Small, AI-Native — 17 Builders Worth Watching

Click any row to jump to its full entry.
	Company	The wedge	Makes money by	Repeatability	Signal
01	GreptileSan Francisco, CA	AI reviews pull requests with full repo context, not just the diff	Seat/usage SaaS to eng teams · ~$1M rev	Medium barrier	Proven
02	Marr LabsSan Francisco, CA	AI voice agents that make outbound calls for lenders/collections	Presumably per-call/volume pricing · no figure public	High barrier	Early
03	VapiSan Francisco, CA	Developer infrastructure for building voice AI agents	Usage-based API pricing · ~$8M ARR	Medium barrier	Proven
04	11x.aiSan Francisco, CA	AI 'digital workers' that do outbound sales prospecting instead of a human SDR	Seat/outcome SaaS to sales orgs · ~$350M valuation	Low barrier	Contested
05	SazabiSan Francisco, CA	Observability built for how AI agents fail, not how normal software fails	Presumably usage-based infra pricing · pre-revenue	Medium barrier	Early
06	Cal AIIndependent (not YC)	Photo of your food in, calorie/macro estimate out	Consumer subscription · $40M+ ARR	Low barrier	Proven
07	Lingo.devSan Francisco, CA	AI translates an app's UI text and opens the pull request itself	Usage-based dev SaaS · ~$1.1M ARR at 7 people	Medium barrier	Proven
08	winfuncSan Francisco, CA	AI that finds a security bug, then proves it's real by actually exploiting it	Presumably usage/engagement pricing; no round beyond YC seed disclosed	High barrier	Early
09	TruewindSan Francisco, CA	AI does the bookkeeping and month-end close instead of a human accountant	Subscription SaaS to startups · undercuts outsourced bookkeeping cost	Medium barrier	Proven
10	OperandSan Francisco, CA	Not software for lawyers — an AI-run law firm that collects the legal fee itself	Collects legal fees directly, not a software license	High barrier	Early
11	LunaBillUnited States	AI automates medical billing and claims coding for healthcare providers	Presumably transaction/claims-volume pricing; not yet public	Medium barrier	Early
12	TomaSan Francisco, CA	AI voice agents that book service appointments and answer sales calls for car dealerships	Presumably per-dealership SaaS · YC says “millions” in revenue	High barrier	Proven
13	LedgerUpSan Francisco, CA	AI agent that runs contract-to-cash billing for B2B SaaS with usage-based pricing	Subscription to SaaS finance teams · ~$750K rev (single, low-confidence source)	Medium barrier	Early
14	HarperSan Francisco, CA	AI-native commercial insurance brokerage that keeps the broker commission itself	Broker commissions, not a license · $6M+ annualized premiums	High barrier	Proven
15	BraviParis, France	AI voice/chat agents that answer phones for home-service installers (shutters, HVAC, solar)	Presumably per-seat/lead subscription; funding not public	High barrier	Early
16	PrismLondon, UK	Not software for recruiters — an AI-run recruiting agency that delivers ready-to-interview candidates	Presumably placement/retainer fees, not a license · $3.3M raised	High barrier	Early
17	BrickantaStockholm, Sweden	AI reads construction bid documents and flags missing items and compliance risks	Presumably per-project/seat SaaS to construction firms · $8M raised	Medium barrier	Early

01

Greptile

Medium barrier Proven YC W2024 · San Francisco, CA · Developer tools

AI reviews every pull request with full memory of the codebase — not just the diff.

Eng team $/seat Greptile delivers AI PR review, repo-aware

~$1M revenue (independent estimate) · $30M raised

Real, dated data · Cumulative funding raised

$4.1M Jun 2024

$30M ~Oct 2025

First customers

confirmedFirst paying customers came before any funding — likely via HN and dev-Twitter, off the founders' Georgia Tech/YC network.

Growing via

guessWord of mouth among engineers; the Series A likely funds a push into larger orgs on top of that.

Copy this?

Medium barrier. Copilot, Cursor Bugbot, and CodeRabbit chase the same wedge — the lead is accumulated repo context, not a moat.

Daksh Gupta, Soohoon Choi, Vaishant Kameswaran · $30M total (Benchmark-led Series A), ~$180M valuation.

02

Marr Labs

High barrier Early YC W2024 · San Francisco, CA · Voice AI (lending)

AI voice agents that place outbound lending & collections calls, built to sound human.

