DeviceClock

1. One-liner

Runs the 90-day countdown between a failed elevator test and the automatic fine, and files the affirmation.

2. Trend signal — why now?

Three things moved at once.

The inspection system is visibly breaking. Massachusetts runs 54 inspectors against 41,000 elevators and reaches roughly 75% of them before certification expires. Los Angeles has 21,000 elevators and 15 LADBS staff — about 1,400 devices per person per year — with roughly 45% past due pre-pandemic. Michigan had 40% of inspector positions vacant and 78% of a sampled 74 routine inspections not performed. Washington: over half of 18,000 conveyance systems uninspected in 2018. Texas employs zero state inspectors, relying on 150 independent contractors who need a $50 fee and a three-day class. When public inspection capacity collapses, enforcement doesn’t go away — it migrates to automated paperwork penalties, which cost the city nothing to issue.

NYC turned that migration into a machine. Under NYC Rule §103-02, a Category 1 test report must be filed within 21 days of the test; defects found must be corrected within 90 days of the test date (the owner may request up to two 45-day extensions); and an Affirmation of Correction must be filed within 14 days of the correction. Miss the window and DOB NOW generates the violation automatically — no warning letter, no grace period. Failure to file an AOC (violation class ACC1) is $3,000 per non-compliant device, plus $150 per month per elevator for untimely filing. Failure to certify runs $1,000–$5,000 per device per month and compounds. A building with three elevators can stack $10,000–$20,000 within a few months of one missed date. DOB granted extended grace periods in 2022 and 2023; those are explicitly not guaranteed going forward.

The consolidation just happened. On 24 August 2026 — three days before this was written — ATIS, a large elevator inspection and consulting provider, acquired AuditMate, the $3.5M-seed startup that had built the software category for elevator maintenance-billing audits. The independent software layer for elevator spend just got absorbed into an inspection incumbent. Nobody replaced the piece neither of them sold: the correction clock between the test and the fine.

And the data to build on is free: NYC Open Data publishes DOB NOW: Elevator Safety Compliance, ~120,000 rows, one per device, with device number, device status, BIN, report filing dates and inspection dates, last updated June 2026.

Provenance:

• Signal 1 (Demand): NYC AOC failure = $3,000/device + $150/month/device; FTC $1,000–$5,000/device/month, compounding, auto-generated with no grace period; 3-elevator buildings hit $10,000–$20,000 in months — https://dobguard.com/resources/elevator-violations-nyc and https://sitecompli.com/blog/dob-issued-updates-on-2023-elevator-filing-penalties/ — 2026-08-27
• Signal 2 (Feasibility): NYC Open Data “DOB NOW: Elevator Safety Compliance” — ~120,000 device-level rows with device number, status, BIN, inspection and filing dates, free and public, updated 2026-06-11 — https://data.cityofnewyork.us/Housing-Development/DOB-NOW-Elevator-Safety-Compliance/e5aq-a4j2 — 2026-08-27
• Signal 3 (Economic): ATIS acquired AuditMate (MassMutual Ventures-backed, $3.5M seed) on 2026-08-24, consolidating elevator software into an inspection incumbent; independent elevator audits reportedly surface 10–60% in annual overcharges — https://www.globenewswire.com/news-release/2026/08/24/3349665/0/en/atis-acquires-auditmate-to-advance-a-next-generation-approach-to-elevator-management.html and https://auditmate.com/are-you-overpaying-for-elevator-service-heres-how-to-find-out/ — 2026-08-27
• Signal 4 (Structural): MA 54 inspectors / 41,000 elevators (~75% reached); LA 21,000 elevators / 15 staff; MI 40% inspector vacancy; WA >50% of 18,000 uninspected — https://propmodo.com/the-persistent-threat-of-uninspected-elevators/ — 2026-08-27 Category: Underserved niche (a hard-dated correction window that sits between two vendors who each sell the adjacent thing) + Workflow automation (a deadline nobody owns, enforced by an automated penalty engine)

3. The opportunity

The gap here is a handoff nobody owns.

The elevator company runs the Category 1 test and files the report. It hands the owner a defect list. Then it goes away — fixing the defects is usually a separate, quoted job, and the elevator company has no duty and no incentive to chase the owner’s filing deadline. The city, meanwhile, doesn’t see the defect list at all in any actionable feed. DOB knows a test was filed; the itemized defects live in the inspection report. The city’s next move is automatic: on day 104, generate the violation.

