Real-Time AI Spend with AI Signals

AI Signals gives your organization real-time visibility into AI spend across providers, models, and users, and connects that spend to the business context behind it.

AI Signals data feeds into CloudZero alongside your cloud and SaaS spend. Allocate AI costs using the same Dimensions and rules as everything else: by team, project, activity, or any category your organization defines.

Costs shown in AI Signals are estimates based on token volume at published market rates. These estimates are directional and can differ from reconciled billing data across the platform.

AI Signals has three tabs, each serving a different stage of the analysis workflow:

Overview

Overview consolidates AI cost, usage, and operational signals into one cross-filterable view, giving you situational awareness of your organization's AI spend without building a query or navigating between tools.

The exact panels you see depend on your connected data sources. Overview can include:

• Headline KPIs: Total cost, projected end-of-month spend, active users, cost per active user, total tokens, and model count. Cost and token KPIs include trend sparklines and period-over-period deltas so you can spot trajectory changes at a glance.
• Spend breakdown: Ranked tables showing which departments, repos, and activities are driving cost, with links to drill into AI Explorer for deeper analysis.
• Month-end forecast: Projected spend for the current month based on recent trends.
• Cost and token trends: A time series chart broken down by model, with a toggle to switch between cost and token views.
• Anomalies: Active cost anomalies from the AI User Monitor appear alongside the spend data with sparklines showing the spend trajectory.
• Top users and top models: Top users ranks individuals by AI spend, with a Users tab and an Activities tab. Top models lists every model in use sorted by cost, showing total cost, token count, price per million tokens, and cache rate.
• Cache economics: Displays the cache-read share of total cost as a headline percentage, alongside a token-level breakdown showing where savings are landing across models.

Filtering and data availability

Click any row in the spend breakdown, top users, or top models panels to refilter the rest of the page to that slice, or use the Add Filter button to apply persistent filters.

You can define custom AI dimensions using CostFormation rules that resolve from AI telemetry. For example, write a rule that groups AI spend by project or product, and that breakdown shows up in AI Signals within minutes. Custom dimensions appear in the dimension pickers on Overview and AI Explorer automatically once published.

Some breakdowns depend on data from your AI gateway or collector. If a panel shows an empty state, your data source is not yet sending that information. Reach out to your account manager for help configuring your setup.

Livestream

Livestream shows every AI inference event as it arrives, within seconds of the call being made. Each row is an individual event with context: who made the call, which model handled it, what the activity was, how many tokens it consumed, and what it cost.

Each event row shows timestamp, caller, model, source, tokens, cache, and cost by default. Additional columns (activity, department, team, project, topic, issue, tool, GitHub repo, git branch, and git user) are available through the Customize Table panel. The screenshot above shows the table with activity turned on.

Filter events by model or source using the dropdowns at the top of the page, or search for specific events using the search bar. Click any event row to open the event detail panel with context for that inference call.

AI Explorer

AI Explorer aggregates recent estimated AI spend so you can answer questions like "which models is each team using?" or "what activity is driving cost growth?" without waiting for reconciled billing data. If you know how to group, filter, and read a cost table in Explorer, AI Explorer works the same way, scoped to AI spend and updated within minutes.

Group and filter

Select a dimension from the Group By picker and set a time range. The chart and table update with your recent AI spend. Add additional group-by levels to break spend across up to eight dimensions at once. Use Add Filter to narrow to a specific model, team, or activity.

The Group By picker shows both built-in AI dimensions and any custom CostFormation dimensions that resolve from AI telemetry. Custom dimensions appear automatically once published; there is nothing to configure on the AI Signals side.

Chart and table

Toggle between a cost view and a token view in the chart to see the breakdown by spend or by usage volume.

The table shows total tokens, estimated market cost, and percentage of total by default. Additional columns (cache and models) are available through the Customize Table panel. Click any column header to re-sort.

To share the analysis with someone outside CloudZero, select Export CSV to download the current view. Export availability and completeness depend on the current grouping and how many rows are loaded.

Estimated vs. reconciled costs

Costs in AI Explorer are estimates based on token volume at published on-demand rates. They appear within minutes. Reconciled costs from your provider billing are available in Explorer for finance-grade reporting.

Putting it together

The tabs in AI Signals are designed to work together. Overview gives you the high-level picture. Cross-filters on any panel narrow the rest of the page to that slice. The department card links directly into AI Explorer for a deeper breakdown. If a cost spike triggered an alert, the Anomalies panel links directly to the associated incident.

When you validate a new integration, use Livestream to confirm events are arriving with the expected context.

Available integrations

Connect your AI platforms to start sending data to AI Signals:

For provider-level billing connections (Anthropic, OpenAI, Cursor), see AI Platforms.