StateLine — AI-hiring notice router for US mid-market HR

1. One-liner

Detects each applicant’s state, drops the legally-correct AI-use notice into the ATS funnel, builds the audit packet.

2. Trend signal — why now?

Three state laws turned on in five months. Mid-market HR can’t keep up.

• Illinois HB 3773 — effective Jan 1, 2026. Amends the Illinois Human Rights Act. Any employer with one or more employees in IL for 20+ calendar weeks must notify applicants and employees whenever AI is used in any employment decision — recruitment, screening, promotion, discharge, training selection. Notices must be in plain language, in languages commonly spoken in the workforce, accessible to employees with disabilities, preserved four years. IDHR draft rules circulated late 2025; formal rulemaking opened Q1 2026. (HB 3773 effective Jan 1, 2026 — Hinshaw; Draft notice rules — Epstein Becker Green)

• Colorado AI Act (SB 24-205) — effective Feb 1, 2026 (employment provisions phase-in to Jun 30, 2026). Employers using “high-risk” AI in any consequential employment decision are “deployers” — must run annual impact assessments, post a public website disclosure, maintain an NIST-AI-RMF / ISO-42001-aligned risk management program, give candidates pre-decision notice with the right to human review and to appeal an adverse decision. (Colorado AIA — Perkins Coie)

• New Jersey N.J.A.C. 13:16 — adopted Dec 15, 2025. Broader than NYC LL 144 — covers “any term, condition, or privilege of employment,” not just hiring. Disparate-impact-by-AEDT counts even with no discriminatory intent, even when relying on third-party developer. Division on Civil Rights enforces with fines plus injunctive relief. (AG Platkin announcement, Dec 2025)

• NYC Local Law 144 — stricter 2026 enforcement. Dec 2025 NY State Comptroller audit shredded DCWP: 75% of 311 calls misrouted, agency caught 1/32 violators while auditors flagged 17+. Major law firms (DLA Piper, Akerman) warning of a step-change in 2026 enforcement, $500–$1,500 per violation per day per use. (NYS Comptroller audit)

• California AB 2930 dead, but FEHC AI regs took effect Oct 1, 2025 with notice and recordkeeping carve-outs effectively similar to Illinois. A reintroduced AB-2930-like bill is expected in the 2026 session per K&L Gates.

• Federal vacuum. Trump’s Dec 11, 2025 executive order attempting to preempt state AI laws hasn’t been litigated; until DOJ or courts act, state laws bind. (HR Morning — Trump EO on state AI laws)

• SHRM State of AI in HR 2026: 67% of HR leaders cite “lack of awareness of AI capabilities” as biggest blocker, 57% of HR pros in AI-regulated states aren’t aware of the policies governing AI in hiring. Mid-size org defined as 100–499 employees. (SHRM State of AI in HR 2026)

• PwC reading: Only 24% of enterprises using AI in HR processes have started formal compliance prep, despite the cliff. AI is in 87% of recruitment funnels. (SQ Magazine — PwC stats)

• 259 active AI hiring bills across 29 states per AI Laws by State, May 2026. (AI Laws by State compliance map)

Provenance:

• Signal 1 (demand): SHRM 2026 — 57% of HR pros in AI-regulated US states unaware of governing policies; 67% blocked by AI awareness gap — https://www.shrm.org/topics-tools/research/state-of-ai-hr-2026 — 2026-01
• Signal 2 (feasibility): LLMs now competent at jurisdiction-detection, plain-language notice generation, multilingual rendering, log retention; Anthropic/OpenAI APIs sub-$0.01/call at SMB volume; ATS webhook ecosystems (Greenhouse, Lever, Ashby, Workable, Workday Recruiting) all expose candidate-creation hooks — vendor docs, 2026
• Signal 3 (economic): Warden AI, Holistic AI, FairNow, Optro (acquired FairNow), BABL AI all funded; bias audits priced $5K–$50K/year per system, all enterprise sales-led; SHRM mid-market category 100–499 employees has no priced product — https://www.warden-ai.com/, https://optro.ai/product/ai-governance, 2026-Q1
• Category: Regulatory arbitrage

3. The opportunity

The bias-audit incumbents (Warden, Holistic, BABL, FairNow/Optro, ComplianceHR) sell upmarket. Their selling motion is sales-led demos to enterprise CHROs at $20K–$150K/year. They audit AI vendor models (Findem, HireVue, Pymetrics, Workday HiredScore) — i.e., they sell to the HR tech vendors and the Fortune 1000.

