AssessmentGovernance

AICost AI Governance

AI governance in 20 modules that run as policy

Cost and governance controls for every AI workflow, compiled into policy your gateway and cloud enforce, and checked every day. Pick a module to see how it runs.

1Automated collection and reports

Read-only connections to Microsoft Entra and Graph and to AWS cost data, plus bill, licence and gateway exports, collect the facts and build the reports.

2Stakeholder interviews

We consult the people who own the work for what data cannot show: purpose, decisions, risk appetite. Their decisions are recorded with name and date.

3Workflows, priced and governed

126 cost and governance engines price every workflow step and keep it inside the governance envelope.

4Executable policy

Cost and governance rules compiled into policy your own gateway and cloud enforce. Approved decisions become the rules the checks run against.

5Continuous governance

Watchers and an evidence journal check cost, vendors, models and policy on a schedule, so drift and new AI show up before the bill or the audit.

Already assessed or governed? We start from what you have. Today that means your registers, assessments and audits as spreadsheet or CSV files. Direct imports from ServiceNow, OneTrust and Credo AI are planned. AICost adds cost control and executable policy across every workflow.

Assess and know

1AI inventory and shadow AI

Every AI app, agent, model, key and AI feature in use, with the spend behind it.

Runs asApproved-app list loaded into the gateway allowlist P1P15 AutomatedStakeholder interviews · 3/5 runs today · 2 controls 2Workflow map and owners

Each AI system mapped to the business workflow it serves, its steps and a named owner.

Runs asOwner and purpose required before a workflow is approved in the gateway P10P3 AutomatedStakeholder interviewsUses your existing assessments · 2/4 runs today · 3 controls 3Use-case value and priorities

Which AI workflows are worth funding, scaling or stopping, ranked by value and risk.

Runs asFunding gates: a workflow scales only above its value threshold P3P16 Stakeholder interviewsAutomatedUses your existing assessments · 3/4 runs today · 2 controls 4Risk tiering and risk register

Every AI system gets a risk tier and sits in one register mapped across frameworks.

Runs asControls switched on by tier: high tier gets review, evals and caps P9P19 AutomatedStakeholder interviewsUses your existing assessments · 2/4 runs today · 2 controls 5Regulatory applicability

Which rules apply to each AI system by country, sector, role and use, with effective dates.

Runs asJurisdiction and model-origin rules compiled into the allowlist P9P19 Stakeholder interviewsAutomatedUses your existing assessments · 3/4 runs today · 3 controls 6Data sensitivity, privacy and residency

What data each AI system can reach, where it is processed and on what terms.

Runs asRedaction, region and model-origin rules per data class P14P5P8 AutomatedStakeholder interviews · 4/4 runs today · 4 controls 7Identity, access and agent permissions

Who and what can call AI, and what every agent is allowed to read, write or approve.

Runs asLeast-privilege scopes and tool allowlists per agent P7P2 AutomatedStakeholder interviews · 3/3 runs today · 3 controls 8MCP servers, tools and AI supply chain

Every MCP server, plugin, model source and library an AI system depends on.

Runs asApproved MCP server and tool list enforced at the gateway P4P2 AutomatedStakeholder interviews · 2/4 runs today · 2 controls 9Vendor and model due diligence

AI vendors and models checked for terms, concentration, price and fit before and after you buy.

Runs asApproved vendor and model list with backup routes P14P11 AutomatedStakeholder interviews · 3/3 runs today · 3 controls

Test and control

10Answer quality and grounding

Answers checked for made-up facts and kept to the data you give the model.

Runs asGrounded-only mode: answer from supplied data or say not found P12 Stakeholder interviewsAutomatedUses your existing assessments · 2/4 runs today · 2 controls 11Security and red teaming

Prompt injection, data leakage, jailbreaks and excessive agency tested and screened.

Runs asInjection and leakage screening on inputs, documents and tool results P2 AutomatedStakeholder interviewsUses your existing assessments · 2/4 runs today · 1 control 12Fairness, explainability and human oversight

Reviews that catch real errors, bias tests where decisions affect people, and decisions you can explain.

Runs asReview gates by risk tier and data class P12P19 AutomatedStakeholder interviewsUses your existing assessments · 1/3 runs today · 2 controls 13Agent safety and runaway control

Loops, retries and long agent runs capped before they cost money or cause damage.

Runs asTurn, token, time and cost caps per run P4P6 AutomatedStakeholder interviews · 3/3 runs today · 2 controls

Cost within the governance envelope

14Workflow cost and TCO

Every step of every workflow priced: models, retrieval, tools, review, compliance and run cost.

Runs asBudget envelope per workflow from its priced baseline P6P13 AutomatedStakeholder interviews · 3/4 runs today · 2 controls 15Cost levers and model routing

Caching, routing, batching and compaction applied where they are safe for the data and the task.