So the owner sits in a 90-day window holding a document with a list of defects, a vendor who considers the job done, and a penalty engine that will fire silently. Compliance-monitoring vendors — SiteCompli, Insparisk Command, KomplyOS — do a genuinely good job of surfacing violations that already exist, because that’s what the agency feeds expose. Insparisk Command is even free for a single building and scales to $20/$199/$799 tiers. But by the time a violation appears in any of those tools, the $3,000 has already attached. They are rear-view mirrors on a road where the crash is scheduled.

That’s the wedge, and it’s narrow on purpose: the interval between “you have defects” and “you have a violation” is the only place in this workflow where software can still save the owner money. ATIS/AuditMate now own the spend-audit side. The compliance platforms own the violation-surfacing side. The correction clock is the seam.

The 10× isn’t a smarter dashboard. It’s that today the defect list is a PDF or a paper sheet in an email from an elevator mechanic, written in trade shorthand (“governor tag missing,” “car top rail non-compliant,” “Cat 1 hydro relief”), and the owner of a six-unit walkup with one elevator has no idea which of those items is a 90-day clock item versus a note. Reading that sheet and turning it into dated, assigned, owner-legible tasks is exactly what a vision-and-language model does well now and could not do cheaply three years ago.

4. Target market

• Primary customer: Owner or managing agent of 1–30 NYC buildings with at least one elevator — the small multifamily landlord, the family-owned mixed-use portfolio, the boutique property manager with 8 buildings and no dedicated compliance staff. Above ~50 buildings they already pay SiteCompli and have a compliance coordinator; below that, the job belongs to whoever answers the phone.
• Why they buy: Because the penalty is automatic, per-device, monthly-compounding, and arrives with no warning. This is not a “should probably get organized” purchase. It’s a “I ate $3,000 last year on a form I didn’t know I had to file” purchase. The rule text is unambiguous, the fine schedule is published, and the failure mode is a date passing.
• Rough TAM reasoning: ~120,000 registered elevator devices in NYC per DOB NOW open data. Devices cluster in buildings; buildings cluster in portfolios. Conservatively that’s tens of thousands of distinct owning entities, of which the sub-50-building segment — the one no compliance vendor sells to seriously — is the large majority by count. Chicago ($200/day certificate lapse, up to $50,000 exposure on the AIC program), Boston/Massachusetts (escalating inspection fees $200→$800 by lateness, 90-day temporary certificates, shutdown on failed re-inspection) are the same shape and are the expansion path.
• Why now for them: DOB extended the filing grace period in 2022 and 2023. It has signalled those extensions are not guaranteed. Owners who were bailed out twice are about to find out what the rule actually says.

5. Product sketch (MVP)

• Drop the inspection report in, get a dated plan out. Upload the CAT1 report PDF (or photograph the mechanic’s sheet). Every defect comes back as a separate line item with a plain-English restatement, its correction deadline, and whether it’s a 90-day clock item.
• The clock itself. One screen per device: test date, day 21 (report filed), day 90 (defects corrected), day 104 (violation fires), plus the two available 45-day extension windows shown as branches, not fine print.
• Extension filing assist. Pre-fills the DOB extension request before the window closes, and tells the owner the honest tradeoff — an extension buys time and costs nothing; missing the date costs $3,000 per device.
• Affirmation of Correction packet. When repairs are done, assembles the AOC with the corrected-defect list, the repair invoice, and the date of correction, ready to file in DOB NOW — the 14-day sub-clock that catches people who did fix everything and still got fined.
• Vendor chase. Drafts the “these 6 defects, quoted or not, status please” email to the elevator company and re-sends it on a schedule, because the single most common cause of a blown 90 days is a repair quote sitting unanswered.
• Portfolio watch from public data. Enter your BINs; the tool pulls DOB NOW Elevator Safety Compliance and flags any device where a test has been filed but no correction activity has followed — including devices the owner forgot they had.
• The receipt. A per-device file — report, defect list, quote, invoice, AOC, filing confirmation — retained and exportable, because DOB asks for exactly this on a complaint-driven or audit inspection.

