Wedge

Start by auditing sales team travel and entertainment expenses for mid-market software companies. Sales teams generate the highest volume of complex, high-dollar receipts with high rates of policy violations, providing fast proof of value. Once embedded as the T&E approval layer, expand into auditing corporate card software subscriptions and procurement vendor invoices.

Timing

Multimodal LLMs now reliably extract structured line-item data from crumpled, poorly lit receipt photos and map them against dense, unstructured corporate expense policy documents in seconds. Earlier OCR solutions failed to understand context, requiring humans to verify the output.

Why This ICP

Mid-market finance teams process enough transaction volume to experience painful financial leakage but lack the massive internal accounts payable compliance teams of Fortune 500s. They adopt tools that offer immediate hard-dollar ROI through recovered cash.

Size Of Prize

~35,000 mid-market and enterprise companies in the US spend roughly $40,000 annually on offshore T&E auditors and absorb unrecovered leakage, creating a $1.4B addressable prize.

Gap Narrative

Finance teams currently sample a small fraction of employee expense reports because manual review is slow, allowing out-of-policy spend and duplicate receipts to slip through. This opportunity provides an autonomous auditor that parses every receipt line item against complex corporate travel policies without human intervention.

Defensibility

Defensibility builds through workflow lock-in as the system embeds directly into the company ERP and corporate card infrastructure to auto-approve or reject spend. The extraction models also compound in accuracy by learning company-specific vendor behaviors and edge-case policy interpretations over thousands of transactions.

Why This Thesis

Service-as-Software replaces the outsourced BPO auditor entirely. It directly executes the approval or rejection workflow in the expense management system rather than just flagging anomalies for a human accountant to review.

Hardest Part

Achieving near-zero false positive rates on policy violations while accurately parsing messy, unstructured receipts against dense, company-specific expense rules.

Min Viable Scope

A read-only integration with a single expense platform that flags anomalous travel and entertainment expenses for human review post-reimbursement. Leave out active approval blocking, corporate card issuance, and recurring vendor invoice matching.

Cold Start Problem

Models lack context on company-specific expense norms and unwritten rules without historical data. Break this by running historically approved shadow-audits on a design partner's past 12 months of data to establish a baseline of acceptable behavior.

Time To First Value

2-3 days of historical ingestion and policy rule mapping

Data Moat Available

true

Technical Difficulty

High

SAM

~$600M-800M representing ~30k-40k mid-to-large legal, consulting, and accounting firms

SOM

~$15M-30M

TAM

~150k US professional services and mid-market enterprises × ~$15k-25k/yr ≈ ~$2.2B-3.7B

Growth Rate

~12-18%/yr, driven by rising travel volumes and the high cost of manual compliance reviews in billable-hour organizations

Paid Comparable Spend

~$45k-90k/yr in fractional or dedicated AP headcount for manual receipt verification and compliance checks, plus legacy expense routing software fees

Kill Thresholds

• Auto-approval rate under 50 percent after 30 days of live usage
• False positive flag rate exceeds 15 percent
• Time-to-first-value exceeds 21 days
• Pilot to paid conversion under 25 percent at $15k ACV
• Customer acquisition cost exceeds $8k in the first 90 days

Leading Metrics

• Auto-approval rate for routine expenses
• False positive out-of-policy flag percentage
• Days from ERP connection to first automated audit run
• Weekly finance hours spent resolving flagged exceptions
• Percentage of flagged expenses requiring manual override

What Proves Right

Mid-market professional services firms deploy the tool and automatically approve over 80 percent of routine expenses without human intervention. Finance teams reduce their monthly close cycle by at least two days because the system catches out-of-policy spend at the point of submission. Customers sign standard $15k annual contracts after a 30-day pilot without demanding bespoke ERP integrations.

What Proves Wrong

Finance controllers refuse to trust the automated decisions and manually review all receipts regardless of the system score. The false-positive rejection rate exceeds 20 percent, forcing finance teams to spend more time resolving employee disputes than they previously spent doing manual audits. Implementation stalls beyond 45 days because mid-market firms require legacy on-premise accounting integrations.