Lenders / collections $/call (likely) Marr Labs delivers Human-indistinguishable calls

No revenue public · ~$2M pre-seed

First customers

guessWarm intros via the founders' prior exit (Vlingo → Nuance, $225M) into voice-tech and BPO circles.

Growing via

guessPilot-by-pilot with a handful of lenders, expanding once compliance and error rates are proven.

Copy this?

High barrier. Needs both the voice tech and hard-won trust in a regulated, compliance-heavy vertical.

Dave Grannan, Han Shu (ex-Vlingo), Ghinwa Choueiter · ~$2M pre-seed.

03

Vapi

Medium barrier Proven YC-backed · San Francisco, CA · Voice AI infrastructure

The plumbing for voice AI — telephony, STT, TTS, latency — so others don't build it themselves.

Developers / companies Usage-based API Vapi delivers Voice-agent infrastructure

~$8M ARR · $500M valuation

First customers

guessEarly adopters likely found it via API docs and developer Twitter, well before enterprise deals.

Growing via

confirmedWon Amazon Ring's business over 40+ rivals (TechCrunch) — now selling into larger enterprise deals.

Copy this?

Medium-high barrier. The stack is copyable; the reliability track record that wins bake-offs isn't.

Jordan Dearsley, Nikhil Gupta · ~$72M total ($50M Series B, Peak XV).

04

11x.ai

Low barrier Contested YC-backed · San Francisco, CA · AI sales agents

AI 'digital workers' meant to replace the entry-level SDR doing outbound prospecting.

Sales orgs Seat/outcome-based 11x.ai delivers AI outbound prospecting

~$350M valuation · results publicly disputed

First customers

guessInvestor- and press-driven inbound, not the outbound motion the product itself is meant to automate.

Growing via

guessFast fundraising and hype look to have outrun proven reliability — ZoomInfo reported failed customer trials.

Copy this?

Low barrier technically — many competitors exist. The unsolved part may be making it reliably work, not building it.

Hassaan Raza · ~$74M raised in under 18 months (a16z among backers).

05

Sazabi

Medium barrier Early YC-backed · San Francisco, CA · AI observability

Debugging tooling built for how AI agents fail — not clean stack traces, but drift and bad handoffs.

Eng teams running agents Usage-based (likely) Sazabi delivers Agent-native debugging

Pre-revenue · $8M seed

First customers

guessFounder's network from his first exit (Opkit → 11x AI) — other AI-agent builders.

Growing via

guessDirect outreach; the buyer category (teams running agents in production) is still small and forming.

Copy this?

Medium barrier. Real problem today, but could get absorbed as a feature by Datadog-style incumbents.

Sherwood Callaway (ex-Opkit) · $8M seed.

06

Cal AI

Low barrier Proven Bootstrapped, 2024 · Independent (not YC) · Consumer health app

Photo of your food in, calorie and macro estimate out — no manual logging.

Individual users $/mo subscription Cal AI delivers Instant calorie estimate

$40M+ ARR · self-funded, later acquired by MyFitnessPal

Real, dated data · Downloads

5M Mar 2025

15M+ ~2026

First customers

confirmedThe founders' own viral TikTok content — not a funded acquisition channel.

Growing via

confirmedOrganic short-form virality — 5M downloads in 8 months, 15M+ since.

Copy this?

Low barrier. The AI is commodity; what's hard to copy is the founders' own distribution instincts.

Zach Yadegari, Henry Langmack (both 18 at launch) · self-funded.

07

Lingo.dev

Medium barrier Proven YC F2024 · San Francisco, CA · Developer tools (localization)

AI translates an app's UI text and opens the pull request itself — no manual i18n.

Eng teams Usage-based Lingo.dev delivers Translated PRs, ready to merge

~$1.1M ARR at 7 people · $4.2M seed

First customers

guessSelf-serve signups via GitHub/HN — a 7-person team has no headcount for outbound.

Growing via

guessClassic PLG: try it on one project, value is obvious, usage expands with the app.

Copy this?

Medium barrier. Translation is commodity now; the PR-generation workflow is the current edge.

Max Prilutskiy, Veronica Prilutskaya · $4.2M seed (Initialized Capital, YC).

08

winfunc

High barrier Early YC S2024 · San Francisco, CA · AI security

AI that finds a security bug — then proves it's real by actually exploiting it.

Security / eng teams Pricing TBD winfunc delivers Verified, exploited vulnerabilities

No funding beyond YC seed · $4.5K in Chrome bug bounties

First customers

guessPrior open-source project (Devika) likely gave the founders an existing technical audience.

