Money for Machines, Markets for Beliefs: Where YC Fintech Went After Payments

The wave right now

The 2023–2026 fintech cohort has quietly abandoned the last decade's playbook. Nobody here is building a neobank, a challenger card, or a cheaper remittance corridor. Instead, the cohort is wiring up money for two economic actors that barely existed three years ago: AI agents that spend, and prediction markets that price beliefs. The third front is less glamorous but probably bigger — AI eating the salaried labor inside regulated financial workflows: underwriting, loan servicing, brokerage, research. If you're choosing what to build this quarter, the question is no longer "which financial product can I make 10x cheaper?" It's "which financial process still assumes a human is in the loop — and what breaks when one isn't?"

The landscape today

1. Agents need a wallet before they need anything else. The card networks assume a human at checkout; autonomous agents break that assumption instantly, and two companies are racing to be the issuance layer. Agentcard issues disposable, per-charge debit cards that agents can spend with under human approval, while Allowance frames the same primitive as scoped credentials — one-time card numbers with enforceable spending rules. The consumer-facing cousin is Uno Wallet, which lets AI pick the optimal rewards card at tap-to-pay. The issuance problem will be solved fast; the durable business is whoever owns the trust layer around the transaction — limits, attribution, disputes.

S26

Agentcard

debit cards for AI agents.

ActiveFintech

P26

Allowance

Scoped payment credentials for AI agents

ActiveAI Agent Payment Credentials1 ppl

P26

Uno Wallet

The AI wallet that picks the best card at checkout.

ActiveAI Credit Card Optimization2 ppl

2. Insurance is the cohort's stealth majority — and agents are its newest risk class. Count the entries: a meaningful slice of this entire fintech cohort is insurance, and it splits cleanly in two. One camp underwrites the AI economy itself: Mount is building an AI carrier with real-time risk scanning for agentic systems, and Klaimee sells liability cover plus certification for autonomous agents — both betting that an agent with no loss history is an underwriting problem worth owning early. The other camp uses AI to attack the brokerage cost structure: Panta runs an AI-native commercial brokerage for construction and logistics, Hedge does the same for hard-to-place specialty E&S risk, and Huscarl brings actuarial intelligence to corporate risk managers. The brokerage plays have clearer near-term revenue; the agent-insurance plays have the bigger prize if liability regulation lands.

P26

Mount

The AI Insurance Carrier

ActiveAI Agent Liability Insurance2 ppl

P26

Klaimee

Liability insurance for AI Agents. You deploy agents, we cover you.

ActiveAI Liability Insurance2 ppl

W26

Panta

AI Native Commercial Insurance Brokerage

ActiveAI Commercial Insurance Brokerage3 ppl

P26

Hedge

AI-Native specialty insurance company

ActiveSpecialty Insurance AI2 ppl

3. Prediction markets got their Wall Street moment — and the picks-and-shovels rush arrived within two batches. Once Kalshi and Polymarket proved real volume, YC funded the entire institutional stack around them almost simultaneously. River Markets is the prime brokerage; Valence is the unified cross-venue terminal with arbitrage detection. Below them, ValCtrl is building the taxonomy-and-search layer that maps fragmented markets to each other, Totalis adds a derivatives layer, and Sequence Markets routes low-latency execution across crypto, prediction, and tokenized venues. This is five infrastructure companies for an asset class younger than most of their founders' careers — exhilarating, and exactly what saturation looks like at the terminal/execution layer.

P26

River Markets

Prime brokerage for prediction markets

ActivePrediction Market Brokerage2 ppl

W26

Valence

Unified prediction markets trading platform

ActivePrediction Market Trading Infrastructure3 ppl

P26

ValCtrl

ValCtrl is the world model for Prediction Finance, and Kassandre…

ActivePrediction Markets Infrastructure2 ppl

4. The quietest pattern is AI replacing regulated back-office labor. Zolvo automates servicing for factoring and asset-based lenders — invoice verification, reconciliation, collections — and Kita automates credit review for emerging-market lenders end-to-end, from document intake to decision-ready memo. On the buy side, Kimpton AI generates portfolio-personalized research and trade proposals, while Standard Signal skips selling the tool entirely and runs the hedge fund itself. The lending-ops wedge is the underrated one: the buyer is desperate, the workflows are document-shaped, and almost nobody else in this cohort is competing for it.

P26

Zolvo

AI that automates servicing for commercial lenders

ActiveCommercial Loan Servicing AI4 ppl

W26

Kita

Automate credit review for lenders in emerging markets

ActiveEmerging Markets Credit Automation2 ppl

P26

Kimpton AI

AI-Native Investment Research

ActiveAI Portfolio Research Platform5 ppl

The cohort signal

This is a deliberate, accelerating program bet, and the batch clustering proves it. Winter 2026 and Spring 2026 between them contain essentially the entire thesis: the prediction-market stack landed as a five-company cluster across those two adjacent batches, agent payments arrived as a pair in Spring–Summer 2026, and AI-native insurance shows up in both W26 and Sp26 with five companies. Spring 2026 alone carries more fintech density than any earlier batch in this window. Notably, many of these — Mount, River Markets, Totalis, Panta — are renames, meaning founders already in the program *pivoted into* these themes mid-flight. When pivots converge on the same three patterns, that's the strongest demand signal YC data gives you.

Lessons from the last cycle

The veterans tell a consistent story about where fintech value actually accrues. Stripe and Razorpay compounded for over a decade because they owned infrastructure, not interfaces — the lesson the agent-payments founders are clearly studying. Coinbase, the only public exit in the set from a new asset class, won by becoming the *regulated chokepoint* for crypto, not the best trading UI — directly relevant to the prediction-market five. And the cautionary beat: Brex, Newfront, and Truebill all built real businesses and all exited via acquisition — brokerage and consumer-finance models tend to consolidate into someone else's balance sheet rather than IPO. Build the rail, not the storefront.

If you're building here

Three concrete openings:

• Post-transaction trust for agent commerce. Agentcard and Allowance solve issuance; nobody in this cohort owns disputes, refunds, chargeback attribution, or "know your agent" compliance. That's the Stripe-shaped layer, and it's empty.
• Credit operations beyond the first two wedges. Zolvo has factoring/ABL and Kita has emerging-market underwriting — that leaves construction lending, equipment finance, and CRE servicing untouched by this entire cohort. Same playbook, open territory.
• Uncontested single-player niches. Palus Finance is alone on SMB treasury yield; ZeroSettle is alone on direct-billing SDKs as app-store fee regulation bites. Both are markets where a second, differently-angled entrant still gets oxygen.

The tarpits, by name: another prediction-market terminal or execution layer (five companies in two batches — the venue-aggregation race is run); another AI research copilot for portfolio managers (Kimpton AI, KelAI, Soria, and Wealor are already fighting over the same PM's screen); and consumer AI money assistants, where Gravy and Uno Wallet face the Truebill precedent — workable product, capped outcome.

What you'd have to believe: that agents transact at meaningful scale by 2027 and card networks don't ship native agent rails first; that prediction markets keep their regulatory clarity through a US election cycle; and that in insurance, carriers and brokers genuinely can't retrofit AI faster than a two-person team can acquire licenses. If you believe all three, the openings above are live this quarter. If you only believe the third, go build in lending ops — it pays either way.