Consolidated AI-tool spend and usage governance for finance and IT, covering the per-seat and per-credit AI subscriptions that never touch a gateway
A June 2 HN thread on Uber capping employee AI spending after blowing through its budget in four months surfaced a recurring B2B pain: finance and IT have no consolidated view of what employees spend on AI tools, and what visibility exists is reactive cost-capping after the bill lands. The LLM-gateway tools (LiteLLM, TrueFoundry, Bifrost) only see API traffic you deliberately route through them. They are blind to the way most AI spend actually happens at companies now: individual Cursor, ChatGPT, Claude, Copilot, and v0 seats bought on cards or expensed, each with its own opaque usage-credit burn. The opportunity is a spend-and-policy layer aimed at the CFO and IT admin that pulls these vendor subscriptions and credit systems into one dashboard with per-team attribution and enforceable caps.
Do not build another LLM gateway. The whole point is the spend that never routes through a gateway. Win on ingestion breadth (vendor billing APIs, card feeds, SSO-based seat discovery) and on giving a CFO a per-team number with an enforceable cap, not on dashboards engineers already have. The buyer is finance and IT, so the product has to make sense to someone who has never heard the word 'token'.
landscape (3 existing solutions)
There are mature tools for routed-API LLM cost control and for generic SaaS subscription inventory, but nothing sits in the middle and gives finance one reconciled, per-team picture of AI tool spend across vendor seats and usage credits with hard caps. The 'track employee AI token usage' framing is now an emerging FinOps category, which signals the demand is real and not yet owned.