The mid-market — 100–499-employee employers with one HR/People Ops lead, using a stack of Greenhouse/Lever/Ashby + LinkedIn Recruiter + some flavor of ChatGPT for resume screening — gets nothing. They have:

• No legal/AI compliance counsel in-house
• No budget for a $30K bias audit
• A Workday/Greenhouse ATS that exposes its own AI features but ships no per-state legal artifact
• One IL employee or one NJ recruiter triggering Illinois HB 3773 / NJ AEDT obligations
• A January 2026 (IL), February 2026 (CO), and ongoing-2026 (NJ + NYC) compliance cliff they’re already past

What they need isn’t a $30K bias audit. They need a thin layer between the ATS and the candidate that:

1. Tags each applicant with the right state(s) — based on residence, the job’s hiring location, and the recruiter’s state.
2. Sends the legally-correct notice — Illinois plain-language + workforce-language render, NYC 10-business-day AEDT notice, Colorado pre-decision deployer disclosure, NJ AEDT notice, in the appropriate channel (career-site banner, application form field, post-application email).
3. Captures the alternative-process / human-review opt-out where the law requires it (NYC, CO).
4. Retains every notice + opt-out for 4 years (IL recordkeeping) with WORM-style audit logs.
5. Generates the annual bias-audit packet for NYC, the impact assessment for CO, and the AEDT log for NJ — pulling EEO-1-style data from the ATS, structuring it the way each state wants.

Sits next to the ATS, not inside it. Doesn’t ask the HR person to learn the law — it routes them to the right artifact.

4. Target market

• Primary customer: US-headquartered companies with 100–499 employees, at least one Illinois, Colorado, NJ, NYC, or California-based employee or open req; HR/People Ops lead is the buyer. Vertical-agnostic, but skewed toward services/tech/healthcare-non-clinical/manufacturing — i.e., industries that do continuous mid-volume hiring (20–200 hires/year).

• Why they buy: They already use AI somewhere in hiring (resume screen, structured interview transcription, AI sourcing, AI-assisted job description). The new state laws don’t care whether the AI is the Workday model or the recruiter pasting candidates into ChatGPT — both trigger notice and recordkeeping. The HR lead reads one of the 50 law-firm alerts about Illinois HB 3773 and panics. Their options are: hire Littler ($800/hr), buy ComplianceHR (sales call, weeks to onboard), or do nothing and hope. StateLine is the third option — fastest, cheapest, paid this week.

• Rough TAM reasoning: US has roughly 200,000 employers with 100–499 employees (Census County Business Patterns; SHRM mid-size band). Trim to those with at least one employee in IL/CO/NJ/NY/CA — those five states hold ~30% of US payroll, so ~150,000 employers in the addressable band. Trim again to those using AI in hiring (SHRM: 27% of HR uses AI for recruiting, 33% use AI to screen resumes — call it 35% of mid-market) → ~50,000 mid-market employers in immediate need. At even 1% penetration × $1,800/year ACV = $900K ARR; 5% × $1,800 = $4.5M ARR. Plenty of room.

• Why now for them: The Illinois law went hot Jan 1, 2026. NJ regs went hot Dec 15, 2025. Colorado is Feb 1 / June 30, 2026. The first IDHR complaints will land in May–July 2026 — public, named in industry press. Every HR lead who reads that headline is a buyer the next morning.

5. Product sketch (MVP)

• Jurisdiction tagger — at application time, infers the state(s) that govern this candidate (applicant residence, job posting state, recruiter state, plus any “remote anywhere” flag). Outputs a per-candidate compliance profile.
• Notice library — pre-written notice text for IL, CO, NJ, NYC, CA, with a lawyer-reviewed update SLA whenever a final rule drops. Auto-translates into the top 6 workforce languages (Spanish, Tagalog, Vietnamese, Polish, Mandarin, Arabic).
• ATS drop-in — Greenhouse, Lever, Ashby, Workable, Workday Recruiting integrations: injects the notice as an application-form field, an automated post-application email, or a career-site banner; captures the candidate’s acknowledgment + alternative-process opt-out where the law requires.
• AI-use registry — HR person logs which AI tools they use in which stage (resume screen, video interview score, sourcing). Maps each to the state’s “AI” definition (IL is broad; NYC AEDT is narrower). Drives whether notice fires.
• Audit-packet generator — annual packet for NYC (bias audit-ready data export with selection/scoring rates + impact ratios), CO (impact assessment template populated with deployer artifacts), IL (4-year retention archive, employee-notice posting log), NJ (AEDT inventory + use log).
• State-rule monitor — feed of bill movements + final rules across the 29 active states; alerts when a new obligation triggers for the customer’s profile.
• Adverse-decision human-review path — for CO and similar, structured workflow that captures the human reviewer’s identity, decision timestamp, and rationale on every AI-influenced adverse decision.