6. AI angle — what’s load-bearing

Remove the AI and this is a calendar with a nag, which is worth roughly nothing — the owner already knows the date exists in the abstract.

The load-bearing work is reading the defect list. CAT1 reports come as scanned PDFs, DOB NOW exports, and in the small-building world literally as photographs of a mechanic’s handwritten sheet. The content is trade shorthand referencing ASME A17.1 sections and device-specific components. Turning “gov. tag missing / car top stanchion NC / pit lite guard” into six discrete, dated, owner-comprehensible obligations — and correctly separating the items that start a 90-day statutory clock from advisory notes that don’t — is a document-understanding and domain-classification task. That is the product. Everything else is a table and a mail merge.

Second AI job, smaller but real: the vendor-chase and AOC drafting, where the model has to write a specific, non-generic email referencing the actual defect list and the actual deadline. Generic reminder emails get ignored by elevator schedulers; itemized ones get quoted.

7. Localization angle (if any)

N/A as a language play — this is US English. But it is intensely jurisdiction-local, which is the same thing wearing a different hat. The 21/90/14/104-day structure, the ACC1 and FTC violation classes, and the DOB NOW filing flow are NYC-specific. Chicago’s AIC program has a December 31 annual certification with $200/day lapse penalties and a different filing portal. Massachusetts escalates the inspection fee by how late you book and issues 90-day temporary certificates with shutdown on re-inspection failure. Each city is a separate rule engine and a separate content moat — which is precisely why a solo builder can own NYC before a national vendor bothers to encode it.

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

• Pricing: $29/month per building for 1–3 devices; $79/month per building for 4+ devices. Portfolio plan at $249/month up to 10 buildings, $599/month up to 30. Annual discount 2 months.
• Rationale for that number: it has to be trivially smaller than one avoided ACC1 fine. $3,000 per device is the anchor. At $29/month a single-elevator owner pays $348/year against a $3,000+ downside — an argument that closes itself on a phone call. It also has to sit under Insparisk’s $199 tier for the small owner, since Insparisk’s free tier is real competition for violation surfacing and this must be positioned as a different job.
• ACV: ~$600 blended (mix of single-building $348 and portfolio accounts).
• Rough math to $1M ARR: ~1,650 accounts at $600 ACV. Against tens of thousands of sub-50-building NYC owning entities, that’s low-single-digit penetration of one city.
• Rough math to $5M ARR: needs NYC penetration plus two more jurisdictions (Chicago and Boston/MA are the obvious pair, both with published per-day or escalating penalties), plus the expansion path below. Roughly 8,000 accounts. This is the point where it stops being a solo project.
• Expansion path: devices per building → buildings per portfolio → adjacent auto-penalty filings the same owner faces on the same DOB NOW portal (boiler, facade/FISP, backflow). The customer relationship is “we run your dated filings”; elevators are the wedge because the penalty is the most brutal per-device.

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

This is the strongest part of the idea and the reason it survived the filter. The prospect list is public and pre-qualified by the customer’s own violation history.

• Mine DOB NOW open data for the already-burned. Pull the Elevator Safety Compliance dataset (~120k device rows) plus DOB violation records. Filter to entities that have already received an ACC1 or FTC violation in the last 24 months and own fewer than 50 buildings. These are not cold leads — they have a receipt for exactly the pain being sold. Cross-reference BIN to owner via ACRIS/HPD registration for contact details. Expect this list to be in the low thousands.
• Lead with their own device number. Email subject line is the device number and the date. Body: “Device 1E-12345 at 214 W 47th filed a CAT1 on 3 March. Your correction window closed 1 June. Here’s the $3,000 line item.” Personalized to a public fact about their building beats any demo video. Target 25–30% open, 5–8% reply on a list this qualified.
• Sell through the people who already get blamed. Small NYC managing agents and expediters (All Boro, DOB Guard and dozens of one-man expediter shops) currently eat these fines as a client-relationship problem. Offer them the portfolio tier at a referral margin. Ten expediters with 30 clients each is the fastest path past 100.
• Show up where the fine gets discussed. The NYC landlord/property-manager forums, BTEANY and local owners’ associations, and the comment threads under every “NYC elevator violations explained” article — currently owned by elevator contractors and law firms selling remediation after the fine. A free “check your device” tool that reads the open data and returns the owner’s actual clock status is the top-of-funnel; it costs nothing to run and it terrifies people accurately.
• The elevator companies themselves, carefully. A small elevator service company that hands clients a defect list and later gets blamed for the fine has a reason to co-brand this. Not a channel to bet on — their incentives are mixed — but worth 5 conversations.

