SignPermit — Sign-code pre-check cockpit for US sign shops

1. One-liner

Pre-checks shop drawings against city sign codes and files permits so small US sign shops stop losing 30 hours per job.

2. Trend signal — why now?

Three things converged in the last 12 months:

1. Vision-capable LLMs got reliable enough to read engineered shop drawings. Claude 4.7 and GPT-4.5 can parse a scaled elevation drawing, extract sign dimensions, illumination type, mounting method, and electrical specs from a PDF, and cross-check those against a municipal sign-code rule set. That capability did not exist at production reliability 12 months ago.

2. The sign industry itself is calling permitting its #1 operational bottleneck. TheSignPack (“How to Fix the Sign Permit Drawing Bottleneck,” 2026) and Signs of the Times’ 2026 State of the Sign Industry both flag permit revision cycles as “a source of frustration, timeline compression, and margin erosion.” Sign Knights (typical sign permit timeline) puts standard turnaround at 3–8 weeks for mid-size cities, 10–12 weeks with variance, and notes 20–40% of applications get bounced for incomplete documentation.

3. AI permit-tech is being funded — but only for big construction. PermitFlow raised $54M Series B in Q4 2025 (TechCrunch, total $91M raised, called “TurboTax for construction permitting”). Pulley raised $4.4M seed. Both target commercial GCs and homebuilders on multi-million-dollar projects. Neither touches the on-premise signage niche — sign-job ACVs are too small for their economic model. The wedge is wide open underneath them.

Sign-maker pain is verbatim on industry forums. From Signs101.com (the largest sign-making forum):

“Pulling permits can be the ultimate bitch even when the political subdivision is staffed with nice folks.” — Signs101 forum member

“I have seen instances when requiring permission can add up to and beyond of an extra 30 hours onto the job!” — Signs101 forum member

“When the customer starts questioning why you are charging so much for just sorting the permission you can end up losing the job.” — Signs101 forum member

“In the last five years our local politics has forced changes in our sign codes about every month.” — Signs101 forum member

Provenance:

• Signal 1 (Demand): Industry trade pubs + sign-maker forum complaints quantifying 20–40% rejection rate, 30 extra hours per job, and category-wide frustration with permit turnaround — Sign Knights + TheSignPack 2026 + Signs101 forum — May 2026
• Signal 2 (Feasibility): Multi-modal vision LLMs (Claude 4.7, GPT-4.5) now parse engineered drawings reliably; municipal e-permit portals (Accela, ePlans/ProjectDox, OpenGov) expose document upload endpoints AI agents can drive — Accela ACA portals — May 2026
• Signal 3 (Economic): PermitFlow $91M total raised, $54M Series B Q4 2025 on construction permitting — sign permit niche left uncovered; sign industry $24B+, ~45,602 sign shops in US (avg 13 staff, $2.5M revenue) — PermitFlow TechCrunch + RentechDigital sign-shop count April 2026 — May 2026 Category: Workflow automation

3. The opportunity

Every on-premise sign that gets installed in the United States has to go through a building/zoning permit. The shop drawing — wind-load calc, elevation dims, mounting detail, electrical cut sheet, structural stamp — has to match a sign ordinance that’s been quietly amended by city council twice this year.

In ~30% of cases the city sends the packet back. Not for design reasons. For checklist reasons: PDF named in the wrong case, structural calc missing, mounting detail incomplete, color sample not attached. The shop owner now has to phone the planner, fix the packet, resubmit. Add a week. Add another week if it cycles again. Add another month if the sign needs a variance and the shop didn’t flag it up front.

PermitFlow doesn’t care about this market — a $1,800 sign permit doesn’t fit a $30K/month enterprise GC contract. The sign-shop management tools (ShopVOX, SignTracker, Cyrious Control) handle estimating and production but explicitly punt on permit submittal. The “expediter” agencies (Permit Place et al.) charge $1,500–$5,000 per commercial permit and operate as services, not software — so they have no incentive to disrupt themselves.

The gap is a sign-specific AI cockpit that knows the sign code of the 200 most-active US jurisdictions, reads the shop drawing, checks the packet for completeness before a planner ever sees it, and drives the municipal e-permit portal to file it.