6. AI angle — what’s load-bearing

Three places AI does real work, not decoration:

1. Jurisdiction-tagging. LLM reads the application context (candidate-supplied address, IP geolocation hint, job posting location, “remote/anywhere” toggle, recruiter location) and outputs a structured “applicable laws” set. A rules engine could do this naively but stumbles on the messy real-world inputs — “Brooklyn, NY” vs “New York” vs “remote, will work US-EST” — and on overlapping jurisdictions where a Colorado-resident applying to an Illinois-posted remote job triggers both. LLM extraction + lightweight verification is the right tool.

2. Multilingual plain-language notice generation. Illinois requires “plain language” and “languages commonly spoken in the workforce.” A static template-per-language doesn’t survive translation review. LLM drafts each notice in candidate’s preferred language, runs back-translation against a lawyer-locked English source, flags discrepancies for the customer’s HR person to approve before send. Once approved, cached.

3. Adverse-impact analysis on the customer’s funnel. Pulls candidate flow + protected-class data from the ATS at year-end, computes selection rates by AI-influenced stage, surfaces likely Title VII / IL Human Rights Act / CO Civil Rights flags. The math is regression-grade; the reading of the results in customer-friendly natural language is LLM-grade.

Strip AI out and you have a CMS with hand-coded forms. The AI is what lets a 1-person HR team handle five states without learning five legal frameworks.

7. Localization angle (if any)

US-only by design — every state is a separate localization within the same product. The “localization” here is per-state notice text, per-state workflow, per-state recordkeeping. Bonus expansion path is EU AI Act FRIA for European subsidiaries of US mid-market customers (deferred to Dec 2027, so v2 problem). Not an India/SEA/LatAm play.

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

• Pricing:

  • Starter — $149/mo — 1 state, up to 50 hires/year, 1 ATS integration, basic audit packet.
  • Growth — $399/mo — up to 5 states, 200 hires/year, all major ATS, state-rule monitor, bias-data export.
  • Scale — $899/mo — unlimited states/hires, multi-entity, SSO/SOC 2, dedicated audit-packet generation, named compliance reviewer (lawyer in the loop).
  • Annual contracts default; lawyer review add-on at $99/notice for one-off custom legal language gets paid on top.

• ACV: Blend lands ~$4,800/year. Starter is the wedge ($1,800), but Illinois-only customers quickly add CO and NJ when they see how cheap the upgrade is.

• $1M ARR math: ~210 Growth customers, or 555 Starters, or 110 Scale. Realistic at 100–499-employee band given the urgency.

• $5M ARR math: ~1,050 Growth-equivalent. Roughly 2% of the 50,000-employer addressable set — credible for a focused mid-market sales motion over 24 months.

• Expansion path: start with state-router + notice; layer adverse-action workflow (Colorado, NJ); layer year-end audit packet ($499 one-time add-on at first, packaged into Scale later); layer EU FRIA module when EU deadline lands; layer non-employment use cases (tenant screening, consumer credit AI notice).

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

1. The “Illinois Cliff” outbound list. Build a list of 5,000 IL-headquartered mid-market employers (100–499 employees) from LinkedIn Sales Nav + Illinois Secretary of State. Cross-reference with public ATS evidence — Greenhouse/Lever apply-with URLs in job posts via Google dorking. Send a one-page Loom from the founder titled “Your Illinois HB 3773 notice is missing — here’s what IDHR will look for in your first audit,” with a 2-minute fix link. Expect 3–5% reply, 30% of those convert to demo, 30% of demos close. ~15 customers off list 1.

2. HR-tech partner channel. Co-sell with the non-ATS HR vendors: LinkedIn Recruiter resellers, AppCast/Recruitics (programmatic job ads), assessment vendors (Wonderlic, eSkill). They have lists of mid-market HR leads and no compliance product to bundle. Rev-share 20% recurring. Target 3 channel deals → 50 customers in months 3–6.

