Overview
Pipefy AI Agent behaviors can now use three different AI Models, Lite, Standard, and Pro, each built for tasks with different levels of complexity.
The Model is chosen per behavior, not per Agent, giving you full control over the reasoning depth used at each step of the process.
What's new?
- Three new AI Models available when configuring each AI Agent behavior.
- Lite, 2 credits per run, optimized for simple tasks and delivery speed.
- Standard, 2 credits per run, the default Model.
- Pro, 3 credits per run, with a promotional rate of 2 credits until October 8, 2026, optimized for complex tasks.
- Each of the up to 5 behaviors per AI Agent can work with the Model that best fits the tasks it needs to perform.
How does this help?
- Reasoning choice per task: Lite and Standard cost the same in credits, so choosing between them is about the type of task, not about savings. Pro stays reserved for behaviors that truly need deeper analysis.
- Speed when it matters: the Lite Model delivers faster responses for simple, well structured tasks, without costing more credits than Standard.
- Consumption visibility: track your organization's AI credit consumption directly in the Admin Panel.
- No charge for zero value scenarios: behaviors whose condition isn't met, or that receive data beyond the supported limit, don't consume credits.
Use cases
1. Fast comment translation with the Lite Model
An AI Agent configured with the Lite Model automatically translates a requester's comment into the support team's language as soon as the card enters the triage phase. Since the task is simple and predictable, Lite delivers the result faster than the other Models, for the same credit consumption as Standard, avoiding queues in high volume phases.
2. Invoice extraction and flow decisions with the Standard Model
In a purchase request pipe, the AI Agent uses the Standard Model to read the attached invoice PDF, extract vendor, amount, and dates, validate that data against the original purchase order, and automatically move the card to Awaiting Approval or Exception. The purchasing team no longer opens every invoice manually to check the data before deciding the next step.
3. Complex quote comparison with the Pro Model
When reviewing three vendor quotes received as PDFs, the AI Agent configured with the Pro Model cross references the RFP requirements against each proposal, scores every vendor, documents the gaps found, and moves the card to the correct approval flow with a written rationale. The purchasing team receives a fully justified recommendation without having to build the comparison in a spreadsheet.
4. Invoice analysis with the Pro Model and document analysis
When analyzing an invoice attached to a card, the Pipefy AI Agent runs a behavior configured with the Pro Model. Total consumption is 4 credits, 3 credits for using the Pro Model plus 1 credit for the specific document analysis action, without anyone needing to manually check the invoice data before moving the process forward.
Most relevant for
- Pipe Admins
- Operations teams
- Purchasing managers
- Organization admins
Documentation
For the full details on the new AI Models and pricing, check the Help Center article.
- For more details on connecting your own AI model, check the Help Center article.
- To understand credit consumption for behaviors that use document interpretation, check this article.
Plans and availability
- Every plan includes 1,000 one time use AI credits. Additional packages of 500, 1,000, 10,000, or 50,000 credits can be purchased as needed.
- Until July 12, 2026, AI Agents use the Classic Model or models connected via BYOM.
- Starting July 13, 2026, the Lite, Standard, and Pro Models become available, and the Classic Model remains active in parallel.
- Starting September 30, 2026, the Classic Model is discontinued. Behaviors configured with Classic automatically migrate to the Standard Model, with the same 2 credit consumption.
- Until October 8, 2026, the Pro Model consumes 2 credits as a promotional rate. Starting October 9, 2026, it consumes 3 credits.
- When using your own model via BYOM, each Agent and Automation run consumes 1.5 credit, plus specific charges for agentic capabilities such as document analysis, calculations, or web search.