10. Build complexity — justification

Low. The document-understanding piece is an off-the-shelf multimodal model with a domain prompt and a defect taxonomy that has to be hand-built once by reading real CAT1 reports. The open-data ingest is a scheduled pull from a public Socrata endpoint. The clock is date arithmetic. The filing packet is document assembly, not portal integration — v1 hands the owner a ready-to-submit packet rather than automating DOB NOW submission, which sidesteps the only genuinely hard integration. A solo builder ships this in 6–8 weeks; the real work is the defect taxonomy and the rule encoding, not the software.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helping an owner meet a published filing deadline. No licensed activity — the product prepares packets, the owner files and signs.
Ethical — no harm / dark patterns	✅	The incentive is aligned with the safety rule: it pushes defects to get fixed on time, not papered over. Worth stating plainly — the product must never help file an AOC for uncorrected defects.
Market exists (evidence above)	✅	Published per-device penalty schedule, ~120k devices, an incumbent just got acquired.
1–5 person team can build this	✅	Solo, 6–8 weeks to v1.
Launchable with <$50K / ₹40L	✅	Under $10K. Open data is free; inference is the main variable cost and it’s per-report, not per-seat.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	17/20	Automatic, per-device, monthly-compounding fine with no warning and no grace period. $3,000 ACC1 + $150/month. Not 18–20 because a single-elevator owner feels it once a year, not daily — it’s a sharp annual spike, not a hair-on-fire continuous bleed.
Demand evidence	15	12/15	Published penalty schedule, DOB’s own 2023 enforcement notice on unfiled 2018–2022 reports, an acquisition in the adjacent space, and a live paid market of expediters selling violation removal. Docked because I found no verbatim owner complaints — the forum voice for this niche is thin, and that’s a real gap in the evidence, not a rounding error.
Build feasibility	15	13/15	Off-the-shelf multimodal extraction, public Socrata feed, date math. Custom work is the defect taxonomy. No portal integration in v1.
Distribution clarity	15	13/15	Genuinely excellent: the prospect list is public, and it can be filtered to owners who already paid this exact fine. Outreach personalizes on a public device number. Docked 2 because owner-contact resolution from BIN via ACRIS/HPD is messier in practice than it sounds.
Revenue mechanics	15	11/15	Pricing anchors cleanly against a $3,000 fine and undercuts Insparisk’s paid tiers. But ACV is low, it’s a per-building SMB sale, and 1,650 accounts of small landlords is a lot of individual closes for $1M.
Time to first revenue	10	8/10	Can pre-sell off the public violation list before the product is finished — the pitch is a fact about the buyer’s building. Realistically 6–10 weeks to first paid.
Defensibility	10	2/10	This is the weak axis and it should be. Encoded rules and a defect taxonomy are a 6-month head start, not a moat. ATIS/AuditMate could bolt this on. The only durable asset is per-device correction history accumulating in the account, and that compounds slowly.
Total	100	76/100

13. Qualitative modifiers

Founder-fit tags

technical-heavy · sales-heavy

Technical to build the extraction and taxonomy; sales-heavy because 1,650 small landlords do not self-serve their way in. Somebody has to run the outbound list and get on the phone. A builder who won’t cold-call NYC managing agents should not do this one.

Key assumptions to validate (3–5)

1. Assumption: Small owners genuinely miss the 90-day window at a meaningful rate — the ACC1/FTC violation volume against sub-50-building owners is in the thousands per year, not the dozens. How to test: Pull the DOB violation dataset, count ACC1 and elevator FTC issuances over 24 months, join to owning entity, and segment by portfolio size. This is a free query and it either proves or kills the idea in a day. Do this first.
2. Assumption: Owners will pay $29/month to avoid a fine they believe their elevator company is already handling. How to test: 40 calls to owners who took an ACC1 hit in the last year. The question is not “would you buy” but “who did you think was filing that form?” If most say “my elevator company does it,” the sale is an education problem and CAC triples.
3. Assumption: CAT1 defect lists are extractable at usable accuracy across the real distribution of formats, including photographed handwritten sheets. How to test: Collect 25 real reports from 5 sources, run extraction, hand-score defect-level recall. Below ~90% recall on clock-starting items, the product is a liability rather than a service.
4. Assumption: DOB does not remove the pain by extending grace periods again. How to test: Track DOB service notices; treat any new blanket extension as a demand shock.

