Assessment · Governance · Cost within the governance envelope

16Budgets, caps and spend guardrails

What stops a bad prompt or loop from running up the bill?

Monthly budgets, daily limits and per-run caps set where AI traffic flows.

AutomatedStakeholder interviews 2 controls 3 engines

How we collect

AutomatedSpend by team, workload and agent from gateway and billsRuns today
Stakeholder interviewsFinance: budgets and who gets the alertRuns today

Status labels show what runs today, what needs an export or setup from you, and what is planned.

What you get

Budget, quota, rate limit and circuit breaker told apart
Vendor cap gaps (Azure OpenAI has no simple hard budget limit)
Runs that would have breached a cap

Questions we ask

Who gets the alert when a budget is close?

We can start from

Budgets from your FinOps tool

Data sources

Source	What we read	Status
AI gateway telemetry	LiteLLM, Cloudflare AI Gateway or AICost CostWall logs: cost per team, workload, agent and outcome.	Partly automated
Questionnaire	The short interview for what data cannot show: purpose, owners, decisions AI informs, regulations in scope.	Runs today
OpenAI Usage and Costs API	Organization usage and cost by project and model, read with an Admin API key.	Wave 2
Anthropic Usage and Cost API	Token usage by model, workspace and API key, and daily cost, read with an Admin API key.	Wave 2

Built on governance engine modules

P6 LiveAI Cost Governance & FinOps

What does AI cost you, and who is accountable for it?

Value & Optimization

AICost governance controls in this module

AICost-GovControl502Runs today

Budgets, caps and spend guardrails

What it is

Monthly budgets, daily hard limits, per-run caps, retry and context limits, and approval before spend above a threshold.

Why it is needed

AI cost moves with usage. Without caps, one loop or one long-context workload can exceed a month of budget.

Why enterprises need it

Enterprises want predictable AI bills and the ability to stop runaway spend without stopping the business.

What you get:Predictable monthly AI spendRunaway workloads stopped automaticallyLimits enforced where traffic flows

Run by: AI Cost Governance and FinOps engineCost Firewall engine Budget Planner engine Agent Cost Reserve engine Retry Waste engine
Data: AI gateway telemetry, Questionnaire
Evidence produced: Guardrail table: workload, monthly budget, daily limit, per-run cap, enforcement point, last breach.
Maps to: FinOps Framework; NIST AI RMF: Manage

AICost-GovControl503Runs today

Cost anomaly and runaway agent detection

What it is

Detection of unusual spend: loops, retries, context growth, new models and sudden usage spikes, with an alert and an owner.

Why it is needed

Cost spikes are often the first sign of a broken agent, an abused key or a misconfigured workload.

Why enterprises need it

Teams want to hear about it the same day, not from the invoice next month.

What you get:Same-day alerts on spikesBroken agents found through costAbused keys spotted early

Run by: AI Cost Governance and FinOps engineAgent Loop engine Retry Waste engine Context Size Cost engine
Data: AI gateway telemetry, OpenAI Usage and Costs API, Anthropic Usage and Cost API
Evidence produced: Anomaly log: date, workload, expected cost, actual cost, cause, action, owner.
Maps to: FinOps Framework; NIST AI RMF: Measure

Frameworks


FinOps Framework and FOCUS	Cost visibility, allocation, budgets, forecasting and unit economics for AI

