New startup ideas · AI and software · Data for AI
startup idea
Registral
Accredited AI agents that turn government records into standardized, machine-readable datasets.
An agent workforce, sold to government agencies through system integrators, that reads scanned permits, inspection reports, court filings and statistical registers and converts them into certified, AI-ready datasets under a published schema.
- AI agent as a service
- Government and public sector
- $10-100B market
- Platform others build on
- Global from day one
3/5
venture judge
1
similar startups, last 2 years (4 all-time)
95%
of 3 nearest real companies still alive
yes
8 matching federal grants and programs
Direction supported by government programs and grants
Test it before you build it
$800 · 6 weeks · 25 prospects
For $800 in 6 weeks, prove that two municipal permitting departments will commit $4,900 each, by P-card or purchase order under their small-purchase threshold, to convert scanned permit backlogs before any accreditation or platform exists.
Riskiest assumption · A municipal permitting official will spend discretionary budget this fiscal year on scanned-permit conversion as a small purchase, instead of leaving it inside a future integrator-led modernization project that takes years to procure.
1Focus group: who and where
The building official or permitting director in a US city of 50,000 to 500,000 whose pre-2010 permits exist only as scans, who fields records requests from appraisers, title companies and state reporting every week while a records-modernization line sits unspent in this fiscal year's budget.
where to find 25 · The Census Bureau's Building Permits Survey list of permit-issuing places, which names every US permit office and its volume; International Code Council regional chapters, where building officials meet monthly; the League of California Cities Annual Conference in October and NAGARA's membership of government records officers, plus the Accela and Tyler Technologies user communities where permitting staff already discuss their systems.
2Sell first, build later
Conversion of up to 10,000 scanned building permits into a machine-readable dataset under a published open schema, 99 percent field accuracy verified on a 200-record audit sample, delivered 30 days after file handoff.
the ask · $4,900 flat per pilot, priced under most municipal small-purchase thresholds; production volume quoted at $0.40 per record.
a real yes · A signed purchase order or a paid P-card invoice counts. Verbal interest, a request to present to the modernization committee, and enthusiasm about the free sample do not.
3Small experiments
The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.
1. Budget-now discovery calls
$300 · 12 days
Pull 25 mid-size cities from the Building Permits Survey list, find the building official through the city site and ICC chapter rosters, and call or email offering to convert their scanned permit backlog at a flat sub-threshold price. One founder runs 25 outreach attempts and books 20-minute calls asking two questions: is there records or modernization money unspent this fiscal year, and what is your small-purchase threshold.
keep going if · 8 of 25 take a call and 5 of those 8 confirm current-year budget and a threshold above $5,000
2. Free 200-permit conversion
$250 · 10 days
File a public records request with two interested cities for 200 scanned permits, then convert them with off-the-shelf OCR plus hand correction into a published permit schema. Send back a clean CSV and the schema doc within 5 days as the accuracy proof.
keep going if · 3 of 5 departments that receive the sample ask for a quote on the full backlog
3. Close two sub-threshold POs
$250 · 20 days
Send each department that saw the sample a one-page fixed-price scope: up to 10,000 permits converted in 30 days for $4,900, 99 percent field accuracy on an audit sample, payable by purchasing card or PO with no RFP. Founder walks the official through signing on a 15-minute call.
keep going if · 2 of 25 departments sign the PO or pay by P-card within 3 weeks of receiving the scope
4Collect a deposit up front
Tesla took $1,000 refundable reservations for the Model 3 and $100 for the Cybertruck before building either: the deposit is the measurement, not the revenue.
$4,900
per prospect, refundable
how · A $4,900 fixed-price pilot charged to the department's purchasing card at kickoff where card use is allowed, otherwise a signed purchase order before any files are converted; municipal buyers cannot wire deposits, but a sub-threshold P-card charge or signed PO is money the department has legally obligated, and the building official can sign it without an RFP.
what it reserves · A slot in the first three-city cohort, a 30-day delivery date, and the department's permit fields written into the first published version of the schema.
refund · If audited accuracy falls below 99 percent, the team corrects the dataset or refunds the full $4,900 within 10 business days.
target · 2 signed POs or P-card payments from 25 departments within 6 weeks
before taking money · Government procurement rules apply, so keep every pilot under the agency's documented small-purchase threshold and touch only public-record document types like building permits, not court, health or juvenile records, until counsel has reviewed data-handling terms.