4. Target market

• Primary customer: Owner-operator or “permit clerk / estimator” at US independent sign shops with 1–25 employees and $500K–$5M annual revenue. Member or prospect of ISA and 17+ state/regional sign associations. Typical shop pulls 5–40 permits per month across illuminated channel letters, monument signs, wall signs, and pylon signs.
• Why they buy: “Every rejected permit kills my margin on a $3,500 sign. I have one person whose half-time job is calling planning departments. I lost two installs last quarter to projects that timed out waiting on variances I didn’t see coming.” Their words: “permit chasing” is the part of the job no one wants to do.
• Rough TAM reasoning: ~45,602 sign shops in US (RentechDigital, April 2026). Realistic SAM = the 12,000 shops with 5+ employees doing 5+ permits/month. At 10% penetration over 3 years → 1,200 paying customers at $200–500/mo ACV = $3–6M ARR. Plus ISA member discount partnership lifts that.
• Why now for them: Margin compression from rising LED component costs + labor, more illuminated-sign zoning scrutiny in 2025–2026, and Gen-2 sign-shop owners (40s) taking over from retiring founders — they’re the first cohort actually willing to pay for SaaS in this category.

5. Product sketch (MVP)

• Drop-zone for shop drawing PDF; vision model extracts sign type, dimensions, illumination, mounting method, electrical load, sq-ft of sign face, height above grade.
• Jurisdiction picker (city + zoning district); pulls latest scraped sign code rule set for that location.
• “Pre-check report” before submittal: green/yellow/red on each ordinance criterion (max sq ft, max height, illumination type allowed, distance-to-property-line, internal vs. external illumination, electrical sealed cut-sheet present, structural calc present, owner-authorization form, PDF naming convention).
• Variance early-warning: when a design exceeds code, flag it before submittal and surface estimated variance lead time + filing path.
• Auto-generated jurisdiction-specific permit packet: cover sheet, applicant info, site plan reference, elevation pages, structural stamp page, electrical cut sheets — packaged in the file naming/format the city’s e-portal will actually accept.
• One-click file via Accela ACA / ProjectDox / OpenGov / municipal e-portal where supported; otherwise generates a print-ready submittal + check list for shops still walking permits in.
• Tracker: where each permit sits (intake, plans review, corrections, ready-to-issue), with auto-pull of correction notices from city portals and a re-package flow.
• Customer-facing status link the sign shop can forward to their end-client so they stop asking “where’s my permit at.”

6. AI angle — what’s load-bearing

Two AI capabilities are doing real work:

1. Vision-LLM extraction from shop drawings. Reading a scaled PE-stamped drawing and pulling out the 15–25 dimensional and material facts that map to a sign-code rule. This is not a chatbot bolted to a CRM — without it the product is a glorified checklist.
2. Sign-code rule reasoning + jurisdiction-specific submittal packaging. The product has to reconcile “Phoenix Zoning Ordinance §705 says max face area is 0.5 sq ft per linear foot of street frontage” against a drawing that says “120 sq ft channel letter face.” That’s structured-rule reasoning across a curated rule corpus that’s only economical to build because LLMs can ingest and normalize 200 city sign codes from raw municipal PDFs.

If we removed the AI, this product becomes a 200-jurisdiction wiki that ages out in 6 months. The AI is the moat against staleness.

7. Localization angle (if any)

N/A — this is a US-only play. The wedge is because every US municipality has its own sign code (~36,000 incorporated jurisdictions, ~200 cover the bulk of commercial signage volume). The product’s value compounds with each jurisdiction’s rule set added, and that’s a US-specific moat. Phase 2 (year 2+) could extend to Canada (similar bylaw-driven model) but I’d resist the EU/UK detour — sign-permitting in Europe is structured differently and dilutes focus.

8. Business model — path to $1M–$5M ARR

• Pricing: $149/mo Solo (1 seat, ≤10 permits/mo, top-50 jurisdictions) → $349/mo Shop (3 seats, 50 permits/mo, all 200 supported jurisdictions, e-portal filing) → $799/mo Pro (unlimited seats, white-label customer status pages, API for sign-shop management software integration).
• ACV: Blended $3,500–$4,800/yr.
• Path to $1M ARR: ~250 shops × $349/mo × 12 = $1.05M. That’s <2% of the 12K-shop SAM.
• Path to $5M ARR: ~1,200 shops × ~$350/mo blended + 5% on Pro tier + sign-code rule data licensing to ShopVOX/SignTracker for in-product pre-check ($150K–300K/yr). Realistic at year 3.
• Expansion path: Per-permit overage above plan; e-portal filing add-on for harder jurisdictions; “variance concierge” managed service tier at $499/permit (sign-shops happily pay this — Permit Place charges $1,500+); ISA referral program revenue share.