Growing via

confirmedReal, Google-verified Chrome VRP bug bounties — credibility before any sales motion.

Copy this?

High barrier. Exploiting (not just flagging) a vulnerability needs rare agentic reasoning.

Mufeed VH, Vivek R (ex-Devika) · YC seed only, publicly.

09

Truewind

Medium barrier Proven YC 2023 · San Francisco, CA · AI accounting

AI does the bookkeeping and month-end close — instead of a human accountant.

Startups / SMBs Subscription Truewind delivers Automated bookkeeping

$17M+ raised · one unverified high-revenue claim, not repeated as fact

First customers

guessWarm intros via the YC/investor network — other startups needing bookkeeping.

Growing via

guessReferrals plus SEO around “AI bookkeeping”; Thomson Reuters (a Series A investor) is a plausible future distribution channel.

Copy this?

Medium barrier. QuickBooks and Xero are bolting on AI fast — the edge has to be being meaningfully better.

Alex Lee, Tennison Chan · $17M+ total ($13M Series A, Rho Capital & Thomson Reuters Ventures).

10

Operand

High barrier Early YC 2024 · San Francisco, CA · AI-native law firm

Not software for lawyers — an AI-run law firm that keeps the legal fee itself.

Legal clients Legal fee (not license) Operand delivers AI-performed, attorney-overseen work

Seed-stage, 8 people · figures not public

First customers

guessFounders' own network, or a narrow legal task where the trust bar for a new provider is lower.

Growing via

guessLikely capped by attorney-oversight capacity, not marketing spend — the norm in licensed services.

Copy this?

High barrier — and not mainly technical. Licensed-attorney oversight and liability exposure are the real gate.

Ram Gorthi, Akhil Iyengar, Arjun Sahney · seed-stage, figures not public.

11

LunaBill

Medium barrier Early YC F2025 · United States · Healthcare AI (billing)

AI automates medical billing and claims coding for healthcare providers.

Healthcare providers Per-claim / % (likely) LunaBill delivers Automated billing & coding

$5.4M raised · founder's prior co. (Skiesoft) reached real scale, bootstrapped

First customers

guessFounder's existing healthcare network from Skiesoft (Taiwan) — not a cold start.

Growing via

guessNo independent data yet for LunaBill itself; strongest evidence is the founder's prior track record.

Copy this?

Medium-high barrier. Good AI fit, crowded category — the founder's trust-building skill is the real edge.

David Day (previously bootstrapped Skiesoft) · $5.4M (Accel, YC, Sequoia scout).

12

Toma

High barrier Proven YC W2024 · San Francisco, CA · Voice AI (auto dealerships)

AI voice agents that answer dealership phones — booking service, answering sales questions.

Car dealerships Subscription (likely) Toma delivers AI phone agents, sales & service

$17.5M raised (a16z-led) · YC says “millions of dollars” in revenue

First customers

guessLikely direct dealership outreach — both founders came from Scale AI/Uber/Lyft/Amazon, not the auto industry, so credibility was probably built deal-by-deal.

Growing via

confirmedTechCrunch reported the voice agents “took off” at dealerships, drawing a16z's investment off real usage, not just a pitch.

Copy this?

High barrier. Auto dealerships are relationship-driven and skeptical of tech — the wedge is trust and workflow depth (DMS integrations, phone systems), not the voice AI itself.

Monik Pamecha, Anthony Krivonos (both ex-Scale AI) · $17.5M total (a16z-led).

13

LedgerUp

Medium barrier Early YC S2024 · San Francisco, CA · AI billing / revenue ops

AI that runs the messy contract-to-cash billing work for B2B SaaS companies with custom pricing.

B2B SaaS finance teams Subscription (likely) LedgerUp delivers Automated billing & revenue ops

Funding not public · ~$750K revenue, one source, not corroborated

First customers

guessFounder Bailey Spell has publicly described customers with quantified billing gaps (one CEO “missing $35K in revenue”) — plausibly the actual pitch that landed early customers.

Growing via

guessA 5-person team at this revenue level suggests self-serve or founder-led sales into SaaS finance teams, not a build-out sales org.

Copy this?

Medium barrier. Usage-based, custom-contract billing logic is genuinely fiddly, but well-funded players (Stripe Billing, Maxio, Chargebee) are adding AI too — the edge has to be depth on the messiest contracts, not the AI itself.