Stakeholders

CFOFinOps

← 15. Cost levers and model routing All modules 17. ROI and unit economics →

Vendor / Model	Field	Why it’s inferred
Anthropic — Claude Sonnet 4.6	`cachedInput`	Derived at 10% of input rate — Anthropic publishes 90% cache-hit discount on this tier.
Anthropic — Claude Sonnet 4.5	`cachedInput`	Derived at 10% of input rate; same 90% cache-hit convention as Sonnet 4.6.
Anthropic — Claude Sonnet 4.5	`batchInput`	Derived at 50% of standard input — Anthropic documents uniform 50% Batch discount.
Anthropic — Claude Sonnet 4.5	`batchOutput`	Derived at 50% of standard output — Anthropic documents uniform 50% Batch discount.
Anthropic — Claude Haiku 4.5	`cachedInput`	Derived at 10% of input rate — Anthropic 90% cache-hit discount convention.
OpenAI — GPT-5.4 Mini	`cachedInput`	Derived at 10% of input — OpenAI documents automatic 90% discount on cache hits across GPT-5.x tier.
OpenAI — GPT-5.4 Nano	`cachedInput`	Derived at 10% of input — OpenAI 90% cache-hit convention.
OpenAI — GPT-5.4 Nano	`batchInput`	Derived at 50% of input — OpenAI Batch API uniform 50% discount.
OpenAI — GPT-5.4 Nano	`batchOutput`	Derived at 50% of output — OpenAI Batch API uniform 50% discount.
OpenAI — GPT-5.4 Pro	`cachedInput`	Derived at 10% of input — OpenAI 90% cache-hit convention.
OpenAI — GPT-5.4 Pro	`batchInput`	Derived at 50% of input — OpenAI Batch API uniform 50% discount.
OpenAI — GPT-5.4 Pro	`batchOutput`	Derived at 50% of output — OpenAI Batch API uniform 50% discount.
OpenAI — GPT-5.2	`cachedInput`	Derived at 10% of input; no residency uplift.
OpenAI — GPT-5.2	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5.2	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5	`cachedInput`	Derived at 10% of input.
OpenAI — GPT-5	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5.5 Pro	`cachedInput`	Derived at 10% of input — OpenAI does not publish a cached rate for *-pro models; using the family convention.
OpenAI — GPT-5.5 Pro	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5.5 Pro	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5.2 Pro	`cachedInput`	Derived at 10% of input — pro-tier convention.
OpenAI — GPT-5.2 Pro	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5.2 Pro	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5.1	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5.1	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5 Pro	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5 Pro	`batchOutput`	Derived at 50% of output.
OpenAI — GPT-5 Nano	`cachedInput`	Derived at 10% of input.
OpenAI — GPT-5 Nano	`batchInput`	Derived at 50% of input.
OpenAI — GPT-5 Nano	`batchOutput`	Derived at 50% of output.
Google — Gemini 3 Flash	`cachedInput`	Derived at 10% of input — Google caching discount convention ~90%.
Google — Gemini 3.1 Flash-Lite	`cachedInput`	Derived at 10% of input — Google caching convention.
Google — Gemini 3.1 Flash-Lite	`batchInput`	Derived at 50% of input — Google Batch API uniform 50% discount.
Google — Gemini 3.1 Flash-Lite	`batchOutput`	Derived at 50% of output — Google Batch API uniform 50% discount.
Google — Gemini 2.5 Pro	`cachedInput`	Derived at 10% of input.
Google — Gemini 2.5 Flash	`cachedInput`	Derived at 10% of input.
Google — Gemini 2.5 Flash-Lite	`cachedInput`	Derived at 10% of input — Google caching convention.
Google — Gemini 2.5 Flash-Lite	`batchInput`	Derived at 50% of input — Google Batch API uniform 50% discount.
Google — Gemini 2.5 Flash-Lite	`batchOutput`	Derived at 50% of output — Google Batch API uniform 50% discount.
xAI — Grok 4 (legacy)	`cachedInput`	Extrapolated at 25% of base.

Vendor / Model

Field

Why it’s inferred

Anthropic — Claude Sonnet 4.6

cachedInput

Derived at 10% of input rate — Anthropic publishes 90% cache-hit discount on this tier.

Anthropic — Claude Sonnet 4.5

cachedInput

Derived at 10% of input rate; same 90% cache-hit convention as Sonnet 4.6.

Anthropic — Claude Sonnet 4.5

batchInput

Derived at 50% of standard input — Anthropic documents uniform 50% Batch discount.

Anthropic — Claude Sonnet 4.5

batchOutput

Derived at 50% of standard output — Anthropic documents uniform 50% Batch discount.

Anthropic — Claude Haiku 4.5

cachedInput

Derived at 10% of input rate — Anthropic 90% cache-hit discount convention.