Risk flags

1. Defensibility: Low, openly. The rule encoding is copyable and ATIS just bought the company best positioned to copy it. The bet is speed and segment focus — ATIS sells to the portfolios that pay for inspections and consulting, not to a six-unit walkup at $29/month. That gap is real but it is not permanent.
2. Single-jurisdiction concentration: v1 is NYC-only. NYC could soften enforcement (it has twice), and the whole demand curve moves. Chicago and Boston/MA are the hedge, and they should be started before NYC is saturated, not after.
3. Liability: Telling an owner a deadline and being wrong is a bad failure mode. The product must present itself as a tracker and packet-preparer, with the owner signing and filing. Errors-and-omissions coverage before the first paying customer, not after.
4. Channel conflict: Elevator service companies may see this as adversarial — it creates a documented record of how long their quotes sat unanswered. That is a feature for the buyer and a reason the most obvious partnership channel may stay closed.
5. Evidence thinness: No verbatim customer voice was found for this specific pain. Every number in this proposal is from a rule text, a penalty schedule, a government dataset, or a law-firm/vendor explainer — all sourced, none fabricated — but I have not heard an owner say it in their own words. That is why confidence is Medium and why assumption 2 is a phone call, not a survey.

14. Structured verdict

Score:                  76/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical solo founder who will personally cold-call NYC managing
                        agents; NYC-based is a real advantage for the first 50 accounts
Time to revenue:        6–10 weeks
Capital to launch:      <$10K
Top 3 assumptions to validate first:
  1. ACC1/elevator-FTC violation volume against sub-50-building owners is in the
     thousands/year — free query against NYC Open Data, one day, do this before anything else
  2. Owners believe someone else files the AOC — 40 calls to recently-fined owners,
     ask "who did you think was filing that form?"
  3. Defect extraction hits ≥90% recall on clock-starting items across 25 real reports
     including handwritten sheets
Kill criteria:
  - Abandon if fewer than ~500 ACC1/FTC violations per year land on owners with <50 buildings
  - Abandon if fewer than 8 of 40 recently-fined owners will put money down after the pitch
  - Abandon if defect extraction recall stays under 90% after two iterations on real reports
  - Abandon if DOB announces another blanket filing extension covering the next cycle
  - Abandon if ATIS ships an owner-facing correction tracker under $50/month before v1

15. Next step — 1-week validation sprint

• Day 1: Query NYC Open Data — DOB NOW Elevator Safety Compliance joined to DOB violations. Count ACC1 and elevator FTC issuances over 24 months. Join to owning entity via BIN, segment by portfolio size. This is the whole idea in one query. If the sub-50-building segment isn’t absorbing hundreds to thousands of these a year, stop here and don’t spend day 2.
• Day 2: Build the target list — owners with a recent elevator violation and fewer than 50 buildings. Resolve contacts via ACRIS/HPD registration. Measure what fraction of the list yields a usable email or phone. If contact resolution is under 30%, the distribution score above is wrong and needs re-scoring before proceeding.
• Day 3–4: 40 calls, using their own device number and violation date as the opener. One question carries the sprint: “When you got that defect list, who did you think was filing the correction?” Then ask for $290 for a year, prepaid, before the product exists.
• Day 5: Collect 25 real CAT1 reports from those calls and from public sources. Run extraction. Hand-score defect-level recall on clock-starting items.

Go/no-go, falsifiable: proceed only if (a) day 1 shows ≥500 relevant violations/year against sub-50-building owners, and (b) ≥8 of 40 owners put money down, and (c) extraction recall on clock-starting defects is ≥90%. Any one of the three failing means no build. Getting 8 verbal “sounds great, send me info” and zero prepayments counts as a failure, not a maybe.