Municipal plan check — product brief

Product context: Estym8 is built from scratch as an AI-first construction preconstruction platform—not a legacy takeoff stack with AI bolted on. Canonical framing: AI-first positioning. Estym8’s buyer remains estimators / GCs / full-project precon (multi-trade takeoffs from plan sets and bid folders). This brief is a separate product for municipalities.

Status: Idea / discovery (2026-07-17; refreshed 2026-07-24 with Paul’s suite notes — back-check, comment log, parcel/feasibility, competitor clusters; Revit / BIM import elevated to a flagship municipal capability). Triggered by Mike Faust, Peoria City Manager; deepened by Paul (design QA / municipal-adjacent workflow).
Positioning decision: Same plan-reading engine, separate product surface — do not reposition Estym8 as “city software,” and do not sell city estimation / BOMs.
Working brand: ClearToPermit (cleartopermit.com / .ai — confirm registrar). Secondary: ReadyForPermit. Do not lead with AHJ*.
Share with Paul / partners: estym8.ai/docs/municipal-plan-check (after deploy).

Jump links: Peoria discovery questionnaire (ask Mike) · Staffing positioning · UX: workflow, not chatbot · Peoria workflow captured

Related: paul-notes-construction-ai-suite-2026-07-24.md · plan-intelligence-skus.md · ae-design-review.md · Estym8 folder workflow · AI_FIRST_POSITIONING.md.

---

Staffing positioning — empower reviewers, never replace them

ClearToPermit makes government reviewers more efficient and better at their jobs. It does not replace plan examiners, eliminate front-desk or discipline review roles, or take the stamp.

This is a political and product constraint for Peoria and every city: sell capacity and quality for the people already doing the work, not workforce reduction. Same culture as Estym8’s human-in-the-loop estimator story.

All ClearToPermit copy must reflect this

When we build the product (UI, emails, onboarding, marketing, pilot decks, public docs), every customer-facing string follows the voice rules in docs/MICROCOPY_GUIDELINES.md — ClearToPermit / municipal copy voice: preferred language, banned replacement/automation phrases, UI pattern examples, and the copy-audit gate before pilot release. No exceptions for “marketing-only” or “internal pilot” surfaces that cities will see.

---

UX principle — cohesive workflow with AI assist, not a chatbot

Suite-wide (ClearToPermit, ClearToIssue, Estym8, Bodi): we design a cohesive product experience and real job workflow. AI assists inside that workflow. We do not ship “a chatbot that talks about permits.”

For ClearToPermit specifically: front desk sees a completeness workspace; examiners see a review workspace with draft comments and sheet evidence; AI fills and suggests; humans drive every stage. Chat may exist inside a case — it never is the product.

Same bar for ClearToIssue (design QA workflow) and any future suite app. Detail also in plan-intelligence-skus.md and microcopy guidelines.

---

One-liner

Ingest a permit package — PDFs and native Revit / BIM when applicants have them → help city staff flag completeness gaps, drawing conflicts, and locale-code concerns faster — with citation-backed deficiency packages and resubmittal back-check. Reviewers stay in charge. Not a bid estimate for the city. Not a replacement for examiners.

---

Job (and what it is not)

Success metric = reviewer efficiency + deficiency quality + trust (false-positive rate) — not “FTEs removed,” and not F4c vs human takeoff.

Flagship differentiator — Revit / BIM import (HUGE)

Most municipal AI peers are PDF- (and sometimes CAD-) first. ClearToPermit should treat native Revit (.rvt) / BIM import as a headline capability, not a v3 footnote:

Honesty: Estym8 today is PDF-strong. Revit import is a priority spike + product investment for ClearToPermit (licensing, APS/Forge or equivalent extract path, security for city-held models). Do not claim GA Revit until the spike proves extract + a first rule pack on a real pilot model. Do claim the roadmap loudly with Mike — this is a huge reason cities would pick us over PDF-only copilots.

---

Competition — suite vs municipal lane

Paul (2026-07-24) flagged the municipal field and sharpened it into three clusters. We already tracked CodeComply / Permittable. Use the table for lane awareness; lead investor/partner narrative with no suite peer.