Go: build it if
2 of 25 departments sign a PO or pay $4,900 within 6 weeks and at least 3 more request formal quotes after seeing the sample.
Kill: stop if
0 payments after 25 departments contacted, or fewer than 3 of 25 confirm any current-year budget, or every interested department routes the $4,900 pilot into a formal RFP anyway.
5 Scripts to run itoutreach message, landing copy, deposit terms · click to open
outreach message
Your department is sitting on decades of scanned permits that staff re-key by hand every time an appraiser, a title company or a state report needs them. I convert scanned permit backlogs into a clean machine-readable dataset under a published schema: 10,000 permits in 30 days for $4,900 flat, under your small-purchase threshold, no RFP. I will convert 200 of your permits free so you can judge accuracy first. Do you have 20 minutes this week?
landing page
Your scanned permit backlog, converted to a clean dataset in 30 days $4,900 flat for up to 10,000 permits, payable by P-card or PO, 99 percent field accuracy verified on an audit sample Send 200 sample permits and get them back converted, free, in 5 days
deposit terms
The $4,900 fixed price covers conversion of up to 10,000 scanned permits into the published schema, delivered within 30 days of file handoff, payable by purchasing card or purchase order. If accuracy on the 200-record audit sample falls below 99 percent, we correct the dataset or refund in full within 10 business days. Signing reserves your city's slot in the first three-city cohort starting October 2026.
Would you run this test?
One tap. The yes-share feeds the Demand pillar of this idea's score; nobody sees who answered.
Budgets are out-of-pocket estimates for a team of one to three, US market. Size the deposit to the deal, and check the terms before taking money in a regulated line.
Scorecard
One score that balances how trendy the idea is, the demand for it and its potential for 100x, with competition measured relative to every other idea in the catalog. Recent startup trends first, government priorities second.
81
Idea Score, 0-100 · raw 50.5 x 1.61
Warm
competition: more crowded than 20% of ideas · headwind x0.90
+4.2
government priorities, secondary (64 matching grants)
Trend
59
Is the wave forming now? 2025-26 entrants vs 2023-24, rounds since 2025, the sector's live-batch direction, the 2026 trend analyst.
- Entrants 2025-26 vs 2023-24 (similar companies)70
- Rounds announced 2025+ in the sector15
- Sector direction (live batch)100
- 2026 trend analyst50
Demand
46
Does anyone want it? YC's current RFS, companies already paid for something similar, the operator judge, founders' yes-rate in decks, readers who would run the test.
- YC asks for it (current RFS: idea / sector)30
- Someone already pays (similar companies, recent / all-time)33
- Operator judge: real pain75
100x potential
50
Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.
- Venture judge50
- Market size axis67
- Moat axis80
- Neighbours still alive6
- Technologist judge50
Score = 100 x cbrt(Trend x Demand x 100x) x (1 - 0.5 x crowding) + government bonus (max 5), calibrated so the 95th-percentile idea scores 90 (order never changes). A geometric mean: a weak pillar cannot be papered over. Percentiles are among the 272 ideas in the catalog; the terms matched were accredited, government, records, standardized, machine-readable, datasets, workforce, sold.
The idea in full
- What
- An agent workforce, sold to government agencies through system integrators, that reads scanned permits, inspection reports, court filings and statistical registers and converts them into certified, AI-ready datasets under a published schema. The deep-tech core is extraction ML tuned to degraded legacy documents and multilingual official formats; the platform layer is the open schema and API that govtech vendors build their own applications on. Security accreditations in each jurisdiction take years to earn and become the license moat that keeps procurement doors closed to newcomers.