3. SHRM chapter and HR-influencer SEO. SHRM has 700 local chapters; ~140 in IL/CO/NJ/NY/CA. Sponsor monthly chapter lunches at $1K each. Recruit 5 mid-market HR influencers (the ones who run a 20K-follower LinkedIn newsletter on “HR compliance”) for paid review + affiliate link. Founder-led webinar series “30 minutes to your Illinois AI-hiring notice” — gates the lead via the audit packet generator. 30–50 customers from this in months 4–9.

4. Inbound from law firm referrals. The 50 employment-law boutiques (DLA Piper, Akerman, Hinshaw, Epstein Becker, Fisher Phillips) that wrote alerts about IL/CO/NJ regulations all face the same client question — “what tool do I recommend?” — and have nothing to point at. Free counsel-tier (full feature set for the firm’s own 50-employee in-house counsel use) in exchange for being the listed referral on their alert page. 10–20 customers from this in months 6–9.

5. Cold paid-search on the panic queries. “Illinois HB 3773 notice template,” “NYC AEDT bias audit how to,” “Colorado AI Act employer compliance.” Bid hard, land on a free Illinois notice generator that captures email and ATS, upsells to paid. Low monthly spend, high intent. ~$500/customer CAC at first.

10. Build complexity — justification

Low. Off-the-shelf AI (Anthropic + OpenAI for the LLM work), standard Postgres + S3 + Next.js, well-documented ATS webhook integrations (Greenhouse, Lever, Ashby, Workable, Workday Recruiting), no custom infra. The hard work is content + legal review, not engineering — keeping the 5-state notice library current, getting a lawyer to bless the multilingual renders, monitoring 29 states’ rulemaking. v1 in 10–14 weeks with two engineers + a part-time employment lawyer on retainer. Mid-market sales motion is the longest pole.

11. Gating checklist

Gate	Pass?	Note
Legal in target market	✅	Helps employers comply with state law; not subject to bar admission rules because we ship templates + workflow, not legal advice.
Ethical — no harm / dark patterns	✅	Pro-candidate (more notice, clearer opt-out paths) and pro-compliance for employers.
Market exists (evidence above)	✅	Three new state laws Jan–Feb 2026, $5K–$50K incumbent pricing, 76% awareness/adoption gap per PwC.
1–5 person team can build this	✅	Two engineers + part-time lawyer in 10–14 weeks.
Launchable with <$50K / ₹40L	✅	$30–45K to v1 incl. lawyer retainer for content review.

All five pass.

12. Feasibility score

Axis	Weight	Score	Notes
Problem intensity	20	16/20	Hits HR weekly (every req, every hire). Penalties material ($500–$1,500/violation/day NYC, IL IDHR civil penalties). Pain felt now — IL law went live 5 months ago. Not “hair on fire” for everyone yet because enforcement still ramping, hence 16 not 19.
Demand evidence	15	12/15	SHRM 2026 quantified gap (57% awareness, 67% blocker, 24% prep’d), Warden/Holistic/FairNow/Optro all funded chasing enterprise tier, 259 active state bills. Mid-market voice underdocumented because they don’t have communities the same way — minus a point.
Build feasibility	15	13/15	LLM tasks well within current capabilities, ATS APIs mature, web stack standard. Content velocity (keeping 29 states current) and lawyer-in-the-loop SLA are the only friction.
Distribution clarity	15	11/15	Outbound to IL/CO/NJ mid-market viable, channel partners plausible, law-firm referrals real, paid-search on panic queries known to convert. Lose a point because mid-market HR is famously slow to switch tools and has 90-day procurement; lose another because LinkedIn outreach saturation makes founder cold email harder than it was three years ago.
Revenue mechanics	15	12/15	Pricing benchmarks against Mineral ($79–149) and ComplianceHR (enterprise); $4.8K ACV credible. 50K addressable shop, 2% penetration = $5M. Churn risk if Trump EO succeeds + states get preempted.
Time to first revenue	10	8/10	Founder pre-sells before code ships — Illinois urgency is the wedge. First paying customer week 6–8 with a credible pitch.
Defensibility	10	7/10	Soft moat: state-by-state legal content library + ATS integration depth + audit-log accumulation. Workflow lock-in is real once year-1 packets archive. No proprietary data; the rule-text is public. A copycat in month 12 has to redo lawyer review for every state, which slows them.
Total	100	79/100	GO.

13. Qualitative modifiers

Founder-fit tags

technical-heavy · domain-expertise-required — needs someone fluent in employment law (or comfortable building a lawyer-on-retainer relationship) plus a strong product engineer. Sales-heavy in year 2.

