Merchant-kiosk origination OS for Mexican non-bank lenders to launch minute-scale small-ticket loans without adding branches.

1. Series A capital earmarked for geographic expansion and new products means kiosk-led lending is moving from niche experiment to rollout phase, creating a platform window for suppliers.
2. AI-guided video applications processed within minutes make assisted origination fast enough to support small-ticket lending economics instead of just slower branch workflows.
3. A 23-state footprint and 300,000 completed applications show the bottleneck is no longer proving borrower demand; it is operating distributed access points with consistent quality and risk controls.
4. Loans sized at $100-$1,000 for informal-economy consumers and small businesses reveal a segment large enough to justify purpose-built infrastructure but too small-ticket for branch-heavy distribution.

Catalyst. Aviva's $18 million raise, 23-state footprint, and 300,000 completed applications show kiosk-plus-AI underwriting is no longer an experiment, making the enabling software layer newly sellable right now.

The product sits between a lender's loan-origination system and every merchant-hosted kiosk, tablet, or video application point. It gives merchant staff a scripted assisted-application flow, real-time document quality checks, fraud flags, and instant escalation to remote loan officers when a case falls off the happy path. Headquarters gets location-level conversion, turnaround, and quality dashboards plus remote QA playback to retrain or suspend weak merchant points quickly. The first deployment is a read-only workflow layer on top of the lender's existing stack, so customers can launch a new municipality or partner network without replacing their core system. Over time, the startup builds a proprietary dataset on merchant-point behavior, drop-off patterns, and approval outcomes that improves rollout, routing, and risk controls.

What's different. Generic loan-origination systems assume either self-serve digital flows or staffed branches; they do not manage merchant scripts, application quality, remote QA, and location-level risk across a kiosk network. This startup owns the control plane for assisted origination at the retail edge, where the valuable data is not just borrower attributes but merchant behavior, drop-off patterns, and launch performance municipality by municipality. That operating dataset compounds into better rollout playbooks, fraud controls, and switching costs over time.

Jobs to be done

flowchart LR
  Buyer[COO at non-bank lender] --> Pain[Underbanked borrowers need assisted onboarding but branches are too expensive]
  Pain --> Product[Merchant-kiosk origination OS]
  Product --> Outcome[Faster expansion with standardized risk and conversion]

Executive takeaways

• Aviva’s new $18M Series A and earlier IDB Lab backing show that kiosk-assisted lending is now scaling beyond pilot stage: the company says it operates in 23 states, has processed more than 300,000 applications, and is broadening its product base.[51][95]
• Demand exists because formal credit access is still thin relative to need: IDB Lab says fewer than 35% of individuals and 10.5% of businesses have formal credit, while ENIF-based summaries still put formal-credit penetration at 37.3% and heavy informal saving at 36.6%.[50][74][90]
• The wedge is not better scoring alone but disciplined merchant-edge operations. Mexico already uses correspondents, OXXO, Yastás, and store-linked finance to extend reach; what most lenders still lack is a neutral workflow layer for scripts, QA, and fraud controls across distributed points.[37][43][46][49][118]
• The market is real but crowded with substitutes: digital lenders, store-lenders, generic LOS vendors, and in-house field ops all attack pieces of the problem, so the startup wins only if it shortens state launches and reduces risk-ops variance faster than those alternatives.[20][21][39][109][111][116]

Market definition

Software and workflow infrastructure for Mexican non-bank lenders that want to launch or govern assisted small-ticket lending through merchant-hosted kiosks, tablets, or video booths—sitting between a lender’s LOS and a distributed physical network to standardize scripts, document capture, QA, routing, and compliance.[50][95][109][111][116]

Customer and buyer

The practical buyer is the COO or head of expansion / credit operations inside a SOFOM, fintech lender, or retailer-linked credit group that is pushing beyond self-serve digital into branch-light physical access. Aviva’s SOFOM disclosure, Konfío’s SOFOM disclosure, and CONDUSEF’s SIPRES registry all point to an operating environment where process control matters as much as demand generation.[58][14][75]

Buying triggers

• A new-state or new-partner rollout makes merchant consistency and remote QA more urgent than branch hiring. [4][50][95]
• Fraud, incomplete applications, or complaint-handling risk grows when assisted points proliferate without standardized scripts and audit trails. [58][75][97][116]
• Management decides to piggyback on merchant or correspondent networks rather than fund more owned branches. [46][49][118]

Willingness to pay

This looks like an operations-and-risk budget line, not a marketing widget. Lenders already pay for credit ops, identity checks, and origination software; Konfío publishes enterprise credit economics, LoanPro and TurnKey pitch compliance-heavy LOS stacks, and Aviva / FEMSA are already investing real capital into physical-plus-digital lending infrastructure. [51][53][66][109][116]

Category dynamics

Growth signal Banking-app usage among Mexican account holders rose from 54.3% in 2021 to 69.1% in 2024.

