadmap. We are also not a HIPAA covered entity or business associate and don't sign BAAs. Need a DPA or a security questionnaire? Email the founder.

Your data is yours to export. From each project you can download its documents as a ZIP and pull its registers (RFIs, change orders, submittals, pay-app and WIP data) as CSV, and you can request a full export of your account data. After you cancel we retain your workspace so you can still export; contact us to pull a full copy or to request deletion. (We don't yet offer a one-click whole-workspace archive — it's on the list; for now the per-project exports plus an account export cover your records.)