New startup ideas · AI and software · Horizontal AI assistants

startup idea

Registrum

Federated agent network for public agencies: every resolved case trains every jurisdiction.

Registrum runs caseworker agents for public agencies (permit review, benefits eligibility triage, procurement document checks) inside a sovereign compute enclave the agency controls.

4/5

venture judge

0

similar startups, last 2 years (5 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

$900 · 6 weeks · 20 prospects

For under $1,000 and six weeks, prove that a hand-run agent can match staff decisions on business permit renewals and that at least two US city permitting directors will pay $4,500 up front for a pilot before any enclave exists.

1Focus group: who and where

Development services director or permit manager at a US city of 100,000-500,000 population who runs business permit and license renewals, has 1-3 vacant counter positions, and is measuring a renewal backlog in weeks this quarter

where to find 20 · LinkedIn Sales Navigator (title 'Development Services Director' OR 'Permit Manager' OR 'Licensing Manager', industry Government Administration, US), the ELGL (Engaging Local Government Leaders) community, and GovLoop; build the city list from population data and each city's published permit fee schedule

2Sell first, build later

A 90-day concierge pilot on business permit renewals only: up to 300 cases decided with rulebook citations, each returned within 48 hours, run by hand behind a simple portal, starting within 30 days of signature; no enclave, no integration, files exchanged through the city's existing secure channel

the ask · $4,500 per pilot, invoiced 100% up front (equivalent to $15 per case at full volume)

a real yes · A real yes is a paid invoice or an issued purchase order with a start date and a named staff contact for case handoff; not a yes: 'the innovation team loves this', an unpaid sandbox, or an invitation to respond to a future RFP

3Small experiments

The first one attacks the riskiest assumption; each ends with a number that says whether to run the next.

  1. 1. Fifteen permitting director calls

    $150 · 12 days

    Book 15 calls with directors sourced from Sales Navigator and ELGL using a records-request-style opener about renewal volumes. Ask three numbers on each call: renewals per year, days per decision, and cost per case; then ask who signs a purchase under the city's small-purchase threshold.

    keep going if · 8 of 15 quantify a backlog of 2 weeks or more and put staff cost above $40 per renewal decision; 6 name the person who can approve a sub-$5,000 purchase without an RFP

  2. 2. Concierge renewal decision run

    $300 · 14 days

    Obtain 50 recent business license renewal case files from 2 cities (renewal records are public; request redacted copies or file a records request). Run each decision by hand against the city's published rulebook with an LLM assist, and deliver each director a memo with your 50 decisions, citations, and a match rate against what their staff actually decided.

    keep going if · Your decisions match staff outcomes on at least 45 of 50 cases, and both directors ask what a pilot would cost

  3. 3. Paid pilot one-pager

    $100 · 21 days

    Email a one-page offer to 25 directors: 90 days, business permit renewals only, up to 300 cases decided with rulebook citations within 48 hours each, fixed at $4,500 so it fits under most cities' small-purchase or p-card threshold. Follow up by phone within 3 business days.

    keep going if · 4 directors forward it to their purchasing officer with a question about invoicing, and 2 sign with a start date

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.

$4,500

per prospect, refundable

how · Paid pilot invoiced 100% up front at $4,500, deliberately priced under the $5,000-10,000 small-purchase threshold most US cities allow on a p-card or direct PO without competitive bidding; signed by the department director and cleared with the purchasing officer on the same email thread set up: Stripe Invoicing

what it reserves · One of the first 3 pilot slots, the city's rulebook encoded first, and a locked $15-per-case price for year one

refund · Fully refundable if the decision match rate on the first 100 cases falls below 90% against staff review

target · 2 paid pilots from 20 director conversations within 45 days

Go: build it if

2 pilots paid or on issued POs, concierge match rate at 90% or better on 50 cases, and at least 5 more directors on a second call; build the thin enclave story only after that

Kill: stop if

Zero payments or POs after 45 days, fewer than 5 of 20 directors take a second call, or every interested city routes the $4,500 offer into a 6-month RFP anyway

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.