Validation signals

• Aviva says more than 300,000 people have already completed loan applications through kiosks across 23 Mexican states.
• IDB Lab expected Aviva to reach at least 150 kiosks by year-end after its venture-debt support.
• Yastás markets itself as a nationwide neighborhood-store network for payments, deposits, withdrawals, and more than 2,000 services, showing consumers already accept merchant-edge financial behavior.
• FEMSA is explicitly pairing OXXO’s omnichannel reach with a dedicated lending venture and QED’s fintech expertise.

Regulatory & technical constraints

• Third-party assisted origination still needs lender-level compliance, complaint handling, and clear entity-of-record disclosures, as Aviva and Konfío themselves disclose their SOFOM status and customer-service structures.
• Remote identity and document capture must meet KYC and AML expectations; major vendors market this as built-in compliance rather than optional UX.
• Cash-heavy merchant environments and rural digital gaps mean the workflow has to tolerate mixed online and offline evidence collection.

Direct Mexico-first assisted-origination software still appears sparse, but substitutes are strong. Aviva proves the phygital model as a vertically integrated lender; Kueski and Klar attack borrower convenience through pure digital channels; Banco Azteca and FEMSA show the power of owned physical distribution; and TurnKey Lender, LoanPro, and Mambu already sell general lending stacks. The gap is a neutral merchant-edge control plane optimized for third-party points, not owned branches or app-only funnels.[50][95][21][39][43][109][111][116]

Why incumbents do not win by default

• Generic LOS / lending cores. Platforms such as LoanPro, Mambu, and TurnKey Lender automate origination and servicing, but they are not opinionated about merchant scripts, remote QA playback, or municipality-level rollout discipline.
• Digital lenders. Kueski and similar apps remove paperwork for digital borrowers, but their model assumes self-serve mobile completion rather than assisted onboarding at neighborhood merchants.
• Retail financial networks. OXXO/Spin, Yastás, and Banco Azteca already own dense customer touchpoints, yet their incentive is to distribute their own ecosystem or captive products, not become neutral orchestration software for rival lenders.
• Identity / fraud vendors. Incode, Entrust, Trulioo, and similar vendors solve document capture and verification, but not end-to-end merchant governance, routing, or launch analytics.
• In-house field ops. Manual field-agent and call-center follow-up work for a few locations, but Aviva’s scale and Mexico’s correspondent footprint show that distributed assisted lending becomes a workflow-control problem once networks stretch across states.

Merchant Kiosk Origination OS targets Mexican non-bank lenders and retailer-credit groups that want to launch $100-$1,000 loans through 50-200 merchant-hosted points without adding branches. The researched signal is that borrower demand is already real—Aviva reports a 23-state footprint and more than 300,000 completed kiosk applications—but rollout discipline, QA, fraud control, and complaint handling become the bottleneck once assisted origination spreads across third-party points. The first product should not replace the lender's LOS or underwriting stack; it should overlay guided scripts, document capture, remote QA playback, exception routing, and location-level analytics on top of the systems buyers already use. The first customer is a growth-stage SOFOM or fintech lender entering new states or merchant programs while a central risk team still audits assisted applications manually, and the first contract should be a paid rollout-control pilot tied to faster activation and fewer bad applications. This beachhead is intentionally narrow because Mexico already has strong substitutes—generic LOS vendors, field teams, digital lenders, and captive retail networks—so the company must win on specific operating KPIs rather than a broad AI-lending narrative. The modeled Mexico-only TAM of roughly $30.0M and year-three SOM of roughly $1.2M are enough for a wedge but not yet enough for a venture case unless the same control plane expands into renewals, collections, adjacent products, and possibly other geographies after proof. The biggest disconfirming risks are that too few lenders already operate assisted networks large enough to buy software and that integrations or compliance requirements make deployments services-heavy. Research also leaves real gaps around active buyer count, point-level fraud baselines, and exact budget ownership, so the first 18 months should prioritize repeatable deployment and paid conversion before aggressive hiring.

Problem

• Branch and field-agent economics break when lenders try to originate small-ticket credit across many tier-2 and tier-3 municipalities, yet self-serve digital flows still miss borrowers who need assisted onboarding.
• Once dozens of merchant points are live, application quality, fraud controls, disclosures, and complaint handling become inconsistent because generic LOS tools do not govern merchant behavior or remote QA.
• Expansion teams therefore face a speed-versus-control tradeoff that slows new-state rollouts exactly when phygital lending is becoming financeable.