Paul’s three municipal clusters (lane competitors)

Detail by player (Paul + open research)

Where ClearToPermit sits (honest)

Verdict (municipal lane): There is competition at “AI for permitting.” None are “Estym8 for cities” (full-project organic takeoff → BOM). Many are already closer to ClearToPermit’s job than our estimating SKU is. Our edge in this lane is not “nobody else does AI.”

Verdict (suite): There is no AI-first competitor for the suite we are creating — Estym8 + Bodi + ClearToPermit + ClearToIssue on one greenfield spine as workflow apps. Point tools fill lanes; nobody assembles the portfolio. Canonical wording: plan-intelligence-skus — Suite competitive position.

ClearToPermit edge in-market:

1. Reuse a battle-tested plan-set engine (folder ingest, classify, sheet intel, organic legend/schedule/conflict evidence) already paid for by the estimating product.
2. Honest, evidence-first concerns + back-check (Prime Directive culture; Paul’s highest-value municipal workflow) rather than black-box “fail the permit.”
3. Native Revit / BIM import as a HUGE municipal differentiator — relational model QA + Revit↔PDF consistency, not glyph-only.
4. Separate SKU so Estym8 stays the full-project estimating platform while ClearToPermit owns municipal GTM.
5. Do not compete as “another Archistar/CivCheck” on day one — win the Peoria-style pilot on messy package completeness, conflicts, citation-backed comment log, resubmittal back-check, and (as soon as spiked) Revit-aware review.
6. Respect Symbium’s lane — if Mike cares about solar/EV/panel instant permits, that is a different product shape (deterministic narrow scopes); partner or defer, don’t fake it with LLMs.

Do not claim empty-field for municipal AI alone. Do claim no AI-first suite peer. Sell speed-to-value from shared engine + Revit path + design-partner depth + defensible flags — and suite optionality.

---

Can it share Estym8’s pipeline?

Yes — share the spine, swap the job head.

[ Shared ]                    [ Municipal job head ]
folder / PDF ingest      →    permit-type routing
classify process/skip    →    jurisdiction checklist completeness
sheet / discipline intel →    required exhibits & sheet gaps
legend / schedule /      →    evidence for conflicts
  organic reads                 (not bid line items)
concerns / conflict log  →    city comment packages + severity
project intelligence     →    staff package overview
                             + locale rule packs / amendments
                             + queue, assign, audit, export

Architecture rule: shared packages / services; separate app surface, auth, pricing, and roadmap. Do not full-clone the Estym8 repo into a second monolith “city Estym8.”

Prime Directive still applies to organic reading. Authority for governance here is city checklists + adopted codes / local amendments — never invent violations; never use a human estimator BOM to “prove” compliance.

---

Where AI earns its keep

This product is AI-assisted review, not “AI stamps the permit.” Use models where they compress messy plan packages into evidence-backed work product for humans. Prefer the same organic-read culture as Estym8: cite the sheet, show the crop, label uncertainty.

A. High-value AI jobs (ship early)

B. Strong mid-term AI jobs (after pilot trust)

C. Where not to lean on AI (liability / trust)

• Auto-approve / auto-deny permits
• Silent “learning” that changes live rule packs without city/ops review
• Structural calc verification, life-safety engineering judgment, or PE stamp replacement
• City cost estimation / BOM (wrong product)
• Inventing violations without drawing or checklist/code evidence
• Using Estym8 F4c / Jonas BOM accuracy as the municipal success metric

D. AI architecture patterns (reuse Estym8)

1. Temperature 0 + consensus on any extraction that feeds a deficiency (sheet inventory, schedule rows, conflict facts).
2. Citation contract — every flag must carry: rule/checklist id · sheet ref · evidence snippet/crop · confidence band · suggested vs confirmed.
3. Human-gated promotion — staff-missed findings and dismissals become training candidates, not silent weights.
4. Per-jurisdiction isolation — Peoria feedback does not silently rewrite another city’s packs.
5. Audit-first — model + rule-pack versions on every run (same append-only trail as §E).
6. Multi-model routing — cheap classify/harvest vs stronger vision for conflict crops; keep cost visible for city procurement.