OpenAI — GPT-5.4 Mini

cachedInput

Derived at 10% of input — OpenAI documents automatic 90% discount on cache hits across GPT-5.x tier.

OpenAI — GPT-5.4 Nano

cachedInput

Derived at 10% of input — OpenAI 90% cache-hit convention.

OpenAI — GPT-5.4 Nano

batchInput

Derived at 50% of input — OpenAI Batch API uniform 50% discount.

OpenAI — GPT-5.4 Nano

batchOutput

Derived at 50% of output — OpenAI Batch API uniform 50% discount.

OpenAI — GPT-5.4 Pro

cachedInput

Derived at 10% of input — OpenAI 90% cache-hit convention.

OpenAI — GPT-5.4 Pro

batchInput

Derived at 50% of input — OpenAI Batch API uniform 50% discount.

OpenAI — GPT-5.4 Pro

batchOutput

Derived at 50% of output — OpenAI Batch API uniform 50% discount.

OpenAI — GPT-5.2

cachedInput

Derived at 10% of input; no residency uplift.

OpenAI — GPT-5.2

batchInput

Derived at 50% of input.

OpenAI — GPT-5.2

batchOutput

Derived at 50% of output.

OpenAI — GPT-5

cachedInput

Derived at 10% of input.

OpenAI — GPT-5

batchInput

Derived at 50% of input.

OpenAI — GPT-5

batchOutput

Derived at 50% of output.

OpenAI — GPT-5.5 Pro

cachedInput

Derived at 10% of input — OpenAI does not publish a cached rate for *-pro models; using the family convention.

OpenAI — GPT-5.5 Pro

batchInput

Derived at 50% of input.

OpenAI — GPT-5.5 Pro

batchOutput

Derived at 50% of output.

OpenAI — GPT-5.2 Pro

cachedInput

Derived at 10% of input — pro-tier convention.

OpenAI — GPT-5.2 Pro

batchInput

Derived at 50% of input.

OpenAI — GPT-5.2 Pro

batchOutput

Derived at 50% of output.

OpenAI — GPT-5.1

batchInput

Derived at 50% of input.

OpenAI — GPT-5.1

batchOutput

Derived at 50% of output.

OpenAI — GPT-5 Pro

batchInput

Derived at 50% of input.

OpenAI — GPT-5 Pro

batchOutput

Derived at 50% of output.

OpenAI — GPT-5 Nano

cachedInput

Derived at 10% of input.

OpenAI — GPT-5 Nano

batchInput

Derived at 50% of input.

OpenAI — GPT-5 Nano

batchOutput

Derived at 50% of output.

Google — Gemini 3 Flash

cachedInput

Derived at 10% of input — Google caching discount convention ~90%.

Google — Gemini 3.1 Flash-Lite

cachedInput

Derived at 10% of input — Google caching convention.

Google — Gemini 3.1 Flash-Lite

batchInput

Derived at 50% of input — Google Batch API uniform 50% discount.

Google — Gemini 3.1 Flash-Lite

batchOutput

Derived at 50% of output — Google Batch API uniform 50% discount.

Google — Gemini 2.5 Pro

cachedInput

Derived at 10% of input.

Google — Gemini 2.5 Flash

cachedInput

Derived at 10% of input.

Google — Gemini 2.5 Flash-Lite

cachedInput

Derived at 10% of input — Google caching convention.

Google — Gemini 2.5 Flash-Lite

batchInput

Derived at 50% of input — Google Batch API uniform 50% discount.

Google — Gemini 2.5 Flash-Lite

batchOutput

Derived at 50% of output — Google Batch API uniform 50% discount.

xAI — Grok 4 (legacy)

cachedInput

Extrapolated at 25% of base.

16Budgets, caps and spend guardrails

How we collect

What you get

Questions we ask

We can start from

Data sources

Executable policy

Continuous checks

Engines that run it

Built on governance engine modules

AICost governance controls in this module

Frameworks

Stakeholders

Methodology

Primary sources

Inferred values (marked with * in calculator tables)

Immediate steps you can take