AI Cost Management (AICM)
AI Cost Management
AI Spend reads token counts, model usage and billing across AI provider consoles — live with customers today. Per-user, per-model, per-cost attribution; never prompt or completion content. The newest meter, on a platform built for meters.
What it covers
- AI-spend tracking for supported providers (including Anthropic, OpenAI, Google Gemini, M365 Copilot and GitHub Copilot): AI usage per user and per model with cost, in one view, at the level of usage data the provider exposes.
- Per-user, per-model and per-cost attribution, so consumption can be traced to an accountable owner rather than a single corporate card.
- Subscription seats, pooled allowance top-ups and application/API consumption separated, because each is governed by a different lever.
- Contract end-date tracking to support negotiation.
- Cost reporting at user and provider level.
The 20–40% AI-inference range
For AI/inference cost specifically, Alekstra works with an illustrative 20–40% savings range. It is an illustrative range, not a guarantee — actual savings are verified inside the engagement from your own reconciled data, and you approve every action that realises them.
Does your AI read our content?
For AI inference, Alekstra sees token counts (input and output), per-user and per-model attribution, and billing — via read-only vendor usage and billing keys. Prompt content, completions, embeddings, fine-tuning data and conversation history are part of a separately activated module that is not available in the EU.
Isn't this just a dashboard?
If a card platform already shows your AI token spend, keep it. The difference is that a dashboard shows spend; a ledger reconciles it — against the contract, to an accountable owner, across the rest of your technology cost, and into an approved GL posting. Alekstra treats AI as one meter in a twenty-year reconciliation discipline.
This describes the service in general terms; fixed price, timeline, service levels and limitations of liability are confirmed at order. Delivery requires the customer's cooperation — providing the necessary data, access and authorisations on the agreed schedule. Personal data is processed in the EU by default.
How AI spend forms
“AI spend” is usually three different pricing mechanics landing on one budget line. Per-seat subscriptions are a licence. Pooled top-ups are what you buy when a plan's allowance runs out mid-month. Application and API usage is priced per token and scales with what your software does, not with headcount. They are governed by entirely different levers, so seeing them apart is the first useful step.
Your numbers
Copilot, ChatGPT, Claude or similar, on a named-user plan.
Monthly list or contracted price per seat.
Extra capacity bought when a plan's included allowance runs out.
Token-priced consumption by your own applications and agents.
How the cost composes
€21,000per month
€252,000 per year
- Per-seat subscriptions€7,50036 %
- Pooled top-ups€1,5007 %
- Application / API usage€12,00057 %
What this shape tells you
- Most of this is consumption, not licences. That cost moves with model choice, prompt and context size, retries and caching — not with how many people you employ. Adding or removing seats will barely touch it.
- Top-ups are large relative to subscriptions. That usually means plans are sized below real use, so you are buying the same capacity twice at two different unit prices.
- Application and API usage typically arrives on one card or one team's account, regardless of who caused it. Until it is attributable to a service or owner, no one can act on it.
If you would like us to look at this with you, you can send these figures to our team. We will see the numbers you entered and which areas the calculator highlighted — nothing else, and no contact details unless you give them to us separately.
Illustrative and local to your browser: the sliders send nothing, and your figures stay on this device unless you press the button above. Defaults are placeholders, not benchmarks; replace them with your own figures. This shows how your spend is structured, not what you would save. Actual outcomes are established inside an engagement from your own data.
See it on your own numbers
The Alekstra audit is where most engagements start — a fixed-scope review of the costs, contracts and risks in the areas you choose. No obligation.