AI Units included
Upgrading your Box subscription tier unlocks advanced foundational reasoning, orchestration capabilities, and larger baseline AI Unit allocations.
Key Capacity Considerations:
- Entitlements: AI Units are allocated for a defined entitlement period (e.g. monthly) and are comprised of your base plan allotment plus any additional units purchased via a service order.
- No Rollover: Unused AI Units expire at the end of each entitlement period and do not roll over to the next period.
- Aggrerated Use: If you utilize multiple chargeable AI features (e.g., Q&A AI API and Text Gen AI API), your unit consumption is aggregated against your enterprise-level entitlement pool.
Included vs. Chargeable Features
To manage your budget effectively, it is critical to understand what triggers an AI Unit charge. Chargeability is determined by the feature and mode of access (Box Web App UI vs API).Features included in your plan
In plan interactions do not deduct from your monthly AI Unit pool:- Box Agent (Standard Mode): Standard interactive end-user prompts within the UI.
- Box Agent (Pro Mode): Exclusive to Enterprise Advanced. When enabled, the agent utilizes advanced reasoning capabilities for complex tasks like strategic analysis or coding. Note: Responses may take longer due to increased processing, but it remains included in your plan when run interactively by a user in the Web App. The Box Agent via API is chargeable, however.
- Custom Agents (Standard Mode): Tailored agents used within the Web App, provided they are not configured in Expanded Mode.
- Box Apps and Search: Built-in platform conversational search capabilities.
- Box Shield Pro & Auto-Classification: Includes native security classifications, the AI Classification Agent, and the AI Threat Analysis Agent. These automated protections run in the background at zero AI Unit cost.
Platform stability impact: While these features do not consume your monthly AI Unit allocation, Box reserves the right to enforce daily AI limits on in-plan features to maintain performance stability. Detailed reporting for in-plan usage will appear in the Admin Console, with daily limit enforcement starting July 2026.
Model Context Protocol (MCP) Partners Integrations
For third-party platform integrations connected via the MCP framework, capacity metering is split by call type and feature category:- Only AI Usage Consumes AI Units: Traditional platform transactions (such as file fetches, folder navigation, or standard structural metadata requests) triggered by an MCP app do not affect your AI pool.
- AI Extraction is Always Chargeable: Any metadata extraction (MDE) task orchestrated by an MCP tool is executed via API and consumes AI Units under standard extraction rules.
- AI Q&A and Text Generation Metering:
- Current State: Because MCP integrations function as API-driven developer access, all Q&A and text generation tasks orchestrated through the Box MCP server consume AI Units. (Note: Standard interactive use of Box AI directly within the native Box Web App UI remains included in your plan).
- Future State: The metering framework will transition to differentiate between Published MCP Apps (from the public marketplace) and Custom-Built MCP Apps, altering how unit deductions are calculated.
Chargeable Features (Consumes AI Units)
AI Units are consumed when actions are performed at scale, are automated, are not directly managed user-driven, or leverage advanced enterprise configurations:- Box AI APIs: Any access via the Q&A AI API or Text Gen AI API.
- Box Automate Workflows: Any agent execution integrated into an automated Box workflow process.
- Expanded Mode Agents: An optional configuration in Box AI Studio for Box or Custom Agents that removes standard query restrictions, allowing the agent to process massive, content-heavy datasets.
- Standard and Enhanced Extract Agents: A cost-effective tool designed for simple, structured, or short documents (e.g., receipts, invoices, passports, salary slips, or resumes typically under 5 pages). It is optimized for high-volume, time-sensitive pipelines where information can be directly read or normalized without complex reasoning or calculations.
Admin Console model labeling and Auto selection: Within the Admin Console and Box AI Studio, models are explicitly labeled as Standard or Premium to remove any guesswork. When building or configuring a custom agent, selecting the Auto model configuration defaults to a Standard model tier to optimize performance speed and unit cost-efficiency. Choosing a Premium model increases the rate of AI Unit consumption, but only when paired with one of the chargeable features listed above.
Monitoring AI Unit usage
Only Account Admins have visibility into AI Unit consumption. To view current consumption against your monthly entitlement:- Go to the Admin Console.
- Select Home and then scroll to the AI Insights card.
To access historical data, Admins can now download the historical AI units consumption report with an ability to filter by date, user and AI Capability from the Reports page in the Admin Console.
Important Tracking Discrepancies to Note:
- Data Lag: The AI Unit Insights card operates with an approximate 2-hour data lag. It does not track consumption in absolute real-time.
- AI Studio Dashboard Window: The “Monthly Queries” chart within the AI Studio dashboard reflects a rolling 30-day window, whereas your official AI Unit allocation resets strictly on the first day of the calendar month.