9. Go-to-market wedge — first 100 customers

1. Cold-DM the Signs101 forum + The Sign Syndicate forum. Both have ~50K+ active sign-making members. Specifically target threads tagged “permitting” — there are 200+ active threads in the past 24 months. Personalized Loom showing pre-check on their actual rejected drawings. Realistic 4–6% conversion on warm permit-pain threads.
2. ISA state-chapter sponsorship. 17 state/regional sign associations (Northeast States Sign Association, etc.) host quarterly meetings with 30–80 owners each. $2K sponsor fee + demo slot = 20–30 leads per event. Target 5 chapters in first 6 months.
3. ISA Sign Expo Orlando booth (April 2027). 20K attendees, ~600 vendors. Booth + permit-pre-check live clinic on attendee drawings. Expensive ($15K all-in) but the entire buying population walks the floor; one event can seed 200+ pipeline.
4. Integration with ShopVOX / SignTracker. Build a “Pre-check this drawing” button inside their workflow; revenue share with the management-software incumbent. They have ~6K paying shops between them.
5. Permit-rejection lead capture via city open-data. Many US municipal e-portals publish rejected permit applications. Scrape, identify the sign shop, send a free pre-check report on the exact rejected drawing. Realistic 3–5% reply rate.

If 4–6 weeks of (1)+(2)+(5) doesn’t produce 25 paid pilots, the wedge is wrong and we pivot the GTM order.

10. Build complexity — justification

Medium. v1 needs three real workstreams: (a) vision pipeline that extracts ~20 fields from PE-stamped sign drawings reliably across template variations; (b) a curated machine-readable rule set for the top 50 jurisdictions’ sign codes (one-time data labor — hire a paralegal-grade ops person for 8 weeks); (c) Accela/ProjectDox e-portal filing automation for ~15 portals that cover the bulk of the top 50 jurisdictions. None of this is research-grade. Two engineers + one ops person ship v1 in 14–16 weeks; jurisdiction expansion is post-launch ongoing.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	We’re a tool helping shops file their own permits; not impersonating licensed PE; structural stamps still come from real engineers.
Ethical — no harm / dark patterns	✅	Reduces rejection-cycle friction for SMBs and municipal staff alike.
Market exists (evidence above)	✅	45K shops, $24B industry, verbatim pain quotes, funded adjacent players.
1–5 person team can build this	✅	2 engineers + 1 ops person for jurisdiction rule curation.
Launchable with <$50K / ₹40L	✅	~$15K infra + LLM credits, ~$20K for the first 50-jurisdiction rule corpus, rest is founder labor.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	”30 extra hours per job,” 20–40% rejection rate, repeat pain on every permit. Owners actively pay for workaround (Permit Place). Hair-on-fire is too strong — but they’d pay this week.
Demand evidence	15	12/15	Industry pubs + forum quotes + funded adjacent players + agency outsourcers proving willingness-to-pay. Knock for: no direct survey data on sign-shop SaaS spend yet.
Build feasibility	15	11/15	Vision-LLM extraction is the riskiest sub-system; rule-set curation is unglamorous but doable; portal automation is mostly browser-agent work. 14–16 weeks to v1.
Distribution clarity	15	12/15	Forums + state-chapter sponsorships + ISA Expo are named, sized, and conversion math is plausible. Drag: aging buyer demographic slow to adopt.
Revenue mechanics	15	12/15	$349/mo blended ACV, 1,200 customer path to $5M, expansion via per-permit + variance concierge. Pricing is benchmarked vs. Permit Place’s $1,500 per permit.
Time to first revenue	10	8/10	Pre-sold pilots achievable in 6 weeks via direct outreach; paid v1 by week 14–16.
Defensibility	10	8/10	Soft moat: 200-jurisdiction rule corpus that compounds; portal-filing automation depth; ISA integration; sign-shop workflow lock-in via ShopVOX/SignTracker partnerships. Not patent-grade.
Total	100	80/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy (vision pipeline + browser-agent portal automation) · domain-expertise-required (sign industry advisor or ex-sign-shop ops person is essential to nail the data model).

