Back to Assessments & Documents
Technology Finance and Assets Active

AI Spend Management Maturity Assessment

A 68-question diagnostic across 9 domains that follows AI cost from governance and the operating model through usage metering, budgeting and allocation, AI service cost models, model selection and placement, token and in...

Start Assessment

Overview

A 68-question diagnostic across 9 domains that follows AI cost from governance and the operating model through usage metering, budgeting and allocation, AI service cost models, model selection and placement, token and inference efficiency, compute cost, vendor, licensing and product economics, and AI FinOps performance. Written for the CFO, CTO and head of AI FinOps who need to answer: do we know what each AI service, agent and model costs us and what value it produces, or are we paying invoices without control?

Detailed Description

AI is one of the fastest-growing costs in enterprise technology, and it often grows without the controls applied to other large expenses. Model API calls, agents, GPU capacity, retrieval systems, AI assistant seats and SaaS AI credits are bought by many teams, billed in units few people understand and rarely tied to the value they produce.

This assessment looks at the whole system for managing AI spend. It starts with the spend strategy, ownership, governance forum, spending policies and cost literacy. It then examines whether every AI provider, call and agent is inventoried, metered through a gateway and attributed, including multimodal usage. It looks at how AI spend is budgeted, forecast, allocated, capped at the provider and limited for experiments, and how it is investigated when it spikes. A dedicated domain tests whether each AI service has a standing cost model and whether that model feeds catalog rates, costed service levels and the technology budget. It then covers the economics of choosing, routing and placing models, the engineering practices that cut token and inference waste, including reasoning budgets, provider pricing features and protection against denial-of-wallet abuse, and the governance of GPU, retrieval, idle and energy costs. It then looks at commitments, invoices, seats, SaaS AI credits, agent task economics, product return and customer margin. It closes with AI FinOps measures, executive reporting, benchmarking and a savings register.

How AI spend joins the wider FinOps practice is covered by the IT FinOps and Cloud Financial Management Maturity Assessment; this model covers the AI-specific mechanics in depth.

Each question is scored on three dimensions: Documented, Implemented and Effective. This exposes practices that exist on paper but are not followed, work that depends on one experienced person, and practices that run but are never measured.

Assessment Details

Audience

CFO / CTO / Head of AI FinOps / VP of AI / Head of AI Platform. Secondary: AI product managers, finance business partners and IT finance, service catalog and service level managers, procurement and vendor managers, AI and platform engineering leads, sustainability leads.

Purpose

Finds unmetered and unattributed AI spend, including agents, multimodal use and SaaS AI credits, uncapped keys, AI services with no unit cost, catalog prices and service levels set without cost, token, compute and idle waste, unused commitments and seats, and AI products and agents with no measured return. Gives each gap a concrete next step and a measurable target, so AI cost moves from invoice surprise to managed, data-driven decisions.

Effort

Initial assessment: 5-7 hours of facilitated sessions with finance and FinOps, AI engineering and platform, AI product management, procurement and an executive sponsor, with input from service catalog, service level and IT budget owners for the cost model domain. Evidence gathering: 2-3 weeks (metering and usage data, cost models, catalog and SLA records, budgets, contracts and invoices). Reassessment: 2-3 hours.

Cadence

Every six months, with a quarterly check of the KPIs, and within 30 days of a major new AI provider or agent platform, a significant pricing change, a material spend overrun or an abuse incident on a public AI feature.

Framework Alignment

FinOps Framework 2026 FOCUS 1.4 (FOCUS 1.5, planned for late 2026, is expected to add AI token usage data) COBIT 2019 NIST AI RMF 1.0 ISO/IEC 42001:2023 ISO/IEC 19770-1:2017 OWASP Top 10 for LLM Applications 2025 OWASP Top 10 for Agentic Applications 2026 ISO/IEC 21031:2024 GHG Protocol Corporate Value Chain (Scope 3) Standard

KPIs

  • AI Spend Visibility - Share of total AI spend, including agents, multimodal use and SaaS AI credits, captured through metering and attributed to an application and business unit. Target: 95% or higher.
  • AI Forecast Accuracy - Variance between forecast and actual monthly AI spend. Target: within 10%.
  • Unit Cost Trend - Change in cost per business transaction for AI services with a current cost model. Target: falling quarter over quarter.
  • Commitment and Seat Utilization - Share of committed AI spend and paid AI seats actually used. Target: 90% of commitments, 85% of seats.
  • AI Product Return - Share of AI products and production agents with measured return against a baseline in the last 12 months. Target: 100%.

Dependencies

IT FinOps and Cloud Financial Management (v5.9.6): provides the cost data, allocation, commitment and reporting foundations and covers how AI spend joins the FinOps practice; this model covers AI-specific mechanics in depth. Technology Service Catalog: owns the catalog where AI offering rates from the AI Service Cost Model are published. Service Level Management: owns the SLA standard and template whose AI section requires costed tiers before sign-off. IT Budget: owns the technology budget plan that AI cost model drivers feed. AI Governance (companion): sets AI policy and risk tiers and covers AI contract risk terms, which are not repeated here. AI Pipeline and MLOps (companion): covers model compression, fallback and serving reliability; this model assesses their cost and value angle only. Recommended: Run the IT FinOps and Cloud Financial Management assessment alongside this one, and the AI Governance assessment where AI risk is material.