E. Design-partner AI experiments (Peoria-shaped)

Concrete experiments that prove AI value without building a full ICC expert system:

1. Completeness-only pilot — checklist map on 10–20 historical packages; measure time-to-first-deficiency-list + false-positive rate.
2. Top-10 conflict classes — pick Peoria’s most common bounce reasons; AI flags only those with evidence.
3. Draft vs confirm A/B — examiner time with AI-drafted comments vs blank form.
4. Resubmittal clearance assist — did the new set actually address last cycle’s list?
5. Staff-missed capture — count how many examiner pins become reusable checklist rules after admin review.

Success for these experiments = reviewer minutes saved + trusted flags, not “AI found every code issue.”

---

Who buys / who uses

Buyer = municipality (or regional shared-services plan-check). User = plan-check staff. Estym8’s existing buyer stays on the estimating product.

Who submits the plan set (industry norm)

Formal permit packages usually arrive from the applicant of record, not from “the city’s estimator”:

v1 product surface is staff-facing intake + review (city uploads or pulls a package into the queue). An applicant portal (pre-check before official submittal) is a later dual-sided play — same engine, different auth and UX.

Product surface (architecture reminder)

• Share: folder/PDF ingest, classify, sheet intel, organic legend/schedule/conflict reads, multi-model routing.
• Separate: app UI, auth/tenancy, pricing, queue/assign, deficiency export, jurisdiction rule packs, roadmap.
• Do not ship a city-skinned Estym8 estimator or city BOM features.

---

Core capabilities (v1 → v2)

A0. Revit / BIM import (HUGE — flagship; spike → pilot)

Treat this as a first-class municipal product pillar, not “later geometry.”

Guardrails: human-in-the-loop; never invent model elements; if extract fails, fall back to PDF path with explicit “model unread” concern. Shared spine can later feed ClearToIssue / Estym8; ClearToPermit leads the investment because cities feel the Revit gap hardest.

A. Package intake & completeness (MVP)

• Upload permit set (folder or multi-PDF) on the shared ingest spine — plus .rvt / BIM when provided (see A0).
• Detect disciplines and sheet designators (A / S / E / M / P / C / fire, etc.).
• Against a jurisdiction permit checklist: missing sheets, obsolete revisions, blank sheets, wrong package type; model required / model present / model unreadable.
• Output: completeness score + must-fix deficiency list.

B. Concerns & internal conflicts (MVP)

Reuse Estym8 forensic / concern patterns for a city audience:

Each concern: severity, sheet refs, evidence, suggested reviewer note, optional applicant-facing wording.

C. Conflicts with city permit codes (MVP → v1.5)

Municipal differentiator — not a full ICC expert system on day one.

1. Jurisdiction profile — adopted codes (IBC / IRC / NEC / IMC / IPC year) + local amendments.
2. Rule packs — machine-checkable rules first; soft rules escalate with citations.
3. Citation-backed flags — code section or amendment ID + drawing evidence + confirm / dismiss / escalate.
4. Honest limits — suggested conflict vs proven violation labeled clearly.

Out of scope early: replacing PE/architect stamp judgment; automatic approval; full structural calc review; any city cost estimation.

D. Reviewer workflow (v1)

• Queue (new / in review / returned / advanced).
• Assign + full audit trail (see below).
• Export comment letter / deficiency report (PDF + JSON).
• Primary UX = workflow surfaces (completeness board, concern list, comment log, back-check) — not a chatbot shell.
• Optional scoped assistant on this package + this city’s rule pack (helper only).
• Staff-added findings — flag issues the AI missed (sheet pin + note + optional code cite); those become training/governance signals (see Staff feedback loop).

E. Submittal audit trail (MVP — first-class)

Every package cycle must be reconstructible for liability, FOIA-ish transparency, and pilot metrics:

Audit records are append-only. Suggestions never auto-approve or auto-deny a permit.

F. Staff feedback loop (missed findings)

Cities will catch things we miss. That must improve future packages — without inventing violations or overfitting one examiner’s taste.

In-product (v1):

1. Examiner selects sheet (and optional crop/pin) → describes the missed issue → optional code/checklist cite → severity.
2. Finding is stored on the package audit trail and can be included in the deficiency export.
3. Status: staff_added (distinct from AI-origin concerns).

