Pre-writeoff cure orchestration for auto lenders that personalizes outreach and repayment paths before loans roll to write-off.

1. When large enterprises can have millions of overdue accounts and 10-15% of revenue tied up, delinquency control already has a board-level budget.
2. A distinct software layer is forming between payment due date and write-off, so startups can wedge into pre-collections without replacing the entire servicing stack.
3. Claimed 11.5% lower write-off rates and 13.6% higher customer lifetime value turn recovery tooling into an ROI-backed revenue product, not just a cost-center tool.
4. Two years and more than 200 million interactions suggest the training data now exists to personalize channel, message, and timing instead of blasting one-size-fits-all campaigns.
5. Auto-lending expansion and existing FICO integration mean first customers can adopt through an overlay workflow rather than a long core-servicing replacement project.

Catalyst. KredosAi's financing, measured write-off and customer-lifetime-value lift, and expansion into auto lending with FICO integration show that behavioral pre-collections tooling has moved from an abstract AI promise into a buyable servicing system.

The startup plugs into loan-servicing data and FICO-style collections queues to score each late account before it reaches write-off. It recommends or executes the next best cure action across SMS, email, voice, and collector worklists, tailoring message, channel, timing, and repayment path to each borrower cohort. Managers can see which segments respond to light-touch nudges, hardship-plan offers, or live-agent escalation and can run controlled tests by delinquency bucket and book type. Because the product is an overlay, lenders can keep their servicing core, dialers, and agency relationships while improving cure rates and reducing avoidable write-offs. Over time, the company builds a proprietary response graph across borrower behavior, vehicle economics, and treatment outcomes that turns collections policy into a learning system.

What's different. Dialers and collection agencies maximize contact volume, while most collections suites expose rules and queues. This startup owns the behavioral treatment layer between them: it decides which borrower should receive which message, through which channel, at which moment, and when to escalate to a human. Because it is designed as an additive layer over FICO-driven workflows, it can win without forcing lenders to replace their core stack. Its moat compounds from lender-specific treatment-response data, repayment-path outcomes, and the workflow graph of which interventions cure which delinquency cohorts.

Jobs to be done

flowchart LR
  Buyer[Head of collections] --> Pain[Static outreach misses cures before write-off]
  Pain --> Product[Pre-writeoff cure orchestration]
  Product --> Outcome[Higher cure rates and lower write-offs]

Executive takeaways

• Auto lenders are already managing a very large and stressed book: the New York Fed put U.S. auto balances at $1.685T in Q1 2026, Equifax saw 87.4M outstanding auto accounts and a 1.37% 60+ DPD rate in May 2026, and the CFPB still counted more than 100M active auto-finance accounts in late 2024 ([2], [4], [6]).
• The pain is concentrated where the thesis points: New York Fed analysis says delinquencies are primarily concentrated in non-captive auto finance companies, while Equifax estimates subprime and deep-subprime borrowers represent 22.1% of outstanding auto-loan debt and dealer finance/monoline portfolios remain majority-subprime ([3], [5]).
• A wedge exists because incumbent stacks optimize broad collections workflows, not auto-loan-specific next-best-treatment loops: FICO, C&R, and Qualco all market end-to-end recovery control, while TrueAccord and InDebted prove digital-first collections can lift outcomes without making the lender's internal cure engine the center of gravity ([8], [9], [10], [13], [17], [19]).
• Budget justification is feasible when sold as avoided loss and better liquidation: KredosAI claims 11.5% lower write-offs and 13.6% higher CLV, TrueAccord says Snap Finance improved liquidation 25-35%, InDebted says 3 in 4 Snap customers resolve digitally and that partner scorecards correlate with 20.4% higher liquidation, and Qualco advertises +35% cash-flow lift ([1], [12], [16], [17], [34]).

Market definition

Software and workflow orchestration that sits between a missed auto payment and charge-off, using lender data, policy rules, and omnichannel communications to choose the next best cure action before repossession or outsourced recovery ([8], [9], [13], [17], [19]).

Customer and buyer

Primary users are collections strategy, servicing analytics, and digital servicing leaders at large auto lenders; the economic buyer is usually the head of servicing, chief collections officer, or COO because cure performance affects write-offs, staffing, and agency spend ([2], [3], [30], [31]).