- Why now
- Governments are now funding this exact problem: the NSF opportunity 'Unlocking Dataset Value for AI-Enabled Scientific Discovery (AI Datasets)' closes 2026-11-04, NSF's Integrated Data Systems and Services runs to 2027, and NIH's curation-at-scale R01 closes 2027-04-15 - public money is explicitly chasing dataset standardization while agency records remain overwhelmingly unstructured.
- Wedge: first customer and entry point
- Partner with one mid-size system integrator to convert a single record type - building permits - for three municipal governments, and publish the schema so other vendors start consuming the output.
- Path to 100x
- Government data and records modernization is a $10-100B global spend, and the company that owns both the accreditations and the schema becomes the platform every govtech vendor must integrate with - each certified jurisdiction and each vendor building on the API raises switching costs for the next. Agents priced per record processed scale with volume, not headcount, which is what makes the model venture-scale rather than a services firm.
- Ceiling
- Fragmented national procurement could force country-by-country rebuilds that keep the company a collection of mid-size regional contracts rather than one platform.
- Closest real companies, as the generator saw them
- Amorphous standardizes unstructured clinical data for healthcare buyers; Netter positions as an AI-native Palantir for the mid market; Next Data provides an OS for autonomous data products in enterprises. None sells accredited agents to government through integrator channels or publishes a schema others build on.
- Main risk
- Procurement cycles and accreditation timelines could burn the company's capital before the first multi-agency contract lands.
Five judges
Each judge scores every idea in the catalog with a named rubric; the venture judge decides whether a card is shown at all (4-5 is venture-grade).
Venture investor
3/5
Per-record agent pricing scales, but jurisdiction-by-jurisdiction accreditation sold through integrators reads as a chain of regional contracts, not one platform.
Bootstrapper
2/5
Per-jurisdiction security accreditations take years and municipal procurement can outlast the cash a two-person team can raise.
Operator
4/5
Agencies already pay integrators to key in permits by hand, and per-record pricing rides the procurement channel they already buy through.
Technologist
3/5
Degraded multilingual document extraction is real ML, but the durable barrier is per-jurisdiction accreditation rather than anything technically hard to copy.
Risk
4/5
Per-jurisdiction security accreditations are a genuine license moat, and integrator channels plus a published schema spread distribution across many govtech vendors.
trends
3/5
Dated NSF and NIH solicitations earn the secondary grant point, but the core timing - unstructured government records - was equally true in 2023.
Similar startups in the directory
Companies whose pitch matches most of the idea's terms (accredited, government, records, standardized, machine-readable, datasets, workforce, sold): 4 all-time, 1 from the last two years. Same matching as Idea Check.
Speech foundation models that understand people: words, tone of voice, emphasis, intent, and emotion.
The AI workforce for freight dispatch
Operator of a contract workforce platform intended to build the standard for easy contract employment. The company's platform offers employers cover every part of the employment process, from finding talent to screening and interviewing, onboarding, benefits, payroll, human resource support and outplacement, enabling clients and employees with the best contract employment experience.
Big-Data without Data Scientists. Our product is toorPIA that lets anyone analyze big data without help of data scientists.
The generator's reference companies
Real companies the model named as closest when it wrote the card, with their fate. A check mark is a company the radar could verify in its directory.
Public money in this direction
US federal grants, SBIR/STTR awards and open opportunities from the radar's public-money feed, matched to the idea's terms; the sector totals give the context.