Key assumptions to validate (3–5)

1. Mid-market HR will pay $149–899/mo for this without enterprise procurement. How to test: 30 LinkedIn DMs to 100–499-employee HR leaders in IL who posted a job in the last 30 days; ask “would you pay $X/mo to fix this in a week?” — pre-sell intent letters at the Growth tier; target 8/30 yes.
2. Trump EO doesn’t kill state-level enforcement. How to test: Watch DOJ filings + 9th-Circuit/2nd-Circuit briefs through Q3 2026. Set up a Google Alert + Bloomberg Law watch. Kill criterion in section 14 below.
3. ATS vendors won’t ship this themselves in 12 months. How to test: Map roadmap announcements from Greenhouse, Lever, Ashby, Workable through their public changelogs + customer-advisory-board minutes leaked to G2; verify none has shipped per-state notice routing by Sep 2026.
4. Annual audit-packet generation is a $499–$2K upsell, not a free table-stake. How to test: Price-anchor in the first 20 closed-won deals; measure attach rate of audit-packet add-on at three price points.
5. Lawyer-in-the-loop SLA can be priced at margin. How to test: Negotiate retainer with two employment-law boutiques at the 50-state level for under $4K/mo; back-solve customer pricing against that floor.

Risk flags

1. Regulatory whiplash. Trump’s Dec 11, 2025 EO directs federal agencies to seek preemption of state AI laws. If DOJ wins fast, demand collapses. Mitigant: focus first on IL HB 3773 (Human Rights Act amendment — civil-rights statute base, harder to federally preempt than a stand-alone AI statute) and on EU AI Act FRIA expansion as a second leg.
2. Platform dependency on ATS APIs. If Greenhouse or Workday acquires a competitor and locks down webhooks, integration cost spikes. Mitigant: ship a “no-ATS-integration” mode (career-site iframe + email-only), so we’re never blocked by ATS.
3. Market timing — federal preemption + bias-audit incumbents move down-market. If Holistic/Warden/FairNow pivot to a self-serve mid-market tier in 2026, the wedge narrows. Mitigant: speed and price — Warden’s enterprise sales cycle is 4–6 months; we close in 2 weeks.
4. Content-update load. 29 states with active bills + 50 federal/agency moves = real ongoing burden. Mitigant: structured rulemaking-feed scraper + lawyer-of-record retainer with weekly review cadence; this becomes the moat over time.

14. Structured verdict

Score:                  79/100
Verdict:                GO
Confidence:             Medium
Best-fit builder:       Technical founder with employment-law network or co-founder; 100–499 mid-market HR-sales chops second
Time to revenue:        6–10 weeks (pre-sell before MVP ships)
Capital to launch:      $30–45K ($4–6K Anthropic/OpenAI API runway, $20K lawyer retainer for content, $10K product + cloud)
Top 3 assumptions to validate first:
  1. Mid-market HR (100–499 employees) will pay $149–899/mo on a credit card — 30 IL HR DMs, target 8/30 pre-sell letters
  2. Trump EO doesn't preempt state law fast enough to kill demand — track DOJ filings + 2nd/9th Circuit through Q3 2026
  3. ATS incumbents don't ship per-state notice routing in 12 months — public-roadmap audit Sep 2026
Kill criteria:
  - Abandon if a federal court permanently enjoins Illinois HB 3773 + Colorado AIA before Q4 2026
  - Abandon if Greenhouse, Lever, or Workday ships native per-state notice routing in Q3 2026 before our v1 hits 50 customers
  - Abandon if <15 of first 100 cold outreaches book a demo by week 8

15. Next step — 1-week validation sprint

• Day 1–2: Pull 200 IL-headquartered 100–499-employee employers from LinkedIn Sales Nav. Cross-reference for an active Greenhouse or Lever job post. Manually scrape the application form to verify no Illinois HB 3773 notice is present today. Document hit rate.
• Day 3: Draft 4-variant founder Loom + 1-sentence email: “Your Greenhouse application is missing the HB 3773 notice — IDHR can fine $X — 2-minute fix here.” A/B subject lines.
• Day 4: Send 100. Track open + reply rate; book demo calls for Day 5–7.
• Day 5–7: Run 5–10 demo calls. Pitch the $399/mo Growth tier. Ask for credit-card-on-file LOI for an Aug 2026 v1 launch.

Go / no-go bar: 8/100 cold outreaches reply, 4 demos booked, 2 LOIs with credit card hold by day 7. Anything less = the wedge isn’t sharp enough; rework or PASS.