Buying triggers

• A rise in 30+/60+/90+ delinquency or roll rates creates immediate pressure to improve cures before accounts tip into repossession or charge-off. [2][3][6][36]
• Operations leaders get pushed to add collectors, agency spend, or repo volume when manual treatment selection no longer keeps up with early-stage delinquency. [4][30][31]
• Lenders want to shift from phone-heavy, manually governed outreach into compliant digital channels without losing auditability. [11][15][21][25][28][29]

Willingness to pay

Budget exists when the system is sold as avoided loss, lower manual-touch cost, and faster liquidation rather than as generic AI. Category vendors repeatedly anchor value in revenue lift, liquidation improvement, or cash-flow gains instead of seat-count efficiency alone. [1][12][16][17][34][35]

Category dynamics

Growth signal Auto balances +2.4% YoY as of May 2026, with non-prime ABS still weaker YoY in March 2026

Validation signals

• KredosAI says enterprise deployments already deliver 11.5% lower write-off rates and 13.6% higher customer lifetime value.
• TrueAccord says Snap Finance improved liquidation rates by 25-35% with a digital-first, ML-driven approach.
• InDebted says 3 in 4 Snap customers resolve digitally and that partner scorecards correlate with 20.4% higher liquidation.
• Qualco markets +35% cash-flow lift and 3x more cases handled from its collections suite, showing buyers already value automation outcomes.

Regulatory & technical constraints

• Debt-collection communications and validation must fit Regulation F and the borrower-contact rules it codifies.
• Robocall and robotext workflows must support easy revocation of consent and one-to-one consent governance under evolving FCC/TCPA rules.
• Bank buyers must treat the vendor as a third party under a full lifecycle of planning, due diligence, contracting, and monitoring.
• SMS programs need explicit opt-in, confirmation, opt-out, and privacy controls aligned with CTIA and vendor-rail requirements.

The field splits into five classes: incumbent decisioning/workflow suites (FICO, C&R, Qualco), digital-first collections agencies (TrueAccord, InDebted), lender analytics/data tools (Equifax, TransUnion), messaging rails (Twilio/CTIA ecosystem), and lender-specific in-house rule stacks. The gap is an auto-loan-specific pre-writeoff control layer that improves message, channel, timing, and repayment-path decisions without forcing a full core-stack replacement ([5], [8], [9], [10], [13], [17], [18], [19], [32]).

Why incumbents do not win by default

• Decisioning and workflow suites. FICO, C&R, and Qualco already own broad recovery workflows, but their products are horizontal systems of record; a startup can still win with an auto-loan-specific overlay focused on treatment experimentation in the pre-writeoff window.
• Digital-first agency models. TrueAccord and InDebted prove that AI-driven, omnichannel collections can outperform static outreach, but they often sit as outsourced recovery motions rather than as an internal cure engine embedded inside lender operations.
• Credit analytics and portfolio tooling. Equifax and TransUnion provide risk, delinquency, and portfolio-health data, but they do not make lender-specific next-best-treatment decisions or own escalation policy execution.
• Messaging infrastructure. Twilio- and CTIA-class infrastructure provides the rails for compliant outreach, but not the policy engine that decides who should receive what treatment next.

Auto-loan Cure Orchestration should be built as a U.S.-first behavioral control layer for non-prime auto lenders that need to improve 15-59 day cures before accounts roll into repossession or write-off. Research supports a real and specific pain signal: U.S. auto balances exceed $1.6T, non-captive lenders carry the highest delinquency stress, and incumbent stacks already route collections through systems such as FICO rather than homegrown greenfields. The most credible first customer is a chief collections officer or head of servicing at a lender with 150,000-plus active contracts, worsening roll rates, and an existing FICO-style queueing workflow. The first product should not try to replace the servicing core, the dialer, or the agency network; it should start in recommendation mode on one delinquency bucket, prove lift with holdouts, and only then automate approved digital and voice treatments. Pricing, sales motion, and product scope must stay aligned to one enterprise wedge: a paid baseline diagnostic and 90-day pilot that converts into annual software when the lender sees lower write-offs, fewer manual touches, and better cure performance. The researched market is meaningful but not broad enough to justify a diffuse launch, with an estimated $58.0M TAM, $35.0M SAM, and a realistic year-three SOM of about $5.0M if the company lands five lighthouse customers. The reason to believe is that broad suites, agencies, and messaging vendors each solve only part of the workflow, leaving room for an auto-specific next-best-treatment layer. The main reason to doubt is that the core ROI evidence still comes from vendor claims and adjacent case studies rather than published non-prime auto holdout data, while consent-data completeness and procurement friction could slow deployment materially.

Problem

• Non-prime auto lenders still treat early-stage delinquencies with static rules, generic scripts, and outsourced dialing even as rising 30+ and 60+ roll rates put write-offs, repossession volume, and staffing budgets under pressure.
• Collections leaders lack a control layer that decides the next best message, channel, timing, and repayment path at the cohort level while preserving auditability and customer lifetime value.