Learning pipeline (v1.5 → v2) — human-gated:

Guardrails (non-negotiable):

• No silent auto-learning into production rule packs — city admin or Estym8 ops reviews promotions.
• Never train on “approve/deny” outcomes as ground truth for code compliance.
• Per-jurisdiction isolation by default (Peoria feedback does not silently change another city’s packs).
• Opt-in anonymized cross-city learning only with contract language.
• Prime Directive: staff pins are evidence for review, not a license to invent counts or violations on other sets.

G. Back-check workflow (Paul — highest municipal value)

After initial review, cities keep a running comment list. On resubmittal, compare three inputs:

1. Original submission
2. Reviewer comment list
3. Revised submission

For each prior comment, determine: change made? where? fully addresses concern? new coordination problems? unrelated plan areas modified? close vs human still needed?

This eliminates repetitive reopen-find-check work. Spec as v1.5 after completeness + comment log MVP; reuse Estym8 revision/diff patterns where possible. Detail: paul-notes §4–5.

H. Tabular comment log (Paul)

Reviewers work from spreadsheets. AI pre-populates rows; humans confirm:

Copilot: reviewer pins one line (“conduit through rated wall without penetration detail”) → system fills sheet, wall type, code, detail refs, similar occurrences.

I. Parcel / zoning / feasibility (v2+)

Validate title-block address, parcels, plat dimensions, setbacks, coverage, easements, landscaping separations against public records; early developer feasibility (buildable envelope) before design. Discovery — needs data partners. See Paul notes §8.

J. Later (v2+)

Zoning exhibits · multi-department routing · applicant portal · design-build phased packages · anonymized cycle-time analytics · Accela/Tyler · supervised promotion of staff-missed patterns into jurisdiction rule packs · discipline deep-dives (panel load, plumbing slope, civil geometry).

(Revit / BIM is not “later” — see A0.)

---

Naming

What “AHJ” means

AHJ = Authority Having Jurisdiction (the city/county/agency that enforces the codes). Correct industry term; poor consumer brand — most city managers and elected officials do not lead with “AHJ” in plain speech.

Available domains you mentioned

Recommendation: keep AHJClarity/AHJClear as backup / secondary domains if you want industry SEO later — do not lead the brand with AHJ.

Recommended direction (clear / fast / code-compliance — 2026-07-17 refresh)

Prior shortlist (PermitClarity, ClearSubmittal, LocaleClarity, AHJ*) hit taken/unavailable at registrar. Fresh candidates below checked via whois/DNS as likely open (confirm at registrar before buying — parking/premium holds can still block).

Also strong backups (whois-open signal): ClearForPermit, GetPermitClear, CivicClear.ai, ReviewClear.ai, PermitComply.ai / .com, PassFirst.ai, PackageReady.ai, ClearLocale.com.

Avoid leading with: AHJ*, generic Clearance.com (taken), PermitReady (taken), ClearPermit.com (taken).

Tagline patterns (any brand):

• “Plan-set intelligence for faster permitting.”
• “Ingest the package. Flag locale concerns. Send a clear deficiency list.”
• “Not an estimate — a first-pass permit review assist.”
• “Clear the path to the permit.”

Brand relationship to Estym8: “Powered by the same plan-reading engine as Estym8” in technical/partner materials; separate logo, domain, and sales motion for cities.

---

Marketing plan

Positioning

• Category: AI-assisted municipal plan review / permit package intelligence.
• Against incumbents: faster first-pass + citation-backed concerns; human stays in the loop.
• Against CodeComply / Blitz / Permittable: win on messy full-folder organic reads + design-partner depth + honest evidence UX; do not pretend we already own 850-city distribution.
• Against “Estym8 for cities” confusion: never lead with takeoff/BOM; lead with permitting outcomes.

Audiences & messages

Channels (phased)

1. Design partner (Peoria / Mike Faust) — discovery → 10–20 historical packages → measurable pilot. Case study is the #1 asset.
2. Peer cities — warm intros from city managers; ICMA / NLC / state municipal leagues once proof exists.
3. Content — “What a first-pass AI deficiency list looks like” (anonymized); locale-amendment stories; not estimating content.
4. Procurement path — only after pilot: cooperative purchasing, RFP language, export-to-Accela/Tyler story.
5. Do not run Estym8 estimator ads for this SKU; separate site, LinkedIn, and email lists.

