GreyScape.ai

For finance teams

Where is our AI spend actually going?

Provider invoices arrive at the organisation level. Personal subscriptions bury themselves in expense feeds. AI features quietly appear inside SaaS you already pay for. Add it up and most companies are 30–50% over their declared AI budget within the first six months. GreyScape gives finance teams per-user, per-team, per-project spend attribution — plus duplicate-sub discovery, budgets that warn instead of break, and chargeback-ready exports.

The four questions

The AI spend questions finance teams need to answer

How much did we spend on AI last month?

One number on the dashboard, with a 30-day trend and an end-of-month forecast. Provider invoices, expense-card AI charges, and inferred subscription spend all rolled into a single figure.

Where did the spike come from?

The Users and Teams views sort by month-to-date or by percentage change vs. last month. The cause of any spike is usually two clicks away — a runaway agent, a new model deployment, or one engineer's experimentation.

Who exactly is using each tool?

Provider usage is tagged to service accounts and workspace memberships, then mapped to your SSO directory by email. Per-person and per-team views with rolling 30-day spend.

Are we paying twice for the same thing?

Duplicate detection across providers and seat-based tools. Three teams expensing Cursor become a single row with the savings opportunity quantified. Personal-tier ChatGPT seats become a migration prompt to your corporate workspace.

Where the dollars hide

Five categories of hidden AI spend — with typical recovery

Averages below are from GreyScape pilot data. Your numbers will vary, but the categories don't. Each one has a clear discovery surface and a clear remediation path.

Duplicate subscriptions

Multiple teams paying for the same tool independently. Typical mid-market finding: 3–7 duplicates × $20–$200/seat/month.

✓ $1.2–$8k/month recoverable

Personal-tier seats outside SSO

ChatGPT Plus, Claude Pro, Cursor — paid personally and expensed at $20–$25/month each. Typical mid-market: 15–40 such seats.

✓ Migrate to a single workspace seat at the same cost — recovers governance and consolidates billing

Forgotten free-trial conversions

A trial signed up for a project, never cancelled, now billing for 18 months on autopay. The longest GreyScape has surfaced is 27 months at $99/month.

✓ $50–$3k/month per find

Runaway model spend

A workload moves from gpt-4o-mini to gpt-4o without a budget conversation. Cost goes 10×. No alert because there's no budget. Anomaly detection catches these within hours.

✓ Variable. Single runaway can be $500–$50k before someone notices.

Model-routing inefficiency

A team using GPT-4o for a task gpt-4o-mini handles 95% as well at one-tenth the cost. Smart routing recommendations surface these patterns automatically.

✓ 20–60% of variable LLM spend

Unmapped invoices

OpenAI bills $14,000 to the organisation. No way to know which team caused which dollar — until you connect the admin API and let GreyScape attribute it.

✓ Enables chargeback; recovers spend visibility entirely

How attribution works

Per-user, per-team, per-project — exportable as chargeback

Provider usage is tagged by service-account or workspace membership, then mapped to your SSO directory by email. The result is a clean ledger: who used what model, when, in which project, for how much. Export as CSV today; live NetSuite and QuickBooks push coming soon.

Budgets at four scopes

Organisation, team, person, project. Monthly periods. Soft alert at 75% utilisation, hard alert at 100%. Email and Slack notifications to the budget owner.

Approved workloads auto-budget

Every AI workload approved through the platform automatically creates a scoped project budget at the recommended monthly amount. No manual budget creation.

Chargeback exports

CSV with cost-centre, person, team, project, provider, model, tokens, and dollar amount. Drops straight into NetSuite or QuickBooks chargeback codes today; live API push coming soon.

See your savings in two minutes

The ROI calculator takes three inputs — headcount, declared monthly AI spend, and your best guess at tool count — and gives you a defensible range for what GreyScape would recover in the first 90 days. No sign-up required.

Start with one provider, scale to all of them

Paste an OpenAI or Anthropic admin API key and the dashboard starts populating within minutes. The first chargeback CSV is ready before month-end. From $3/user/month, 10-user minimum, 14-day refund guarantee.

Frequently asked

What finance teams ask first

How fast do we see the first attribution data?

About 10 minutes from the moment you paste a provider admin API key. The first sync pulls the trailing 30 days of usage, attributes it to users via your SSO directory, and the dashboard fills in. First chargeback CSV ready before month-end.

What if the AI providers don't expose attribution data?

They do — OpenAI exposes per-project and per-user usage via the Admin API; Anthropic exposes per-workspace-member usage. GreyScape calls those endpoints every 15 minutes and joins the usage to your SSO directory by email. For providers without admin attribution (some Azure deployments), we fall back to service-account-level tagging plus our own discovery surfaces.

Can we export to NetSuite or QuickBooks?

CSV export today, with the cost-centre, person, team, project, provider, model, and dollar amount columns laid out for direct paste into either system. Live API push to both is on the roadmap — already scoped, customer-led prioritisation.

What's the pricing — is it usage-based?

No. Linear per-user pricing: $3/user/month (or $2.50/user/month on annual), 10-user minimum, self-serve up to 999 users. We don't bill on AI tokens, calls, or dashboard views. Per licensed user, full stop.

What's the payback?

Most customers recover the GreyScape cost within the first 90 days from duplicate-subscription consolidation alone. Median pilot finding: 3-7 duplicate subscriptions at $20-$200/seat/month. The ROI calculator quantifies it for your headcount in two minutes.

Continue reading

Related pages