AcademyLead

IT Economics: Cost, Value and AI Spend

How to trace technology spend to products and consumers, apply FinOps beyond public cloud, manage the cost drivers of generative AI, agree the accounting treatment, and report value that finance accepts.

L201Practitioner13 slidesAbout 12 minutes

Start the course Download the PDF
1 / 13
PDF
AxiareteAcademy

L201Lead · Practitioner

IT Economics: Cost, Value and AI Spend

Cost transparency, unit economics and value realization across cloud, SaaS, data cloud platforms, data centers and AI.

FormatKnowledge pack · 13 slides
ForIT and finance leaders
LevelPractitioner
Axiarete Academy · L20102 / 13

The problem

Technology cost is easy to total and hard to explain

IT spend is usually reported by ledger account. Saying what a business capability costs, who consumes it and what would change if it stopped requires an allocation model.

AllocationSpend traced to products, services and the people who consume them.
Unit costCost expressed per transaction, user or outcome.
ValueBenefits evidenced in terms finance recognizes.
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20103 / 13

Cost model

Trace every dollar from ledger to consumer

TBM Taxonomy 5.0.1 (TBM Council, July 2025) names these layers and adds an AI Models sub-tower.

Cost poolsGeneral-ledger spend by type: staffing, outside services, cloud services, hardware, software and SaaS.
Resource towersWhat the spend buys: compute, storage, network, data, applications, end user, security.
SolutionsBusiness-facing products and services; applications now sit in the towers.
ConsumersBusiness functions, value streams and external customers that use the solutions.
Why it mattersAllocation shows each consumer what its demand costs, so the people who create the demand can change it.
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20104 / 13

FinOps

FinOps now reaches well beyond public cloud

The FinOps Framework now gives guidance for five technology categories: public cloud, SaaS, data cloud platforms, data center and AI.

InformIngest billing data, ideally in FOCUS format, then allocate, forecast and benchmark unit costs.
OptimizeIdentify where rates, usage and workload placement can improve efficiency and value.
OperateAct on the chosen optimizations, with engineering, finance and business accountable.
PrinciplesTwo of six: “Business value drives technology decisions” and “Everyone takes ownership for their technology usage.”
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20105 / 13

Levers

Each cost lever has a different owner and risk

LeverWhat it changesTypical ownerMain risk
RatePrice per unit, through commitments and negotiated discountsFinOps team, procurementUnused commitments
UsageUnits consumed, through rightsizing, scheduling and removing idle resourcesEngineering teamsPerformance regressions
ArchitectureHow cost grows with volumeArchitectsMigration effort
PortfolioNumber of systems, through retirement and consolidationPortfolio governanceMissed dependencies
DemandConsumer behavior, through showback and policyBusiness partnersFriction with the business
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20106 / 13

Unit economics

Unit costs connect spend to volume and revenue

Prefer a unit the business already manages, so cost can be read alongside volume and revenue.

Cost per transactionRun cost of a product divided by the transactions it processes.
Cost per active userPlatform or SaaS cost divided by active users, by business unit or role.
Cost to serveTechnology cost per customer, account or segment.
Cost per inferenceTotal inference cost divided by the number of inference requests.
Run cost to revenueTechnology run cost as a share of the revenue a product earns.
Cost per environmentNon-production cost per environment, and its share of the total.
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20107 / 13

Generative AI

Generative AI has a different cost structure

Spend follows consumption at a fine grain, so small design choices multiply across millions of calls.