Solution

• Provide a workflow overlay between the lender LOS and each merchant-hosted kiosk, tablet, or video point with guided scripts, document capture, low-bandwidth assist, QA playback, and exception routing.
• Give headquarters location-level dashboards, merchant scoring, suspend or retrain controls, and audit logs so central risk can scale third-party points without building branches.
• Launch as a read-only or lightly integrated overlay first, then automate approvals, renewals, and adjacent workflows only after the base rollout playbook is repeatable.

Why we win

• The startup targets the missing merchant-edge control plane that vertically integrated lenders, digital apps, and generic LOS vendors do not offer as a neutral product.
• Merchant-point conversion, QA, and fraud data can compound into a cross-lender operating dataset and rollout playbooks that are hard for a single lender stack to replicate.
• A software-first overlay avoids taking lending risk and shortens time to value versus building branches, captive networks, or a replacement origination core.

Milestones

flowchart LR
  Wedge[50-200 point lender rollout pilot] --> MVP[Merchant-edge workflow overlay]
  MVP --> Proof[Faster launches, lower QA variance, fewer bad applications]
  Proof --> Expansion[More states, more points, adjacent lending workflows]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least 6 of the first 20 target lender or retailer-credit accounts already run or are launching 50 or more assisted points and lack a purpose-built governance layer.
• One launch-stack integration path can bring a pilot live in 60 days or less without replacing the lender's LOS.
• The product can reduce manual QA touches by 30% or more and improve completed applications by 15% or more in the first three pilots.
• Buyers will convert from pilot to annual contracts at roughly US$120k or more ARR because the software maps to operations and risk budgets.
• The same control plane can expand into renewals, collections handoff, or adjacent onboarding workflows without becoming a bespoke services business.

Open diligence questions

• How many active Mexican lenders today operate or plan more than 50 assisted points, and what tooling do they use?
• Which budget owner signs the first contract in practice: COO, head of risk operations, CTO, or a retail-partner P&L owner?
• What are current fraud-loss, coached-application, abandonment, and manual-QA baselines by merchant point?
• Which LOS and identity vendors cover the majority of reachable beachhead accounts?
• Will merchant-network operators refer neutral software, or will the best locations stay locked inside captive ecosystems?

Model sanity

• Revenue engine. Base revenue comes from growing from 3 paying logos at Y1 exit to 8 at Y3 exit while mature logo value moves from the $150K ACV midpoint toward a roughly $1.3M ARR exit rate through usage and premium-control upsell.
• Must go right. The funding ask only works if the overlay stays deployment-light enough that 8 logos can be supported with 8 FTE and gross margin clears 70% by late Y3.
• Model breaks if. The downside case and sales-cycle sensitivity show that slower partner conversions or fragmented integrations can push cash toward roughly $0.4M while holding the company at only 7 logos by Y3 exit.
• Next-round proof. A credible seed story exists once the pre-seed gets the company to 5 production logos, one partner-assisted second-stack deployment, and sub-45-day implementations by Month 24.

Scenarios

Sensitivity

flowchart LR
  QualifiedAccounts --> PaidPilots
  PaidPilots --> ProductionLogos
  ProductionLogos --> UsageFees
  ProductionLogos --> PremiumControls
  UsageFees --> Revenue
  PremiumControls --> Revenue
  Revenue --> GrossProfit
  GrossProfit --> Cash

Flags: Revenue per FTE is still only about $125K in Y3, which is below software benchmarks and implies the workflow remains partly services-like until reuse or adjacent modules lift ARPU. · A25 is a real risk: if one solutions engineer and one merchant-ops lead cannot support 8 logos across two launch stacks, both gross margin and the funding ask deteriorate quickly. · Research still leaves buyer-density and budget-owner uncertainty unresolved, so fewer than roughly 6 beachhead accounts with near-term rollout plans would lengthen CAC and the sales cycle. · The researched $1.2M SOM is better read as an exit-rate lens than a full-year revenue number because the eighth logo lands late and Q4Y3 ARR is about $1.3M while full-year Y3 revenue is about $1.0M.

• Merchant-point fraud. Third-party kiosk hosts can coach applicants, mishandle documents, or generate synthetic volume that poisons underwriting. Mitigation: Start with remote QA playback, location scoring, configurable approval thresholds, and a one-click suspend workflow for risky points.
• Thin lender software budgets. Small-ticket lenders may treat assisted-origination tooling as a cost center unless it clearly beats branch or field-agent economics. Mitigation: Sell into state-launch moments and price by live point or completed application so ROI is tied to faster rollout and lower abandonment.
• Workflow fragmentation. Each lender's origination stack, risk rules, and compliance process may differ enough to make deployments look services-heavy. Mitigation: Launch as a read-only workflow overlay with a narrow integration set, then expand automation only after one repeatable playbook proves out.