AI Tool Spend Report

Overview

The AI Spend Report is a beta feature that provides visibility into your organization's investment in AI coding tools. This report helps you track and analyze spending across developer tools like Cursor, GitHub Copilot, and Claude Code, breaking down costs into subscription-based and usage-based components.

With the AI Spend Report, you can:

• Track total AI tool costs over time

• Understand spending patterns across your organization

• Analyze cost per developer to optimize tool allocation

• Break down costs by tool and AI model to identify usage trends

• Monitor base subscriptions vs. overage costs for budget planning

Navigate to Organization → AI Transformation → AI Spend to access the report.

Setup

Integration setup

• Claude code OTEL collector setup

• Claude code API integration

• Cursor API integration

• GitHub Copilot integration

AI spend integration settings

In order to accurately calculate base spend over time, we need to be aware of the license cost per person and billing date. To set this up, you can visit Settings -> Integrations and go to the AI Spend Integration Settings section of the page.

You can configure multiple plan settings over time to accurately reflect plan changes (e.g., upgrading from Team to Enterprise). Each plan setting has a Valid From and Valid To date range. When calculating base spend, Span applies the correct plan rate for each time period, so historical queries always reflect the accurate cost. 

You can also set up historical per-person plan settings for Claude code in the Claude Code user mappings section of the page. Those override the default settings so please make sure to set up the most common plan for OTEL and then only select specific people for the different seat options.

Keep in mind that individual accounts can be tracked only for Claude code via OTEL - other tools require an integration for the organization / team.

Available plans by tool

Base spend

Base spend is your predictable, subscription-based cost - the fixed monthly fee charged per seat for each active developer.

Base Spend = Plan Cost per Developer × Number of Active Developers × Time Period

For each tool we calculate daily license spend per person in various ways.

Cursor

• Data source: Cursor's daily usage API endpoint.

• Settings: Per-connection plan settings. Each Cursor integration connection can have its own plan/pricing with a default fallback to Team plan.

• Identifier for each person is email address. Please make sure to use the same email you use in Span for your Cursor account.

Copilot

• Data source: GitHub Copilot seat assignments endpoint.

• Settings: Pricing is determined by plan_type from the seat assignments data and it is not configurable in the app. Only the billing date is available to configure and defaults to the 1st of each month.

• Identifier for each person is GitHub username. Please make sure to add all your used GH usernames in Span.

Claude Code

• Data source: OTEL collector data is used for base spend attribution.

• Why OTEL only: The Anthropic usage report API provides data for API users only, so subscription cost for those users is always $0. OTEL data, by contrast, arrives regardless of license type and allows for a variety of plans to be tracked - both personal and team subscription accounts are available if you have the correct setup. The Claude code API's usage report data is still used for overage cost attribution.

• Settings: Org-level OTEL settings apply to all users by default. Individual users can have person-level plan overrides configured in Settings.

• Identifier for each person is email address. Please make sure to use the same email you use in Span for your work email setting in the OTEL configuration.

• Override merging: When a person has custom plan settings, we splice those overrides into the OTEL base timeline, producing a gapless sequence of settings. Override periods use the override's plan/cost but inherit the billing date from the base setting in effect at that point.

Overage spend

Overage spend represents variable costs for additional usage beyond your base subscription. This includes:

• Token usage charges for AI models

• Premium feature usage

• API calls beyond subscription limits

Cursor

• Data source: Cursor's filtered usage events API endpoint.

• Included events: For non-Enterprise customers, only on-demand events are included in overage cost. For Enterprise, the events are categorized as "Included in Business" even if they do not fit into the licensed usage limits due to pre-committed costs in each contract. We include those in Overage calculation from the point when Enterprise plan becomes valid.

Claude Code

• Data source: OTEL collector provides us the metrics claude_code.cost.usage and claude_code.token.usage. We also use Claude code usage report data (Anthropic API) when available with a higher priority than OTEL.

• Usage report for API keys: If you are using API keys for Claude code, you can set up the API key names in People management the same way you set up your VCS usernames. The usage via API key and email will be combined together in the report.

Copilot

• Data source: Premium requests data is not available through the Copilot API. We allow users to import a CSV that's extracted from GitHub UI in order to track overage spend. Another limitation is that GH does not expose token spend through API or CSV exports.

  • You can see an example of the info we need here: https://github.com/orgs/community/discussions/157693#discussioncomment-13304789

  • You should be able to find the report here (please double check that it matches the report in the community discussion above)
    https://docs.github.com/en/rest/billing/usage?apiVersion=2022-11-28#get-billing-premium-request-usage-report-for-an-organization

Report Features

AI Spend Visualization

The main chart offers multiple views to analyze your spending:

Base & Overage View (Default)

• Stacked bars showing base vs. overage spend trends over time

• Quickly identify when overage costs spike

By Model View

• Line chart showing spend per AI model (e.g., GPT-4, Claude, etc.)

• Understand which models your team uses most

By Tool View

• Stacked bars showing spend per developer tool (Cursor, Copilot, Claude Code)

• Compare tool adoption and costs

Toggle between Bar Chart and Line Chart formats using the chart type selector.

Summary Cards

Three summary cards provide quick insights:

1. Total Base Spend - Shows your total subscription costs with period-over-period comparison

2. Total Overage Spend - Displays variable usage costs with trend indicators

3. Average Cost per Developer - Your per-developer spending average

Spend Breakdown by Tool & Model

Expandable cards provide detailed breakdowns including:

• Overage spend per tool and model combination

• Input token usage

• Output token usage

• Total costs for each category

Filtering & Analysis

Refine your view using available filters:

• Date Range: Select the time period to analyze

• Teams & People: Filter by specific teams or individuals

• Person Attributes:

  • IC Level

  • Manager Status

  • Job Title

  • Location

Data Table View

Switch to table view for a detailed breakdown by person, showing:

• Individual developer spending

• Metrics across selected time period

• Filterable and sortable columns

Troubleshooting

❓ Missing base spend data?

• Verify your tool plan settings are configured

• Check that the correct plan cost and billing date are set

• Ensure the tool is actively being used by developers in Span

❓ Overage costs seem incomplete?

• For Claude Code: Check if there are known data collection issues

• Verify integrations are healthy and collecting telemetry data

• Contact support if gaps persist

❓ Don't see your tools?

• Only tools with active usage data appear in the report

• Configure settings for tools you want to track

• Verify integrations are properly set up

Frequently Asked Questions

Q: Why don't I see any data in the report?

A: Ensure that (1) AI tool plan settings are configured, (2) developers are actively using the tools, and (3) integrations are collecting data properly.

Q: Can I track different plans for different teams?

A: Not currently. The system supports a single plan per tool across your organization. Support for mixed plans is under consideration for future releases.

Q: How far back can I view historical spend data?

A: Historical data availability depends on when you enabled the feature and configured tool settings. Spend calculations begin from your configuration date.

Q: What happens if I change my plan mid-month?

A: The ability to edit historical settings is being developed. Currently, plan changes apply from the update date forward. Contact support for guidance on handling mid-period changes.

Q: Are the costs I see actual charges from the tool providers?

A: The report calculates estimated costs based on your configured plans and observed usage. Base costs are calculated from your plan settings × active developers. Overage costs are estimated from token usage × model pricing. Always verify with your actual bills from tool providers.

Need Help?

If you have questions about the AI Spend Report or need assistance:

• Contact your Span account team

• Reach out to support@span.app