TokensOutput tokens, including reasoning tokens, typically cost several times more than input.
Context lengthInstructions, tool definitions and retrieved text are billed on every call; caching only discounts them.
Model choicePer-token prices vary more than tenfold and tokenizers count text differently; compare cost per task.
Agent loopsEach call, including retries, re-sends the growing history, so cost outpaces step count.
GPU capacityProvisioned capacity is billed whether used or not; on-demand GPUs can be scarce.
Supporting costsEmbeddings, vector storage, evaluation, logging and human review.
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20108 / 13

Controls

Control AI spend without slowing adoption

Route by taskSend each task to the least expensive model that meets the quality bar.
CacheCache stable prompt prefixes; reads can cost a tenth of the input rate, but writes can cost more.
Bound agentsLimit steps, tokens and time for every agent run.
BatchSend non-interactive work to batch APIs, typically billed at half the standard rate.
Measure per outcomeTrack cost per resolved task, not only per call.
TagAttribute every AI charge to a product, team and use case.
Re-test modelsTry cheaper models again as prices and capabilities change.
Set budgets and alertsPer product and environment, each with an owner and an agreed action at the limit.
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20109 / 13

Value

Report value in terms finance accepts

Value typeEvidence requiredCommon mistake
Run-rate cost removedInvoice stopped or asset decommissioned, net of any replacement run costCounting a retirement before the bill stops
Cost avoidedSpend in an approved budget or forecast that will no longer occurConstructing the baseline after the fact
Capacity releasedHours redeployed to named workReporting hours saved as cash savings
Risk reducedExposure removed, such as unsupported systems retiredPresenting avoided breaches as savings
Revenue enabledAn attribution method agreed with finance in advanceTaking full credit for shared outcomes
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20110 / 13

Accounting

Agree the accounting treatment before comparing options

Options with the same total cost can report differently, depending on what is capitalized and what is expensed.

SpendUsual treatmentWhat to check
Cloud, SaaS and AI usage feesOperating expense, as consumed or over the subscription termReplacing depreciated assets with fees lowers EBITDA
SaaS implementation, US GAAPQualifying costs capitalized, then expensed over the hosting term in the same line as the fees (ASU 2018-15)The term includes renewals reasonably certain to be exercised
SaaS configuration, IFRSUsually expensed as the supplier does the work, or over the access period (IFRIC agenda decision, 2021)Code the customer controls may be an intangible asset
Internal-use software, US GAAPFrom annual periods beginning after December 15, 2027 (ASU 2025-06): capitalized once funded and probable to completeSignificant development uncertainty delays capitalization
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20111 / 13

Pitfalls

Common errors in technology business cases

Counting license fees only
Include infrastructure, support, integration and labor
Double counting across programs
One register of benefits, owned by finance
Ignoring parallel running
Budget for old and new systems running together
Missing contract dates
Plan around notice periods, renewals and true-ups
Pricing AI from the pilot
Model cost at production volume and concurrency
Benefits with no owner
Name an owner and review each benefit quarterly
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20112 / 13

Getting started

Build cost transparency in three stages

Stage 1Visibility
  • Map the ledger to cost pools and towers
  • Tag cloud and AI spend; map SaaS contracts to products
  • Identify the 20 most expensive applications
Stage 2Accountability
  • Allocate cost to products and consumers
  • Publish showback every month
  • Introduce unit costs for three products
Stage 3Decisions
  • Budget and forecast by product
  • Require run cost in every investment case
  • Track benefits quarterly against the baseline
IT Economics: Cost, Value and AI SpendLead
Axiarete Academy · L20113 / 13

Summary

Key takeaways

  1. Trace spend from ledger to consumer, so each cost has an accountable owner.
  2. FinOps now spans public cloud, SaaS, data cloud platforms, data centers and AI.
  3. Tokens, context length, model choice and agent loops drive AI cost; design for them.
  4. Use unit costs the business already manages.
  5. Report only value that finance can verify, with an owner for every benefit.
Next course · U101Technology Portfolio IntelligenceUnderstand · Foundation
IT Economics: Cost, Value and AI SpendLead

Further reading

Continue learning

L101 · LeadThe AI-Native IT OrganizationU101 · UnderstandTechnology Portfolio IntelligenceAxiarete AcademyAll courses and tracks

Bring this course to your leadership team

We run private working sessions that apply these frameworks to your own estate, data and pending decisions.

Request a Private BriefingExplore the Academy