8
grants and programs matching the idea
189
startup-relevant grants in Data for AI
$182M
awarded in the sector, tracked
10
opportunities open now in the sector
- Feasibility testing of a novel in-ear sensor system for ambulatory neurological sleep monitoring to enable accessible diagnosis of sleep disordersawardhigh relevance
NIH / NHLBI · SBIR phase I · $314K · posted 2026-09-20
- Novel preservative-free eyedrop with antimicrobial and anti-inflammatory properties for treating Acute Infectious Conjunctivitis and Epidemic KeratoConjuntivitis (EKC)awardhigh relevance
NIH / NEI · SBIR phase I · $307K · posted 2026-09-15
- I-Corps: Translation Potential of a Digital Pet Health Management Software Platformawardhigh relevance
National Science Foundation · I-Corps · $50K · posted 2026-08-10
National Science Foundation · CSCS: Circuits and Systems for · $550K · posted 2026-08-04
- Unlocking Medical AI: A Scalable, Privacy-Preserving Annotation Platform for Clinical and Physiological Dataawardhigh relevance
NIH / NLM · SBIR phase II · $307K · posted 2026-08-01
- SBIR Phase II: AI-Driven Visualization of Rehabilitation Documentation to Support Decision-Making Across Care Settingsawardhigh relevance
National Science Foundation · SBIR Phase II · $1M · posted 2026-07-20
National Science Foundation · SBIR Phase II · $1M · posted 2026-07-16
NIH / NIDCD · SBIR phase I · $306K · posted 2026-07-01
Market signal
What the radar sees in Data for AI: new companies by cohort year, the forming YC batch, and outcomes since the February snapshot.
Data for AI · 50 → 35 → 67 → 62 → 57 new companies 2022 → 2026 · 91% aliveYC F26 live: 8 in this cluster, 6% of the batch (was 4% in S26)Since February, of 167 YC companies here: 1 acquired, 2 shut down, 31 rewrote their pitch
Design attributes
The card is one cell of a designed set: every axis below was chosen before the text was written, and the text had to realize it.
- Buyer
- Government and public sector
- Business model
- AI agent as a service
- Path to 100x
- Platform others build on
- Market size
- $10-100B market
- Capital intensity
- Capital-medium (ops, field teams)
- Speed to revenue
- Revenue in 1-3 years
- Technical depth
- Deep tech: ML, hardware, bio
- Go-to-market
- Partners and channels
- Moat
- License or regulatory moat
- Geography
- Global from day one
- Regulation
- Heavily regulated
- Vibe
- Boring business
Listed under
An idea sits in its own sector and in any sector its text clearly touches.
More ideas like this
AI and software · Vertical AI agents
Bidcraft
Small contractors upload plans and get a priced, win-rate-tuned quote in twenty minutes.
Bidcraft is a self-serve subscription for specialty trade subcontractors and field service firms: drop in a PDF plan set, a site photo set or an inspection report, and the system does takeoff, prices materials from live supplier catalogs in the user's country, applies the firm's labour rates and returns a client-ready quote.
AI and software · AI infra and compute
Gridshift
The marketplace where AI data centers sell their flexibility to the power grid.
Gridshift turns interruptible AI training load into a tradable grid product: data centers enroll clusters, Gridshift's ML forecasts and checkpoint-aware orchestration shift or shed training jobs on minutes of notice, and the flexibility is sold into wholesale electricity markets as demand response.
AI and software · AI infra and compute
Ledgerwatt
Metering and settlement API that turns AI spend into an auditable clearing layer.
Ledgerwatt sits between enterprises and every model provider, neocloud and GPU broker they use, meters each call at the token and job level, prices it against the contracted rate card, and settles a single invoice with per-team, per-product attribution that finance can audit.
AI and software · Vertical AI agents
Coderoot
Self-serve AI agents that maintain legacy enterprise code for a tenth of outsourcing cost.
Coderoot's agent connects to an enterprise's legacy estate - ABAP, COBOL, aging Java, ERP customizations - builds a behavioral map of the system, then works the maintenance backlog: patches, upgrades, regression tests and documentation.
AI and software · Data for AI
Groundfeed
A marketplace where ordinary people sell everyday real-world recordings to AI labs.
A consumer app plus an embeddable SDK that lets anyone record mundane real-world data - chores on camera, receipts, handwriting, phone screen sessions - and sell it into standing dataset orders from model builders.
AI and software · Data for AI
Anvilnet
Capture appliances on factory lines, sold back as an industrial process API.
Anvilnet ships a rugged capture appliance that automation integrators bolt onto existing production and packing lines, tapping machine buses and fixed cameras.
Fictional company written 2026-08-26 from MarkosWeb data; the companies, grants and numbers around it are real and tracked. Treat the idea as a research prompt, not a plan.