Estimating AI Action Usage

AI Actions are the unit of measurement for AI consumption in Appian. MyAppian displays AI usage metrics on the Adoption page to help you monitor AI Action consumption across environments and Appian AI features.

How AI Actions are consumed

Every execution of an Appian AI feature consumes a specific number of AI Actions. A single business workflow might involve multiple executions, such as an AI Agent completing a multi-step task, a DocCenter model extracting data, or Smart Search indexing a document. Your total estimate must account for every execution within your business workflow.

The following table shows how AI Actions are measured for individual features, and how feature implementations may affect the consumption of those AI Actions.

AI Feature	Measurement Trigger	Volume Driver
AI agents	An AI agent is invoked to execute a task or sequence of steps.	Number of agents, number of tools used to execute a task or sequence of steps, number of steps or tasks completed
DocCenter	Document or pages processed for classification or extraction.	Number of documents, average pages per document, features enabled (classification, extraction, AI review, advanced layout)
Composer	AI features are invoked during app development.	App planning sessions, sample data rows generated, test case rules executed
Smart Search	Content is indexed for AI-powered search.	Volume of structured text (rows × fields × character length), volume and size of unstructured documents (PDFs, Word docs)
Prompt Builder	A prompt is submitted to an AI model.	Monthly prompt volume, total context length (document + instructions + AI output), document type (text vs. visual elements), number of AI calls per process
Chat Features (Records Chat, Data Fabric Chat, Enterprise Copilot, Document Chat, a!callLanguageModel)	A chat conversation session is initiated.	Number of monthly conversations, average messages per conversation session

Most AI feature executions start at a baseline of 10 AI Actions. However, consumption varies based on factors like input size, document complexity, and the number of tool calls or sub-tasks involved. Your total process estimate is the sum of all individual AI Actions consumed by each feature within that process.

AI Action consumption by feature

The following table shows typical AI Action consumption rates for each Appian AI feature.

AI Feature	Typical AI Actions per Execution	Notes
Most AI features (standard execution)	10	Baseline rate for most feature executions.
AI Copilot task	10	Applies to tasks performed by the AI Copilot assistant.
DocCenter process (IDP)	37	A single run of a DocCenter process with Advanced layout configuration. Includes base costs for prompt, classification, and extraction steps. Consumption may increase based on the number of pages or complexity of the document.
AI agent run	100	Applies to complex tasks requiring flexible, goal-driven automation. More complex runs can consume more than 100 AI Actions.

How to estimate AI Action usage

Use the following workflow to estimate AI Action consumption while designing and developing your application.

1. Map your AI calls: In your process model, identify every AI-powered smart service node, such as Execute Generative AI Skill. You will perform a separate estimation for each node.
2. Determine the baseline: Identify the baseline consumption for each feature by using the AI Action consumption by feature table.
3. Identify complexity variables: For each call, determine if dynamic data might increase the consumption. For example:
   • Generative AI: Combined prompt and response text consumes additional AI Actions beyond the baseline.
   • DocCenter: Consider if documents are unstructured or contain heavy graphical elements, which require more computational resources.
   • AI Agents: Estimate the number of sub-tasks or tool interactions the agent must perform to reach a successful outcome.
4. Sum for a total process estimate: Add the estimates for every AI feature in your workflow. The result is the estimated AI Action usage for a single, end-to-end transaction.

Use your estimate to make design decisions

Understanding the AI Action cost of a transaction empowers you to make critical design trade-offs.

For example, you might ask:

• Is the added accuracy worth the cost? "My prompt is more accurate with five few-shot examples, but my estimation shows this adds AI Actions, pushing each execution past the threshold and consuming additional AI Actions. For a process that runs 10,000 times a month, is the accuracy gain worth the additional consumption?"
• Can I be more concise? "Does my summary really need to be five paragraphs long, or can I prompt for a three-sentence summary and keep each execution under the threshold?"

Licensing and platform considerations

Keep these points in mind when managing your overall usage:

• Monthly allotment: Your allotted AI Actions are for a specific monthly period based on your license tier. Any AI Actions not consumed by the end of the month will expire and the count will begin again with the new calendar month.
• Service continuity: Appian will not abruptly disable the AI service if you reach your AI Action limit. Instead, we will notify your Appian Account Executive to discuss your usage. Any additional usage beyond the included amount will be billed according to the applicable overage rates.
• Purchasing additional AI Actions: If you anticipate exceeding your limit, you can purchase additional AI Actions. Contact your Appian Account Executive for details.

Validate estimates with the AI usage log

After deploying your application, you can compare your estimates against actual per-execution AI Action counts recorded in the AI usage log. The AI usage log records the number of AI Actions consumed each time an AI feature executes, giving you granular data to validate your planning assumptions.

Use the AI usage log to:

• Confirm estimates match reality. Compare the AI Actions recorded for each feature execution against the baseline rates in the AI Action consumption by feature table.
• Identify unexpected consumption patterns. If actual AI Actions consistently exceed your estimates, investigate whether input complexity (document length, number of agent tool calls, or prompt size) is higher than anticipated.
• Inform design trade-offs. Use actual consumption data to make evidence-based decisions about prompt length, feature configuration, and process design.

The AI usage log complements your estimates — it does not replace them. Estimation helps with planning and budgeting before deployment. The log provides actual consumption data for validation and ongoing governance after deployment.

Feedback