What are Agent Compute Units (ACUs)?

Agent Compute Units (ACUs) are used to pay for the consumption of AI-powered analytics resources on TextQL. An ACU is a unit of measure that represents the computational resources, AI model usage, and business value delivered.

ACU Value Proposition

Unlike traditional compute billing models that charge for infrastructure time, ACUs are designed to align costs with business outcomes:
  • Outcome-Based Pricing: You pay for delivered insights and completed analyses, not idle time
  • Transparent Value: Each ACU represents a specific action, which can be audited later by organization administrator
  • Predictable Scaling: ACU consumption scales with your analytical needs, not infrastructure complexity

ACU Usage and Consumption

Agent operations come in different complexity levels and ACUs are charged based on the complexity of the tasks performed. The ACU numbers shown below represent the cost for each complete operation.
Task CategoryOperationACU Cost
OntologyCreating a Metric2
Creating a Dimension2
Documenting a Column2
RetrievalPlanning5
Search15
Select50
Object Explorer Query50
Metric Explorer Query50
Load CSV25
AnalysisRead Tableau Dashboard50
Cleaning (Python)20
Visualize (Python)35
ML Modeling (Python)125
PresentationApp Creation (Streamlit)250
Generate Summary50
Generate Recommendations100
Create PDF Report250
Schedule Run175

Key billing considerations

  • ACUs are only consumed when agents perform work. When no tasks are being executed, no ACUs are consumed.
  • Each agent task is billed as a discrete unit based on successful completion. Complex operations like Metric Explorer Queries may involve multiple sub-tasks but are billed as a single unit.
  • Current ACU allocation and consumption can be analyzed through the TextQL Usage Page, which is accessible by organization admins.

Optimizing ACUs

Excess ACU consumption is mainly caused by unnecessary dataset discovery and search within Threads; the more context you provide as to what you are looking for and what your desired output is, the less likely it is that Ana spends extra effort on your Thread or message. Below are the most effective levers our customers use for reducing ACU usage while still getting answers fast.
Optimization LeverWhy It HelpsHow to Apply It
Reuse agent context instead of starting fresh ThreadsA running Thread preserves chat history, intermediate calculations, and cached result sets. Building on that context prevents the agent from re‑discovering tables or regenerating plots it already created.
  • Continue follow‑up questions in the same Thread whenever the goal is the same.
  • Start a new Thread only when you change topics or Ana says the Thread is too long.
Leverage a well‑curated OntologyColumns, metrics, and dimensions registered in your Ontology steer Ana straight to the right data and filters, cutting exploration time.
  • Use the Ontology setup guide to register core metrics/dimensions.
  • Keep definitions up to date; encourage analysts to add new ones as they create them.
Embed known SQL snippets or table names in PlaybooksIf Ana already knows where the data lives, she can skip search → sample → profile cycles.
  • Prototype the Playbook in a Thread first.
  • When satisfied, ask Ana to rewrite the prompt with explicit SQL/Python snippets for speed.

Frequently Asked Questions

Pricing and Value

ACU Management

Technical Details