- User-Level Auditing: Admins can download detailed User Activity Reports to break down unit spend by individual user or specific agent. (Data is available retroactively to January 1, 2026).
What happens when you reach your limit?
Box notifies admins at the following consumption thresholds:On track to exceed entitlement
- What happens: The admin sees a purple warning banner in the Admin Console.
- Recommended action: Contact your Box account manager to purchase an expansion block before operations are interrupted.
At full entitlement:
- What happens: Paid/chargeable AI features are blocked until the next entitlement period or until additional units are purchased. Free/included in plan Web App features remain fully functional.
- Admin view: A persistent red warning banner appears in the Admin Console, and an automated email notification is sent.
- User view: Users receive an error message (“AI unit entitlement has been reached”) if they attempt to trigger a blocked feature.
Minor overages: You may occasionally notice that your final monthly usage slightly exceeds your allotment (e.g., 20,320 / 20,000). This occurs due to the 2-hour processing lag and internal business rules designed to let active, in-flight requests finish rather than breaking an active user’s session mid-task. New requests are blocked immediately following.
AI Unit consumption
For information on how AI Units are consumed across Box AI APIs, agents, and Automate workflows, reference the Box AI Units consumption table.Chargeability & Exhaustion Matrix
To manage your budget effectively, it is critical to look at both the feature name and how it is being executed. The table below outlines the definitive chargeability status and the expected operational behavior if your organization exhausts its monthly AI Unit allocation.Best practices for efficient AI Unit management
- Choose models intentionally: Use Standard models (or select the Auto setting) for routine tasks, bulk processing, and basic summarization to optimize cost-efficiency. Reserve Premium models for complex reasoning, multi-step tasks, and long-document analysis.
- Configure Custom Agents smartly: Prevent unexpected drawdowns by setting a maximum query length cap, disabling agents that are not actively used, and restricting agent folder scope (fewer scanned files equals fewer AI Units consumed).
- Monitor proactively: Review the AI Unit Insights card weekly and leverage the downloadable User Activity Report to identify heavy users or runaway automated processes.
How the entitlement resets
- Calendar reset: Entitlements reset at the start of each entitlement period. Unused units do not carry over.
- Mid-period purchases: If you purchase additional AI Units mid-period, access to blocked features is restored shortly after the purchase is processed. These add-on units will carry over through the end of your contract order date.
- To purchase more: Contact your Box account manager or reach out directly to the sales team at sales@box.com for baseline pricing tiers and subscription blocks.
Box AI Units FAQs
Do unused AI Units roll over to the next period?
Do unused AI Units roll over to the next period?
No. AI Units operate on a “use it or lose it” basis and reset on the first day of each entitlement period (e.g. monthly).
Why does my AI Studio dashboard say I have exceeded my queries if my entitlement period just restarted?
Why does my AI Studio dashboard say I have exceeded my queries if my entitlement period just restarted?
The AI Studio dashboard displaying Monthly Queries tracks usage on a rolling 30-day window, which functions independently of your strict entitlement period (e.g. calendar month).
Do AI Units apply to Box AI for documents and notes in the user interface?
Do AI Units apply to Box AI for documents and notes in the user interface?
No. Standard document interactions, standard Custom Agents, and Pro Mode within the Box Web App UI are user-driven and are included in the plan. AI Units are primarily consumed by API calls, Automate workflows, Expanded Mode configurations, and Metadata Extraction Agents.
Is Box Agent Pro Mode available on all plans?
Is Box Agent Pro Mode available on all plans?
No. Box Agent in Pro Mode (and access to Box AI Studio for custom agents) requires an upgrade to the Enterprise Advanced tier.
Are the AI features in Box Shield Pro chargeable against my AI Unit pool?
Are the AI features in Box Shield Pro chargeable against my AI Unit pool?
No. The background AI tools included with Box Shield Pro (like the AI Classification Agent and AI Threat Analysis Agent) are native security features covered by the Shield Pro license and do not deplete your monthly AI Units.
Why did our organization manage to spend over our rigid limit?
Why did our organization manage to spend over our rigid limit?
Box employs a ~2-hour data processing lag and chooses not to terminate in-flight active processes mid-query to avoid data corruption. Once those active sessions conclude, new chargeable requests are blocked.
What should I do if I see the error "AI unit entitlement has been reached"?
What should I do if I see the error "AI unit entitlement has been reached"?
Contact your organization’s Box Administrator. The administrator can contact Box Sales to purchase an immediate allotment block or coordinate internal workflows until the entitlement resets.