Solution

• Overlay FICO-style collections queues with account-level treatment scoring, experimentation, and human-approval workflows that recommend the next best digital, voice, or collector action for each delinquent loan.
• Start with one 15-29 or 30-59 DPD cohort in recommendation mode, prove lift against a randomized control, then expand into automated execution, repayment-path selection, and escalation management across more buckets.

Why we win

• The wedge is narrower and more measurable than a full collections-suite replacement: it plugs into existing servicing and decisioning stacks, uses the lender's own delinquency book as the baseline, and can prove value in one cohort within one quarter.
• As deployments expand, the company builds a proprietary treatment-response graph across borrower tier, vehicle context, consent status, delinquency bucket, and repayment outcomes that incumbents and agencies do not automatically own inside the lender.

Milestones

flowchart LR
  Wedge[15-59 DPD auto cure pilot] --> MVP[Recommendation-mode orchestration]
  MVP --> Proof[Holdout-tested cure and roll-rate lift]
  Proof --> Expansion[More buckets channels and adjacent credit products]

Founding team

Experiment roadmap

Risk assessment

What must be true

• At least 40% of qualified beachhead lenders must already manage early-stage cure decisions through static rules, spreadsheets, or generic dialer playbooks.
• A single delinquency-bucket pilot must show at least 5% relative cure lift or 3% roll-rate reduction versus control within one quarter.
• Target lenders must have enough consent and contact-history data to activate digital treatments on a meaningful share of accounts without a full data-rebuild project.
• At least half of paid diagnostics must convert to production-track pilots because the workflow is already budgeted as loss mitigation, not AI experimentation.
• Five lighthouse lenders must be able to support about $5.0M ARR by year three before the company expands beyond auto.

Open diligence questions

• How often do non-prime auto lenders already use FICO or adjacent systems for early-stage queueing, and what API or file hooks are actually available?
• What level of cure lift versus control is sufficient for a chief collections officer to reallocate budget from agencies or headcount into software?
• How complete are consent, opt-out, and language-preference records across SMS, email, and voice in the first 10 target accounts?
• Will buyers accept recommendation mode first, or do they require automated execution before approving annual software spend?
• Which channel produces the fastest credible sale: direct founder outreach, FICO implementers, or digital collections BPO referrals?

Model sanity

• Revenue engine. Base revenue comes from converting three paid diagnostics into a four-logo production beachhead by Q4Y2 and then expanding five lighthouse lenders toward roughly $1.0M ACV by Q4Y3.
• Must go right. The company needs paid diagnostics to convert into production in about 150 days and at least two lenders to add second buckets or channels, or the Y3 exit ACV assumption breaks.
• Model breaks if. If deployments stay bespoke and the company exits Y3 with four logos instead of five, the downside case keeps EBITDA negative and pulls the cash floor toward roughly $0.8M.
• Next-round proof. The next financing case is three production logos, two partner-sourced pilots or referrals, and at least one expansion inside a live lender by late Y2 or early Y3, which is exactly what the $4.0M seed is sized to reach.

Scenarios

Sensitivity

flowchart LR
  Diagnostics[Paid diagnostics] --> Pilots[90-day pilots]
  Pilots --> Production[Production buckets]
  Production --> Expansion[More buckets and channels]
  Expansion --> Revenue[Revenue]
  Revenue --> GrossProfit[Gross profit]
  GrossProfit --> Cash[Cash and runway]

Flags: The Y3 base case lands the full researched five-logo SOM, so most upside must come from expansion inside each lighthouse lender rather than from a broad logo count. · customersEop includes paid diagnostics and pilots in Y1, so recurring production logos trail the headline count until late Y2. · The model assumes the heavy audit, legal, and integration spend of Y2 tapers meaningfully in Y3; if onboarding stays bespoke, gross margin and EBITDA will miss together. · Revenue concentration is high because five lenders represent essentially all Y3 revenue, so one delayed deployment can move the year materially. · Cash is modeled as EBITDA, so enterprise prepayment timing, procurement delays, or security-review retainers could shift the real cash curve even if the P&L lands on plan.

• Model lift may vary by lender book. A treatment strategy that works for one auto lender's borrower mix, channels, and policies may not produce the same cure lift elsewhere. Mitigation: Start with cohort-specific pilots and randomized holdouts, then expand only after proving lift against each lender's own baseline.
• Incumbents can bundle adjacent features. FICO, servicing platforms, or large collection agencies could add basic personalization features once the wedge is visible. Mitigation: Win by owning the auto-loan-specific treatment dataset and by integrating into incumbent workflows faster than they can ship vertical outcome models.
• Operational and policy friction. Lenders may hesitate to let automated systems touch borrower outreach, repayment options, or escalation rules in a sensitive servicing workflow. Mitigation: Launch first in recommendation mode with full audit logs and human approval, then graduate customers into automation on narrowly defined cohorts.