90-day marketing outcomes

• Named design partner + pilot scope written.
• One public one-pager + private pilot deck.
• Brand + domain locked (prefer ClearToPermit or ReadyForPermit; see Naming).
• 3 discovery calls beyond Peoria.
• Zero claims of automatic code compliance or city estimation.

What not to market

• “We estimate for the city.”
• “Automatic permit approval.”
• “AI replaces plan examiners” / “cut staffing” / “no human needed.”
• “No competition.”
• Estym8 F4c / Jonas BOM accuracy as the municipal proof point.

What to market (staffing-safe)

• “Makes your reviewers faster and more consistent.”
• “Front desk catches incompletes before they clog the queue.”
• “Examiners spend time on judgment — AI drafts the comment list.”
• “Humans always confirm; AI never stamps.”

---

Go-to-market & Peoria discovery

Near-term: Peoria as design partner, not a rewrite of Estym8.

Peoria workflow (Mike meeting 2026-07-24) — captured

Peoria discovery questionnaire

Find this from the top of the doc: jump link → Peoria discovery questionnaire.
Shannon to ask Mike / staff; paste answers under each item when known (or into docs/ops/peoria-discovery-notes.md).
Status: [ ] open · [x] answered.
Reminder when asking: we are positioning as staff efficiency, not job replacement — frame questions that way.

Must-have (locks MVP)

1. Completeness checklist — What’s on the front-desk “complete vs reject” list today? (doc types, naming, seals, sheets, calcs, applications.) Can we get a copy?
2. Pilot permit type — Which package type first? (e.g. new commercial, TI, residential, site.)
3. 90-day win — What would make Mike say this worked? (e.g. cut incomplete accepts, faster front-desk, fewer bounce cycles.)
4. Who clicks — Front desk only for phase 1, or also building/fire reviewers from day one?
5. System of record — Accela / Tyler / other? Is export-only OK for a pilot, or must we live inside their portal?
6. Historical packages — Can we get 10–20 anonymized electronic packages (complete + rejected) with outcomes?

High value (shapes product)

7. Environmental — Confirmed: reviews depending on project type, as required (not every package). Still open: exact trigger list / permit types that pull environmental in.
8. Routing — Parallel reviews or sequential? Who consolidates comments back to the applicant?
9. Comment format today — Spreadsheet, portal markup, letter? Sample of a real deficiency package?
10. Codes / amendments — Which editions + local amendments must appear on a comment? City-authored checklist vs licensed content?
11. Revit in practice — How often do applicants send .rvt? Who opens them? Any rules (“model preferred / optional / rare”)?
12. PDF for field — Who prints/publishes the inspector set? Same sheets as submitted, or a stamped “approved for construction” set?
13. Resubmittals / back-check — How do they track “comment closed vs still open” today?
14. Liability line — What must never be automated or auto-rejected by AI?

Nice-to-have (GTM / ops)

15. Volume — Packages/month; % rejected at front desk; avg days intake → first review.
16. Procurement — Pilot path (design partner / MOU / RFP later)? Timeline they have in mind?
17. Stakeholders next — Who should we meet (permit tech lead, building official, fire marshal, IT)?
18. Competitors they already saw — Any Archistar / CivCheck / Blitz / Accela AI demos?

Pilot shape (default until refined): front-desk completeness on Peoria checklist × 10–20 historical electronic packages → time-to-reject/accept + false incompletes → then one discipline’s first-pass flags.

---

Success metrics

---

Risks

---

Open decisions

• Final product name / domain (recommend ClearToPermit or ReadyForPermit; confirm registrar)

• Brand relationship copy (“powered by Estym8 engine” vs silent)

• First jurisdiction profile schema (Peoria as template)

• MVP permit type for pilot

• Revit / APS spike owner + licensing path (ClearToPermit flagship — do not bury)

• City-authored checklists vs licensed code content

• App home (subdomain vs separate domain)

• Pricing (annual city license vs per-package)

• Staff-missed finding → rule-pack promotion workflow (who approves; per-city vs opt-in cross-city)

• Retention / export of audit trail for FOIA and design-partner metrics

• Copy voice gate — ClearToPermit UI/marketing strings follow MICROCOPY_GUIDELINES — ClearToPermit voice; wire bans into copy-audit when first screens land

---

Document history