Allocex sits between your developers and frontier AI models, classifies every token against ASC 350-40 criteria in real time, and produces the audit-ready evidence packages your controller needs to move qualifying costs from the P&L to the balance sheet.
FASB's ASU 2025-06 replaced the old stage-based test with a principles-based framework that makes qualifying AI development costs capitalizable. But there is no defensible, audit-ready mechanism to act on it.
Replace your Anthropic, OpenAI, or Google endpoint with one environment variable. Single-digit millisecond overhead. Model choice, streaming, and tool use fully preserved.
Each call is enriched with project context from GitHub, Jira, and your IDE. A rules-based engine applies the ASU 2025-06 criteria and classifies every token as capex-eligible or opex.
Monthly close packages link every capitalized cost to an authorized project. Journal entries formatted for your ERP. A Big 4 co-authored methodology position paper covers the treatment with your auditors.
A company spending $30M annually on AI APIs, with 40% attributable to committed software development, can move $12M/year from opex to depreciable capex. At a 15× P/E multiple, that is $144M of enterprise value — for a policy change.
Assumes 5-year straight-line amortization. EV impact = annual opex reduction × P/E multiple.
We are working with a small number of enterprises — ideally a bank, an insurer, and a large retailer — who want to co-develop the methodology, work through their first audit cycle together, and set the standard for how this gets done. Free in year one, in exchange for collaboration.