Assessment Outline

(9 domains · 68 questions)
  1. 1

    1 AI Spend Governance and Operating Model

    AI spend strategy, accountable ownership and operating model, investment forum, spending policies, total cost evaluation and AI cost literacy

    • 1.1 AI Spend Strategy 1 questions
    • 1.2 AI Spend Ownership and Operating Model 1 questions
    • 1.3 AI Investment Governance Forum 1 questions
    • 1.4 AI Spending Policies 1 questions
    • 1.5 Total Cost Evaluation 1 questions
    • 1.6 AI Cost Literacy 1 questions
  2. 2

    2 AI Usage Visibility and Metering

    Provider and model inventory, central gateway, token and cost attribution, shadow AI, token mix, unit economics, dashboards, engineer cost visibility, agent cost attribution and multimodal usage

    • 2.01 AI Provider and Model Inventory 1 questions
    • 2.02 Centralized AI Gateway 1 questions
    • 2.03 Token and Cost Attribution 1 questions
    • 2.04 Shadow AI Detection 1 questions
    • 2.05 Token Mix Analysis 1 questions
    • 2.06 Unit Economics per Transaction 1 questions
    • 2.07 AI Spend Dashboards 1 questions
    • 2.08 Engineer Cost Visibility 1 questions
    • 2.09 Agent Cost Attribution 1 questions
    • 2.10 Multimodal Usage and Cost 1 questions
  3. 3

    3 Budgeting, Forecasting and Cost Allocation

    AI budget lines, forecasting, scenarios, allocation, quotas and rate limits, anomalies, overruns, showback, provider spend caps and experimentation spend

    • 3.01 AI Budget Integration 1 questions
    • 3.02 AI Spend Forecasting 1 questions
    • 3.03 AI Spend Scenario Planning 1 questions
    • 3.04 AI Cost Allocation Model 1 questions
    • 3.05 Spending Quotas and Rate Limits 1 questions
    • 3.06 AI Spend Anomaly Detection 1 questions
    • 3.07 Overrun Management 1 questions
    • 3.08 AI Showback and Chargeback 1 questions
    • 3.09 Provider Spend Caps and Key Limits 1 questions
    • 3.10 Experimentation and Evaluation Spend 1 questions
  4. 4

    4 AI Cost Models and Service Integration

    A standing AI service cost model and how it feeds service catalog pricing, costed service levels, technology budget planning and customer pricing and margin

    • 4.1 AI Service Cost Model 1 questions
    • 4.2 AI Service Catalog Pricing 1 questions
    • 4.3 Costed AI Service Levels 1 questions
    • 4.4 AI Cost Drivers in Technology Budget Planning 1 questions
    • 4.5 Customer Pricing and Margin for AI Features 1 questions
  5. 5

    5 Model Selection and Routing Economics

    Model sourcing, dynamic routing, right-sizing, upgrade governance, cost-quality trade-offs and AI placement and sovereign hosting economics

    • 5.1 Model Sourcing Economics 1 questions
    • 5.2 Dynamic Model Routing 1 questions
    • 5.3 Model Right-Sizing 1 questions
    • 5.4 Model Upgrade Governance 1 questions
    • 5.5 Cost-Quality Trade-off Decisions 1 questions
    • 5.6 AI Placement and Sovereign Hosting Economics 1 questions
  6. 6

    6 Token and Inference Efficiency

    Prompt and context compression, caching, conversation memory, output limits, agent guardrails, prompt template review, reasoning budgets, provider pricing features and denial-of-wallet protection

    • 6.1 Prompt and Context Compression 1 questions
    • 6.2 Semantic Caching 1 questions
    • 6.3 Conversation Memory Management 1 questions
    • 6.4 Output Length Governance 1 questions
    • 6.5 Agent Loop and Runaway Guardrails 1 questions
    • 6.6 Prompt Template Review 1 questions
    • 6.7 Reasoning Effort and Thinking Budgets 1 questions
    • 6.8 Provider Pricing Features 1 questions
    • 6.9 Denial-of-Wallet Protection 1 questions
  7. 7

    7 Compute and Infrastructure Cost Management

    GPU provisioning, auto-scaling, spot capacity, accelerator utilization, training cost, retrieval and embedding costs, idle AI deployments and AI energy and carbon

    • 7.1 GPU Provisioning Governance 1 questions
    • 7.2 GPU Auto-Scaling 1 questions
    • 7.3 Spot and Interruptible Capacity 1 questions
    • 7.4 Accelerator Utilization Monitoring 1 questions
    • 7.5 Training Cost Governance 1 questions
    • 7.6 Retrieval and Embedding Cost Management 1 questions
    • 7.7 Idle AI Deployments and Endpoints 1 questions
    • 7.8 AI Energy and Carbon 1 questions
  8. 8

    8 Vendor, Licensing and Product Economics

    AI commitments, multi-vendor strategy, pricing changes, invoice validation, seats, product return, portfolio review, SaaS AI consumption credits and agent task economics

    • 8.01 AI Commitments and Enterprise Agreements 1 questions
    • 8.02 Multi-Vendor Strategy 1 questions
    • 8.03 Provider Pricing Change Monitoring 1 questions
    • 8.04 AI Invoice Validation 1 questions
    • 8.05 AI Seat and Subscription Licenses 1 questions
    • 8.06 AI Product Return on Investment 1 questions
    • 8.07 AI Product Portfolio Review 1 questions
    • 8.08 SaaS AI Consumption Credits 1 questions
    • 8.09 Agent Task Economics 1 questions
  9. 9

    9 AI FinOps Performance and Improvement

    KPIs, executive reporting, cross-functional review, benchmarking and the optimization register

    • 9.1 AI FinOps KPIs 1 questions
    • 9.2 Executive AI Spend Reporting 1 questions
    • 9.3 Cross-Functional AI Spend Review 1 questions
    • 9.4 AI Cost Benchmarking 1 questions
    • 9.5 AI Spend Optimization Register 1 questions

At a Glance

Category Technology Finance and Assets
Type Free
Roles 6
Domains 9
Questions 68
Start Assessment

Still have questions? Get in touch

ITSGovernance
  • Twitter

Contact us

ITSGovernance