How GreyScape.ai works
Shadow AI discovery and AI budget control — under one architecture.
GreyScape.ai sits beside your AI providers, not in front of them. We pull usage and cost data on a 15-minute schedule, attribute every dollar to the people and teams driving it, and give you the tools to govern it. Shadow AI discovery surfaces every unsanctioned tool in use across five independent signals. AI budget control sets right-sized caps with per-user attribution and a one-minute approval workflow for new workloads. Data leakage protection rounds out the trio. More on shadow AI discovery → More on AI budget control →
Why this exists
The three problems we tackle
Every IT and Finance team we talk to is losing sleep over the same three things. GreyScape.ai is built around them.
The AI bill is a mystery box
Providers bill by the word, in dollars, at the organisation level. No name attached. Finance can't allocate it; IT can't justify it.
Shadow AI is everywhere
Personal-tier signups on expense cards, browser-based agents that bypass network controls, AI features added to tools you already pay for. IT never sees most of it.
Data is leaking through prompts
When an engineer pastes a customer record into a chatbot to reformat it, that data is now on someone else's server outside your retention policy.
What you can do with it
Five capabilities, today and coming soon
Each capability is shipped as a built-and-honest feature today, with a clearly-labelled Coming soon section for what's next on the roadmap. Nothing is hidden behind a footnote.
See every dollar
One dashboard, every AI provider, with a 30-day trend and an end-of-month forecast.
Attribute it to real humans
Each dollar of AI spend mapped to a person, a team, and a project — no manual reconciliation.
Manage spend appropriately
Right-sized monthly allowances at four scopes. Soft warning at 75%, hard warning at 100%. Never a hard block — that's how production apps break.
Approve new AI work in a minute
An employee chats with an AI advisor for one minute. It scopes the work, recommends the cheapest sensible model, and estimates monthly cost. You approve and a key gets provisioned.
Protect against shadow AI and data leakage
Find AI tools your employees are using without approval, and stop company data leaking through prompts — at the network edge when you need it.
The approval flow, in detail
From idea to provisioned key, in about a minute
This is the workflow that turns GreyScape.ai from a dashboard into a system of record for AI use. Most enterprises spend weeks scoping each new AI workload; ours takes minutes — without losing the rigour.
- 1Admin opens /requests
Generates a one-use token, optionally pre-fills the requester's email and name. Hits 'Email this' — the admin's email client opens with subject and body pre-populated.
- 2Requester opens the link
Anonymous, no login required. Sees a short brief and starts chatting with the AI advisor.
- 3The advisor asks five questions
What are you trying to do? How often does it run? On what kind of data? What's the quality bar? Who else needs the output? The conversation takes ~1 minute.
- 4The advisor produces a structured recommendation
Recommended model (filtered by your approved-models policy), estimated input/output tokens per call, monthly cost at expected volume, and a list of cheaper alternatives with quality trade-offs.
- 5Requester submits with ROI note
A free-text 'why this is worth it' line — 'saves the support team ~6 hours a week', 'unblocks the marketing automation roadmap', etc.
- 6Admin reviews on /requests/[id]
Full transcript, model + cost, ROI note, requester's team and email. Approve, reject with a reason, or send back for re-scoping.
- 7Approval auto-provisions
A scoped project budget is created. For OpenAI, a service-account key. For Anthropic, a workspace + workspace-member key. The requester gets an email with everything they need to start.
- 8The workload appears on the dashboard
Spend, utilisation, and anomaly flags from day one — no separate setup. The audit log shows every step from request creation to key delivery.
What we see — and what we never see
Read-only on day one, by design
GreyScape.ai reads billing metadata from your providers. We do not read prompts, completions, or any user-facing AI content. The only conversational text we ever store is the scoping chat between an employee and the AI advisor.
What we see (today)
- Provider cost totals: $/day, $/model, $/project.
- Provider usage totals: input/output tokens, by model, by service-account or workspace.
- Identity attributes: user email, name, team — from your SSO directory.
- Approval-request data: the chat with the AI advisor, recommended model, monthly cost, ROI note, decision.
- Audit log: every admin action — who, what, when, from where.
What we don't see
- Prompts users send to providers in production traffic.
- Completions / responses the model returned.
- Embeddings, files, or images sent to providers.
- Passwords, API key plaintext (keys are encrypted at rest with AES-256-GCM and decrypted in memory only when calling the provider).
- Anything your provider doesn't expose on its Admin API.
See the full data inventory — every piece of data, whether it's mandatory or optional, and what functionality each piece unlocks — on the What we collect page.
Cadence and ops
What runs, when
GreyScape.ai is a Next.js app with a background scheduler. Here's the rhythm so there are no surprises.
Provider sync. Pull usage and cost from each connected provider, update budget utilisation, fire alerts if a threshold has been crossed.
The dashboard recomputes "What now?" suggestions from the latest sync, so the recommendations always reflect the freshest numbers.
CSV exports for chargeback. Approval-request key provisioning. Test email / Slack send from the Notifications settings page.
Who it's for
One product, two readers
GreyScape.ai is read by Finance and IT — sometimes the same person at a smaller company. The page each one cares about is different, but the numbers underneath are the same.
Finance / CFO office
A single defensible number for AI spend per month, attributed to cost centres, and a forecast that doesn't surprise the board.
- Dashboard — total spend, forecast vs budget, top spenders.
- Teams — % share by cost centre, monthly run-rate.
- Budgets — set monthly allowances; alerts before overshoot.
- Export CSV — drop into NetSuite / SAP / QuickBooks for chargeback.
IT / Security / Platform
Approved AI workloads with proper provenance, shadow AI surfaced before it's a habit, and a tamper-evident audit trail for the next compliance review.
- Providers — connect, sync, and rotate keys.
- Policies — set the approved-models allowlist.
- Requests — approve new AI work in minutes.
- Audit — full log of every admin action.
Next step