Key assumptions to validate (3–5)

1. Assumption: Sign-shop owners will pay $349/mo and accept a tool taking permit-filing out of their hands. How to test: 30 cold-calls + 30-minute interviews with shops in 4 metros (Atlanta, Phoenix, Chicago, Orlando). Concrete asks: “If we caught 80% of rejections before submittal and filed via the city portal for you, what would you pay per month?” Hard kill: <40% commit verbally to $200+/mo.
2. Assumption: Vision-LLM extraction can hit ≥90% field-accuracy on real PE-stamped shop drawings across ≥5 drawing-template styles. How to test: Collect 50 historical drawings from 3 pilot shops; run extraction; manually score field-by-field; iterate prompting/fine-tuning for 2 weeks.
3. Assumption: The top 50 US jurisdictions cover ≥60% of commercial signage permit volume nationally. How to test: Sample 200 sign-shop project locations across 5 ISA member states; geocode against jurisdiction list; confirm coverage.
4. Assumption: Municipal e-portal filing automation is reliable enough not to introduce new rejection-cause errors. How to test: Run 25 staged “shadow” submittals via Accela/ProjectDox in pilot mode in 3 cities; verify file format, naming, and required-field acceptance vs. manual baseline.

Risk flags

1. Platform dependency: Municipal e-permit portals change without notice. Mitigation — maintain “print-ready packet” fallback so the shop can always walk a permit in if our automation breaks.
2. Aging buyer adoption risk: Sign-shop owner demographic skews 50s+ with low SaaS-buying muscle. Mitigation — anchor distribution in state-chapter live demos + Gen-2 owner Discord/forums, not pure inbound SEO.
3. PermitFlow horizontal expansion: If PermitFlow extends down-market into signage, they outspend us 100×. Mitigation — speed + sign-vertical depth + ShopVOX/SignTracker integrations PermitFlow won’t build. 24-month head start window.
4. Sign-code change cadence: Cities amend sign codes ~monthly per forum quote. Rule-corpus staleness is an existential risk. Mitigation — diff-monitoring on top-50 city codes via municipal-meeting-minutes scraping + a paralegal contractor on retainer.

14. Structured verdict

Score:                  80/100
Verdict:                STRONG GO
Confidence:             Medium
Best-fit builder:       Technical founder (vision + agent automation) + domain advisor from sign industry (ex-shop owner or ISA staffer)
Time to revenue:        12–14 weeks to first paid pilots; design partners closeable in 6–8 weeks
Capital to launch:      $35K–$50K (LLM credits, rule-corpus paralegal contract, infra, one trade-show booth)
Top 3 assumptions to validate first:
  1. WTP at $349/mo blended — 30 cold owner calls in 4 metros, ≥40% verbal commit at $200+ floor
  2. Vision extraction ≥90% field accuracy on 50 real PE-stamped drawings across ≥5 template styles
  3. Top 50 jurisdiction coverage ≥60% of commercial signage permit volume nationally
Kill criteria:
  - Abandon if <40% of 30 interviewed sign-shop owners commit to $200+/mo pricing
  - Abandon if vision extraction stalls below 85% field accuracy after 8 weeks of iteration
  - Abandon if PermitFlow announces a signage-vertical product before our v1 ships
  - Abandon if 3-pilot integration with Accela/ProjectDox introduces new rejection errors that exceed manual baseline

15. Next step — 1-week validation sprint

• Day 1–2: Build a target list of 60 independent US sign shops (5+ employees) across Atlanta, Phoenix, Chicago, Orlando. Pull from Signs101 forum activity + ISA state-chapter member directories + Google. Draft cold-DM + cold-call script anchored on “30 extra hours per permit” pain.
• Day 2–4: 30 conversations (DM + phone). Ask each: (a) how many permits/month, (b) rejection rate, (c) what do you currently pay anyone to handle permits, (d) would you pay $X/mo for pre-check + filing.
• Day 4–5: Collect 10 historical rejected shop-drawing PDFs from sympathetic owners (offer free pre-check report as the give).
• Day 5: Decide go / no-go on the basis of: ≥12 of 30 interviewed owners verbally commit to $200+/mo AND ≥6 of 10 historical drawings are extractable with current vision pipeline at ≥80% field accuracy on a single afternoon’s prompting work.

The falsifiable outcome is willingness-to-pay (>40% of interviews) and extraction feasibility (>80% accuracy on day-one prompting). If either misses, we adjust the wedge before building.