Bailey Spell, Joseph Johnson · funding amount not public.

14

Harper

High barrier Proven YC W2025 · San Francisco, CA · AI-native insurance brokerage

Not software for insurance brokers — an AI-run brokerage that keeps the commission itself.

Mid-sized businesses Broker commission (not license) Harper delivers AI-run insurance brokerage

$6M+ annualized premiums since Oct (confirmed) · $45M raised (a16z-led)

First customers

confirmedSince acquiring their first customers in October, Harper says it grew to write over $6M in annualized premiums across 35 states — stated directly in their own YC launch post.

Growing via

guessLikely broker-to-broker referrals and direct outreach to mid-sized businesses — the founders' backgrounds (Goldman, Carlyle, Coatue) suggest a relationship-driven enterprise motion, not self-serve.

Copy this?

High barrier — and not mainly technical, the same pattern as Operand (law). Insurance brokering requires licensing, carrier relationships, and underwriting trust built over years; the AI automates internal workflow, but the entry cost is regulatory and relational.

Dakotah Rice, Tushar Nair (both ex-Goldman; previously co-founded Poolit, scaled to $100M AUM) · $45M combined Series (a16z-led).

15

Bravi

High barrier Early YC F2025 · Paris, France · Voice AI (home services)

AI voice and chat agents that answer the phone for home-service installers so no lead goes unanswered.

Installers & manufacturers Subscription (likely) Bravi delivers 24/7 lead capture & quoting

Funding not public · live with installers across Europe/US

First customers

confirmedThe co-founder's own family shutter-installation business, plus his side e-commerce shop, were the first real testing ground before selling to outside installers.

Growing via

guessBest guess: expanding from the founders' home vertical (shutters) into adjacent trades (solar, HVAC, carpentry) via warm intros from the same installer/manufacturer network, rather than broad marketing.

Copy this?

High barrier. The AI itself is a commodity voice/chat stack; the edge is deep, personal founder-market fit in a fragmented, non-technical vertical most outsiders wouldn't have the patience or trust to sell into.

Anas Bouassami, Pierre-Habté Nouvellon (both ex-Snipfeed, $7M raised, acquired 2024) · funding amount not public.

16

Prism

High barrier Early YC P2026 · London, UK · AI-native recruiting agency

Not software for recruiters — an AI-run agency that finds, engages, and delivers candidates itself.

Hiring companies Placement/retainer fee (likely) Prism delivers Pre-screened, interview-ready candidates

$3.3M raised (Variant, Vento, YC, angels)

First customers

guessNo public detail. Best guess: given the founders' own recent hiring/recruiting backgrounds, first clients were plausibly companies in their own early-career network.

Growing via

guessToo early for public data. Part of a small emerging cluster of AI-native recruiting-agency YC bets (Prism, Perfectly) — the same structural pattern as the AI-native law and insurance firms elsewhere on this list.

Copy this?

High barrier — same pattern as Operand and Harper. Sourcing/outreach automation is replicable tech; the harder part is candidates and hiring managers trusting an AI-run agency with something as high-stakes as a job placement.

Theo Kitsberg (ex-Cambridge, crisis-helpline recruiting), Axel La Pira (ex-Oxford, founder associate at Alan) · $3.3M raised.

17

Brickanta

Medium barrier Early YC F2025 · Stockholm, Sweden · AI for construction estimation

AI reads construction estimates and bid documents, flagging missing items before budgets blow up.

Construction estimators / PMs Seat/project SaaS (likely) Brickanta delivers Automated document analysis & risk flags

$8M raised (investors behind OpenAI, Airbnb, Klarna, PlanGrid-Autodesk)

First customers

guessNo public detail on paying customers yet. Best guess: the founding team's prior industry roles (ABB, Fabege, Husqvarna, IKEA, Konecranes since 2018) gave them a direct network of construction firms to pilot with.

Growing via

guessBest guess, per their own launch post: deliberately focused on the Swedish/European market (trained on Eurocodes standards) before expanding — a narrow beachhead, not a broad launch.

Copy this?

Medium barrier. Document-heavy compliance/estimation review is a good LLM fit, and PlanGrid-Autodesk-adjacent investors clearly agree, but construction software has a notoriously long sales cycle and deep incumbent lock-in (Autodesk, Procore) — the tech is copyable, the enterprise trust and workflow integration is the slow part.

Lucas Otterling, Linus Bein Fahlander · $8M raised.