Runs asRouting, cache and batch rules compiled for your gateway P6P11 AutomatedStakeholder interviews · 2/2 runs today · 1 control 16Budgets, caps and spend guardrails

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

Runs asCost firewall: monthly budget, daily warning, hard stop, per-run cap P6 AutomatedStakeholder interviews · 2/2 runs today · 2 controls 17ROI and unit economics

Value and cost per resolved case, document or task, and seats that are actually used.

Runs asScale only workflows above their ROI threshold P3P16P15 AutomatedStakeholder interviews · 2/3 runs today · 2 controls

Run continuously

18Executable policy

Cost and governance rules compiled into policy your own gateway or cloud enforces. AICost stays out of the request path unless you choose the enforcing proxy.

Runs asOne policy compiled for LiteLLM, Cloudflare AI Gateway or Amazon Bedrock model access P6P14 AutomatedStakeholder interviewsUses your existing assessments · 1/3 runs today · 3 controls 19Continuous monitoring and incidents

Cost, quality, drift and policy checked every day, with incidents handled to closure.

Runs asIncident playbooks tied to the controls that failed P18P6 AutomatedStakeholder interviews · 2/3 runs today · 3 controls 20Evidence, audit and executive reporting

A tamper-evident evidence journal, audit packs and board reporting from the same data.

Runs asHash-chained evidence journal for every control P9P20 AutomatedStakeholder interviewsUses your existing assessments · 2/3 runs today · 7 controls

How governance runs

Set the envelope

Budgets, allowlists, review gates and caps per workflow.

Compile

One policy for your gateway or cloud.

Enforce

Your gateway and cloud apply it; AICost never sits in the request path.

Watch

Daily checks on cost, quality, drift and new AI.

Prove

Evidence journal, audit packs and board reports.

Talk to AICost about your AI governance

We run the basic assessment on your tenant and invoices and walk you through the findings, the spend and the first controls to switch on.

Book a governance call Price and govern a workflow Full guide →

Assessment → · 20 modules · 43 controls · 126 engines

📖 Data sources & methodology 172 text models · 9 embeddings · 44 vision · 62 audio · 8 vector DBs across 10 vendor pages · last verified 2026-10-18

Methodology

  • All prices are USD per 1 million tokens, current as of 2026-10-18.
  • Vendor-published values have no mark. Inferred/extrapolated values are marked with * and listed below.
  • Batch API discounts are 50% off standard rates across providers that offer Batch mode.
  • Prompt caching discounts vary by provider (typically 80-90% off cached input tokens).
  • Regional data-residency surcharges (Anthropic 1.1x, OpenAI 1.1x, Google regional tiers) are NOT included in base rates.
  • Long-context pricing tiers apply when input exceeds model threshold.
  • Embedding prices are input-only (no output tokens generated).

Primary sources

Last-verified date is the most recent successful daily snapshot (aicost_pricing_snapshots) or, when no snapshot exists yet, the latest successful crawler run (aicost_crawler_runs). 10 of 10 vendors are currently verified. Aggregator services (TokenCost, AI Pricing Guru, etc.) are not listed.

Anthropic
2026-10-18
https://www.anthropic.com/pricing
Daily snapshot since Sep 2023 · 704 days captured
Anthropic Docs
2026-10-18
https://platform.claude.com/docs/en/about-claude/pricing
Daily snapshot since Sep 2023 · 704 days captured
OpenAI
2026-10-18
https://openai.com/api/pricing/
Daily snapshot since Sep 2023 · 705 days captured
Google AI
2026-10-18
https://ai.google.dev/gemini-api/docs/pricing
Daily snapshot since Dec 2023 · 680 days captured
Google Vertex
2026-10-18
https://cloud.google.com/vertex-ai/generative-ai/pricing
Daily snapshot since Dec 2023 · 680 days captured
DeepSeek
2026-10-18
https://api-docs.deepseek.com/quick_start/pricing
Daily snapshot since May 2024 · 619 days captured
xAI
2026-10-18
https://x.ai/api
Daily snapshot since Nov 2024 · 537 days captured
Mistral
2026-10-18
https://mistral.ai/pricing
Daily snapshot since Dec 2023 · 678 days captured
Cohere
2026-10-18
https://cohere.com/pricing
Daily snapshot since Sep 2023 · 704 days captured
Voyage AI
2026-10-18
https://docs.voyageai.com/docs/pricing

Inferred values (marked with * in calculator tables)

Derived from industry conventions, not directly published by the vendor. Typical conventions: cached input = 10% of base (90% off), Batch API = 50% of base (50% off).

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.

Pricing is cross-verified against the LiteLLM community registry when available. Daily snapshots are kept in aicost_pricing_snapshots; every change is logged to aicost_price_changelog with old & new values for full audit trail. Read the full methodology →