47

Idea Score, 0-100 · raw 28.8 x 1.64

Open

competition: more crowded than 6% of ideas · headwind x0.97

+3.6

government priorities, secondary (33 matching grants)

Trend

7

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)3
  • Rounds announced 2025+ in the sector0
  • Sector direction (live batch)0
  • 2026 trend analyst25

Demand

32

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)42
  • Operator judge: real pain25

100x potential

79

Can it return a fund? The venture judge (double weight), market-size and moat axes, neighbours still alive, the technologist judge.

  • Venture judge75
  • Market size axis67
  • Moat axis100
  • Neighbours still alive81
  • Technologist judge75

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 federated, network, public, agencies, resolved, case, caseworker, permit.

The idea in full

What
Registrum runs caseworker agents for public agencies (permit review, benefits eligibility triage, procurement document checks) inside a sovereign compute enclave the agency controls. Case data never leaves the jurisdiction; what crosses the network is validated procedure patterns and outcome signals, so a permit decision pattern proven in one city improves the agent in the next. Agencies pay per case handled, not per seat.
Why now
Three NSF streams in this cluster fund exactly the missing pieces: 'A New Incentive Design Theory for Learning Agents in Federated Population Games' at ~$530k, 'Collaborative Research: RI: Conditioning Uncertainty on Epistemic Context in Human-AI Communication' at ~$941k, and 'SBIR Phase I: Agentic Artificial Intelligence System' at ~$305k. Of 282 tracked companies in this cluster, the recent cohort is entirely consumer, SMB and enterprise assistants, with no one aimed at public casework.
Wedge: first customer and entry point
One mid-size national agency plus one municipal permits office, single case type: business permit renewals, where the rulebook is public and the outcome is checkable.
Path to 100x
Global government administrative case processing and BPO is a $10-100B spend, and per-case pricing captures the labour line directly rather than a software budget. The network effect is the whole company: each jurisdiction's validated procedures raise accuracy for every other, so the agency that joins the largest network gets the best agent, and the second network cannot catch up on case volume.
Ceiling
If regulators forbid any cross-border learning signal at all, Registrum degrades into a per-country systems integrator with no compounding advantage.
Closest real companies, as the generator saw them
Almanac ('The AI that knows your company') and Dock build shared corporate context but assume one organisation and commercial data handling. Tasklet has agents owning work with no enclave, no jurisdictional isolation and no cross-tenant learning protocol.
Main risk
Public procurement cycles run longer than the R&D runway, and the first three agencies each demand a bespoke enclave that never amortises.

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

    4/5

    Per-case pricing on a $10-100B government BPO line with cross-jurisdiction learning is genuine compounding defensibility, though procurement cycles may outrun the runway.

  • Bootstrapper

    1/5

    Sovereign enclaves plus public procurement cycles longer than the runway means years of spend before the first agency pays.

  • Operator

    2/5

    Public procurement cycles outrun the runway and the value only appears once many jurisdictions join, which is the whole industry moving first.

  • Technologist

    4/5

    Cross-jurisdiction procedure transfer inside sovereign enclaves is genuinely hard federated engineering whose accuracy compounds per case, though the learning signal is legally fragile.

  • Risk

    2/5

    Cross-border learning signal is the entire moat and any regulator forbidding it collapses the company into per-country integration work.

  • trends

    2/5

    The why-now is three NSF awards, which the rubric caps at one secondary point; federated public casework itself rests on no dated 2025-2026 shift.

Similar startups in the directory

Companies whose pitch matches most of the idea's terms (federated, network, public, agencies, resolved, case, caseworker, permit): 5 all-time, 0 from the last two years. Same matching as Idea Check.

  • Arctasef EF New York 2024 · 2024 · Commerce and marketplacesalive

    Government contracting automation for small businesses.

  • Heymirza500global 500G GA 29 · 2021 · Fintechalive

    Developer of a fintech platform designed to deliver tax-efficient employer stipends to a network of caregivers. The company's tools integrate family financial planning and childcare funding, and also provide assistance in corporate policies for parental leave, fertility options, and modern parenting advice for employees, helping working professionals to work towards achieving their financial goals.

  • Untapef EF London 2021 · 2021alive

    Wastewater pathogen monitoring for public health.

  • Markerly500global 500G GA 5 · 2012 · Commerce and marketplacesalive

    Operator of an influencer marketing platform designed to connect brands with people. The company's platform helps to build a marketing network with full campaign management capabilities and influencer CRM (Customer Relationship Management) tools, enabling clients to create authentic experiences and persuasive repurposable content.

