As operations grow with new processes and AI Agents, management faces a practical question: how can teams move faster while maintaining control over costs, security, and stability? Choosing models, tracking consumption, and detecting exceptions become parts of the same responsibility.
The improvements presented at Pipefy Updates connect these decisions. You will learn about capabilities for managing AI usage, bringing operational indicators together, and monitoring executions, with information that helps guide management as processes scale.
Match the model to the task
Classifying a request and analyzing a complex contract require different capabilities. Having model options lets teams consider the nature of the task when deciding how to use AI in a process.
Pipefy AI offers Lite, Standard, and Pro models, selectable for each AI Agent behavior. This brings the decision closer to the work being performed: a model can be evaluated for a specific behavior according to task complexity and the consumption involved.
Bring Your Own Model, or BYOM, extends this management to models connected by the company. Model Registry lets teams register providers and keys, define a default model, and select models by behavior. Adding Anthropic expands the options for companies using Claude models.
Privacy is also part of AI model management. With Zero Data Retention, prompts are not retained. BYOM lets companies connect models using their corporate keys.
IT and process owners can organize model usage around their needs and the providers adopted by the company. In a contract process, for example, classification and analysis tasks can be evaluated separately, without treating every execution as if it required the same capabilities.
Explore Lite, Standard, and Pro, in Portuguese, BYOM, in Portuguese, and connecting Claude models, in Portuguese.
Bring the data together to decide where to act
When several teams use Pipefy, a consolidated view helps explain consumption before resources are allocated or increased usage is investigated. Company Overview brings together API Calls, AI Credits, and automation indicators across multiple Pipefy organizations in a central dashboard.
For customers with the Custom Integration add on, Integration Usage Stats adds detail for integrations, with consumption by pipe and flow, period filters, trends, and data export. In purchasing operations, for example, the team can identify which integrations account for the most consumption and relate that information to process activity.
Identifying inactive users also supports license reviews. That information gives management a starting point for assessing access needs and opportunities to reallocate licenses.
Pipe visibility adds another dimension. The Public column in Usage Stats identifies public pipes and lets teams sort the list by that condition. Consumption, usage, and visibility can therefore support more concrete conversations about operations.
Explore Company Overview, in Portuguese, integration usage indicators, in Portuguese, and the pipe visibility column, in Portuguese.
Connect auditing to the company's routine
Beyond understanding consumption, management needs to know who made a change and what changed. The Pipe Audit Logs API provides access to audit records by pipe, with structured information such as object identifiers, the actor responsible, and before and after values.
These records support connecting the audit trail to the systems used by the team, including monitoring and security tools. During an investigation, the team can work with evidence of changes made to the process.
Audit trail reliability improvements also include controls for uniqueness, delivery confirmation, persistence, and reprocessing. These mechanisms support the consistency of the records used in that analysis.
Learn about the Pipe Audit Logs API, in Portuguese.
Monitor exceptions in the team's tools
When an execution needs attention, context matters to the person investigating it. Webhook alerts send structured information about automation and AI Agent executions, including status and details associated with the pipe and card.
Webhooks are a way to send events to another system. Once the connection is configured, the team can bring monitoring into tools it already uses, such as Slack, Teams, or Datadog, and organize its response around the information received.
If an automation fails after a process change, the alert helps the team locate the card and review the failure reason. That information can be connected to logs and version history to investigate the change and decide how to resume operations.
There are also alerts regarding approaching monthly limits. Email notifications alert you when API call and automation task usage reaches 80%, 90%, and 100% of the monthly limit, helping your team track the consumption of contracted capacity.
Explore automation webhook alerts, in Portuguese, and AI Agent alerts, in Portuguese.
Save a stable version before making changes
Version history lets teams save and name a pipe version as a stable reference before changing its configuration. If a return is needed, restoring recovers that version.
Imagine that cards stop advancing as expected after a change. With a stable version saved, the team can restore the earlier configuration and use logs to investigate what changed.
This capability supports change control. A separate sandbox environment is a future development, distinct from version history and restoration.
Grow with decisions grounded in actual usage
Model choice, consumption indicators, audit records, and execution monitoring work together. They provide context for deciding where to expand AI usage, what to monitor, and which areas need review.
Keep exploring Pipefy Updates
Return to the event recap or explore the other journeys:
