AI GOVERNANCE & COST VISIBILITY
Can you say who used your generative AI spend — and for what?
Multiple AI APIs, multiple departments, and an invoice that shows only a total. Without a breakdown you cannot measure ROI, promote AI where it works, or cut wasteful usage.
Early Adopter ProgramNo setup fee · $500/month (plus applicable taxes) | First 5 companies | Until Nov 30, 2026
SUPPORTED PROVIDERS
- OpenAI
- Azure OpenAI
- Claude (Anthropic)
- AWS Bedrock
- Google Vertex AI
The more you use AI, the more its costs become a black box
- 01
Rising, scattered costs
Departments adopt API usage across multiple providers at the same time, and costs keep climbing. No one has the full picture.
- 02
No breakdown
The invoice shows only a total. It cannot tell you who, which department, or which project used it.
- 03
No way to act
Without ROI, you cannot promote AI to the teams getting results or rein in wasteful usage. You find out about budget overruns only after the invoice arrives.
How many of these apply to you?
- You cannot say right away what each department spends on generative AI per month
- There has been a month when the bill exceeded expectations
- There is no register of who holds which API keys
- Leadership asked about "AI ROI" and you had no answer
- You have AI usage guidelines, but no view of actual usage
If three or more apply, see the solution below.
ManageAI is not a blocking or monitoring tool. It supports governance by making costs and usage visible, without slowing down AI adoption on the ground.
See who is spending how much, and on what — by department
- 1
Visibility across five providers
Usage and spend for OpenAI, Azure OpenAI, Claude, AWS Bedrock and Google Vertex AI, by department, user, project and API key, on a single screen.
- 2
Three alert types, before the invoice
Detects daily cost caps, monthly budget consumption and usage spikes, and notifies you via Slack and email.
- 3
Ready for executive reporting
CSV export and a JPY/USD toggle. Actual and estimated costs are always shown separately.
See “for what” too: usage by purpose
Patent pendingWe plan to offer a feature that visualizes usage by purpose as statistics, without any prompt text ever leaving your environment.
See it in the product
Cost, tokens, requests and share for each department, with a daily trend showing who used what, and when.
* Screens show demo data.
* Actual costs are reconciled against provider invoices. The difference between actual and estimated costs was 1–1.8% (measured on Azure and AWS Bedrock).
Lightweight to adopt, built to pass security reviews
Just register read-only keys. No system changes or resident software required.
No impact on your AI usage. ManageAI is not in the request path, so your AI usage continues even if ManageAI stops.
Prompt text is never collected, by design. We handle usage metadata only; there is no mechanism to store the text at all.
Private database. Hosted on AWS (Tokyo region), with a database that cannot be reached directly from the internet.
Frequently asked questions
- Q.
Does ManageAI receive our prompts?
- A.
No. We collect only usage metadata (token counts, costs, model names and key identifiers). Prompts and generated output are never collected or stored. We only read each provider’s usage statistics APIs, which return aggregated values, so by design we cannot access the text.
- Q.
What does setup involve? Do we need to change our systems?
- A.
You issue read-only credentials in each provider’s console and register them in ManageAI. No system changes, agents or network changes are required. Visibility starts on the day you register your keys, and actual costs appear from the next day onward. Including organization mapping and alert thresholds, it typically takes one to two weeks to be fully operational.
- Q.
Which AI providers are supported? Can you see ChatGPT subscriptions?
- A.
Five providers are supported: OpenAI, Azure OpenAI, Claude (Anthropic), AWS Bedrock and Google Vertex AI, all verified in real environments. Coverage is pay-as-you-go API usage; seat fees for subscriptions such as ChatGPT Team or Claude Pro are not included. Granularity varies by provider: per API key for OpenAI and Claude, per deployment for Azure, and per model for Bedrock and Vertex AI.
- Q.
How accurate are per-user costs?
- A.
Providers report actual costs only down to the project (workspace) level, so per-user and per-key costs are estimates calculated from price tables. The screen always shows “actual” and “estimated” separately and never mixes them. Organization-wide and per-project figures are actual costs that reconcile with your invoices.
- Q.
If ManageAI goes down, does it affect our AI usage?
- A.
No. ManageAI does not sit between you and your AI providers; it only reads management APIs. If the service stops, the impact is limited to visibility and alerts. You can export your data as CSV at any time.
- Q.
How do alerts work? We don’t want endless notifications.
- A.
There are three rule types: daily cost cap exceeded, monthly budget consumption, and recent spike detection. Notifications go to Slack and email, with a cooldown so the same alert does not fire repeatedly. We set thresholds together during onboarding based on your actual usage, and tune them by recording each alert as a true or false positive.
- Q.
How much does it cost?
- A.
Through our Early Adopter Program, open until the end of November 2026, there is no setup fee and the price is $500/month (plus applicable taxes). If you continue, the same price is locked in for three years from the contract date (standard plans are expected to be $1,000–2,000/month). The program runs for six months from the contract date, and you can cancel at any time with one month’s notice.
For other questions, please contact us. A detailed security document for your review is also available.
Early Adopter Program
First 5 companies · Until the end of November 2026
No setup fee, $500/month (plus applicable taxes). If you continue, the same price is locked in for three years from the contract date. The dashboard, three alert types, onboarding support and monthly reviews are included at no additional cost.
Contact us
For onboarding questions or security review materials, feel free to get in touch.
Email: info@manageai.jp