  • neeboorplugandplay · Consumeralive

    neeboor is an open and casual location based social app lets your living area more hackable✌️

Run this as an Idea Check →

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

15

startup-relevant grants in Horizontal AI assistants

$30M

awarded in the sector, tracked

All public money by sector →

Market signal

What the radar sees in Horizontal AI assistants: new companies by cohort year, the forming YC batch, and outcomes since the February snapshot.

Horizontal AI assistants · 30 → 77 → 84 → 57 → 28 new companies 2022 → 2026 · 89% aliveYC S26: 7 in this cluster, 3% of the batch (was 3% in X26) (F26 is still forming: 21 listed)Since February, of 129 YC companies here: 3 acquired, 3 shut down, 30 rewrote their pitch

Horizontal AI assistants: companies, trend and grants →

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
Network effects, winner takes most
Market size
$10-100B market
Capital intensity
Capital-heavy (hardware, bio, infra)
Speed to revenue
R&D first, revenue after 3 years
Technical depth
Deep tech: ML, hardware, bio
Go-to-market
Founder-led sales
Moat
Network effects
Geography
Global from day one
Regulation
Some regulation
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 · Horizontal AI assistants

Subroclaim

Agents that negotiate insurer-to-insurer claim recoveries on both sides of the file.

Subroclaim runs subrogation and inter-carrier recovery as an agent service for US property and casualty insurers: the agent reads the loss file, builds the demand with liability evidence, and then negotiates directly with the other carrier's Subroclaim agent, escalating to a human adjuster only on disputed liability or amounts above a set threshold.

Score 71Warm competitionVC 4/5AI agent as a serviceEnterprisetest: $2.6k · 6w78% of 3 neighbours alive

AI and software · Horizontal AI assistants

Guildhall

The exchange where enterprises hire proven AI agents like contractors.

Guildhall is a self-serve marketplace where enterprises browse, trial, and deploy third-party AI agents - sales research agents, ops agents, finance-close agents - each listed with standardized, independently run performance telemetry instead of vendor claims.

Score 59Warm competitionVC 4/5MarketplaceEnterprise95% of 4 neighbours alive

AI and software · Horizontal AI assistants

Piecework

A marketplace where enterprises hire third-party AI agents and pay per verified task.

Independent builders list agents with a declared task type, a price per completed unit and a verification rule; enterprise teams connect their systems through no-code connectors and buy agent labour without a procurement cycle.

Score 56Active competitionVC 5/5MarketplaceEnterprisetest: $1.5k · 5w95% of 3 neighbours alive

AI and software · Horizontal AI assistants

Procura

API that gives any AI agent legally binding authority to act for a small business.

Procura is the mandate layer agent builders call before their agent signs, pays or files anything: the small business owner verifies identity once, grants a scoped, revocable power of attorney, and Procura issues a cryptographic credential that binds each agent action to that mandate with a qualified electronic seal and an immutable audit trail.

Score 53Open competitionVC 4/5Infrastructure and APIsSmall businesstest: $1.7k · 4w78% of 3 neighbours alive

AI and software · Horizontal AI assistants

Bursar

A personal agent with a license to move your money.

Bursar is a consumer financial agent that does not just read your accounts - it pays bills, disputes charges, cancels subscriptions, negotiates rates, and switches providers, because the company holds money transmitter licenses and runs its own payment rails.

Score 45Crowded competitionVC 4/5AI agent as a serviceConsumer95% of 4 neighbours alive

AI and software · Horizontal AI assistants

Backline

Flat-rate AI back office for small businesses, running on our own GPUs.

Backline sells small businesses a set of working agents - phone answering, scheduling, invoicing, collections, vendor follow-up - for one flat monthly price with a self-serve signup that takes fifteen minutes.

Score 40Crowded competitionVC 4/5AI agent as a serviceSmall business95% of 4 neighbours alive
Swipe ideas like this in the deckTalk to the radar about it

Fictional company written 2026-08-22 from MarkosWeb data; the companies, grants and numbers around it are real and tracked. Treat the idea as a research prompt, not a plan.