Wedge

The beachhead targets SaaS and cloud infrastructure vendor accruals for mid-market technology companies. Cloud computing bills are highly complex, volatile, and consistently arrive late, causing significant month-end variance and acute pain for the finance team. Once the system reliably predicts cloud vendor accruals, it expands into marketing agency spend, legal fees, and eventually supply chain accruals to own the entire accounts payable accrual ledger.

Timing

Context windows in foundation models now support ingesting hundreds of unstructured vendor contracts alongside thousands of historical ERP transactions to detect pricing tiers and billing schedules. Furthermore, modern ERP systems offer robust REST APIs that allow real-time read and write access to journal entries, enabling immediate syncing of predicted accruals.

Why This ICP

Mid-market controllers feel the acute pain of strict reporting deadlines but lack the massive offshore accounting armies of Fortune 500 firms. They possess enough transaction volume to generate statistically significant historical data, making them the ideal testing ground for predictive models.

Size Of Prize

The addressable market consists of roughly 50,000 US mid-market and enterprise companies that spend an average of $40,000 annually on accounting labor and external audit fees tied specifically to accrual calculations and reconciliation. This yields an annual economic prize of approximately $2B.

Gap Narrative

Mid-market and enterprise accounting teams manually calculate month-end accruals by hunting down open purchase orders, historical invoices, and vendor contracts. Existing ERP modules only capture explicitly entered data, forcing controllers to rely on lagging spreadsheet estimates that delay the financial close. An automated system that ingests vendor history and open POs to instantly predict and propose journal entries eliminates a multi-day bottleneck.

Defensibility

Defensibility compounds through proprietary workflow integration and data-driven prediction accuracy. As the API ingests more vendor billing patterns across multiple clients, the model's accuracy for common vendors improves, creating a network effect on precision. Switching costs become high once the finance team relies on the API to close their books days faster, as ripping it out reinstates a manual, multi-day delay.

Why This Thesis

An API-first software thesis fits perfectly because controllers refuse to abandon their existing ERPs as the general ledger system of record. Delivering the capability as an API that integrates directly into NetSuite or SAP ensures the prediction engine acts as a silent backend service, aligning with the ICP's strict compliance and workflow requirements.

Hardest Part

Achieving audit-grade precision in predicting unbilled expenses without generating false positives that artificially inflate month-end liabilities.

Min Viable Scope

V1 focuses exclusively on predicting Accounts Payable expense accruals for recurring software and utility vendors. Deliberately leave out revenue accruals, inventory, and complex multi-year contract amortization.

Cold Start Problem

The model requires labeled examples of missing accruals to learn vendor billing cadences. Break this by running historical backtests on 24 months of a design partner's closed General Ledger data to train the initial heuristics.

Time To First Value

1 full close cycle

Data Moat Available

true

Technical Difficulty

High

SAM

~$250-500M US-based lending, earned wage access, and B2B cashflow fintechs

SOM

~$10-25M

TAM

~20k global fintechs and embedded finance platforms × ~$50k-100k/yr API usage ≈ ~$1B-2B

Growth Rate

~20-25%/yr, driven by the expansion of embedded credit and demand for real-time underwriting visibility

Paid Comparable Spend

~$150k-300k/yr on internal data engineering labor, custom risk models, and raw bank data aggregation fees

Kill Thresholds

• Time-to-first-value > 45 days
• Prediction error rate > 5% after 30 days of live data
• Conversion from sandbox to paid contract < 20%
• CAC > $15k within first 90 days

Leading Metrics

• Time-to-first-API-call
• Prediction variance vs. settled cashflow
• Weekly active API calls per client
• Percentage of predictions auto-approved by risk models
• Sandbox-to-production conversion rate

What Proves Right

Fintech engineering teams route production ledger data through the API and retrieve predicted accrual values within 14 days of sandbox access. Cohorts processing more than 5,000 transactions per month maintain a 90 percent retention rate after 90 days. Customers convert from pilot to paid contracts at a 50,000 dollar annual minimum.

What Proves Wrong

Data science teams extract the raw transaction feeds to build in-house models rather than consuming the calculated accrual outputs. Integrations stall beyond 60 days because data formatting requires extensive transformation from legacy ERP schemas. The prediction variance against settled cash exceeds 5 percent, resulting in manual underwriter overrides and immediate churn.