👤 For people who build, oversee, and improve processes in Pipefy
🔐 For customers with access to Pipefy Copilot during the closed beta
🎯 For people who want to create a process or assess improvements to one already in use
A purchase request is waiting in Approval, and the team needs to find out what is keeping it there. In another workflow, finance gathers receipts sent through separate conversations before reviewing a reimbursement. Both situations call for turning a business need into a process with clear information, rules, and responsibilities.
By the end of this article, you will know how to start a conversation with Pipefy Copilot to improve an existing process or create a new one, assess the recommendations, and decide what to put into practice. The purchasing and reimbursement examples show what to check in each case while staying in control.
For questions about access, capabilities, permissions, AI credits, and support, see Pipefy Copilot: frequently asked questions.
What you will learn:
A conversation connected to your team’s work
Process intelligence, now one conversation away. With Pipefy Copilot, you describe what you need in everyday language and receive recommendations to apply to your processes. It helps diagnose, create, modify, optimize, and analyze process components within the platform.
The conversation brings together your company’s context and the experience Pipefy has built over 10 years of process modeling. You provide the business rules, assess the proposal, and decide how to proceed. People and AI work on the same task: recommendations help prepare a decision, and you stay in control of the changes.
To get started, click the Pipefy Copilot icon next to your profile photo at the top of the screen. The conversation opens in a column on the right. You can start with a pipe your team already uses. A pipe organizes a process into phases, and cards represent the requests moving through them. Explain the difficulty that prompted the analysis, such as requests waiting for approval or recurring questions about supporting documents.
For more on combining rules, AI, and human review in process tasks, read Understand when to use AI in your processes, available in Portuguese.
Starting with a process already in use
Choose a workflow you know and a difficulty you can recognize. In purchasing, for example, the goal might be to understand why a request is still in Approval and what the team needs to arrange before moving it forward.
Suggested prompt: Help me optimize my process.
In the purchasing example, the analysis highlights a software license request that has been in Approval for 12 days and phases without an assignee field. Pipefy Copilot suggests adding that field, creating an alert for requests that spend too long in the phase, or checking the delayed request’s details. These findings belong to the process being analyzed; recommendations for your pipe depend on the available context.
Use the findings to guide your investigation. A stalled request shows where to look more closely. A missing assignee field suggests an opportunity to record who is handling the work, but you still need to understand with the team what prevented the request from moving forward.
When you choose to add the field, review the proposed changes. In this example, they are fields named Responsável in the Triagem, Análise da Compra, Aprovação, and Execução da Compra phases. These labels refer to triage, purchase analysis, approval, and purchase execution. Check whether those phases match the points where your team needs to record who handles each request.
Connect structure to everyday work: creating an assignee field provides a place to record a person. Agree on who will fill it in, who will handle the request, and how the information will be maintained. If the process needs assignment based on a rule, assess that separately. The guide to updating the assignee field through automation, available in Portuguese, explains that separate feature.
An alert also calls for a process decision: how long to wait before sending it, who should receive it, and what that person should do. An alert about time spent waiting and the deadline configured for a phase serve different purposes. Decide which need you want to address before requesting a change.
Creating a process around a business need
When a workflow still needs to be structured, start by explaining what you want to organize. A short request can begin the conversation; you can add context as it develops.
Suggested prompt: Help me create a reimbursement pipe.
In this example, Pipefy Copilot asks how requests should be submitted and approved. Options include requests with receipts and approval based on the amount, approval by finance only, or a specific approval policy. An approval threshold determines who can approve each amount range.
Choose the option that reflects your company’s workflow. If you have a specific policy, explain its conditions. When you are unsure about a rule, check with the person responsible for the process before approving a proposal based on it.
Choosing receipts with approval based on the amount leads to a plan proposed by Pipefy Copilot with phases, fields, automations, and a receipt validation AI Agent that Pipefy Copilot will configure. This combination is an example to review, rather than a required structure for every reimbursement process.
Assess the plan against the work it needs to support. Do employees know which receipts to submit? Does the reviewer receive the information needed? Do the approvers match the current policy? Does each phase have a clear purpose? If a rule differs, request a correction and check the updated proposal.
For example, if the plan suggests approval above R$ 500 but company policy sets R$ 700 for that situation, provide the correct threshold and approver. These amounts illustrate a company rule and do not represent Pipefy limits.
Reviewing changes before putting them into operation
Whether you are creating or improving a process, compare the proposed changes with the need that started the conversation. Adding a field helps when the team needs to record information in order to act. An approval rule is useful when it reflects the policy people need to follow.
In the illustrated flow, you approve the proposal and review the changes, which are prepared as a draft. Publishing is a separate decision: before confirming, check the change summary and decide whether the proposal is ready to publish or should remain in draft. In the reimbursement example, the automations and agent are created when you publish and start running from that point. If the pipe already contains cards that a rule would affect, Pipefy Copilot may leave that item disabled and flag it in the summary.
Include the people who know the affected rules. An approval threshold change calls for review by the person responsible for the approval policy. A new assignee field calls for agreement with the people who distribute and carry out the work. Their input helps you assess whether the proposal can work in practice.
Good to know: Once a change is published, it cannot be undone automatically, and deletions are permanent.
A criterion for moving forward: can you explain what will change, which difficulty it addresses, who will use it, and how to check the outcome? If any of those points is unclear, add the missing context before putting the proposal into operation.
Checking the outcome and choosing the next improvement
Agree on a validation exercise with the people responsible, using a representative request in an appropriate environment. For reimbursements, check that the required receipts reach the reviewer and that the approval rule matches the request amount. In purchasing, check that the assignee field can be filled in and that the team understands who should handle the request.
Once the change is in use, revisit the problem that prompted the conversation. Are requests still waiting because information is missing? Can the team identify who needs to act? This follow up helps you choose the next adjustment without assuming that a structural change resolves every cause of a delay.
Record the decision and the reason for the change so that other people can follow the process’s development. Use the guidance in Document your Pipefy processes for easier maintenance to keep that context available.
Practice choosing between rules, AI, and human review with Panorama da IA, a Pipefy Academy course available in Portuguese. Sign in or create an account to access the course.
Before moving on, confirm that you understand:
☐ You can start with a difficulty in an existing process or the need to create a new one.
☐ Recommendations need to be assessed against the rules, people involved, and outcome the team expects.
☐ Adding an assignee field calls for a decision about how the team will record and take responsibility for requests.
☐ Approving a proposal, publishing changes, and assessing the outcome are separate decisions.
Follow Pipefy on social media
Keep up with content, news, and tips on processes, automation, and AI: LinkedIn | Instagram Brasil | Instagram Global | YouTube | X
Ready to record your first improvement and make future changes easier? → Document your Pipefy processes for easier maintenance


