The Guided Managed Agent Builder in ZipLeads helps you create and edit AI agents using plain-English instructions. It offers targeted edits, guided clarifying questions and a live configuration panel, so you can make changes without rebuilding your agent. The builder also keeps related instructions synchronised and uses resources already in your account. This article explains how these features work and how to use them well.
What is the Guided Managed Agent Builder?
The Guided Managed Agent Builder is an AI-powered experience in Agent Studio. It lets you create and change Managed Agents (formerly Super Agents) through conversation. You don't need to update every setting by hand. Just describe the changes you want, review the results in a live configuration panel, and adjust supported fields directly.
The builder also recognises questions and asks for clarification when needed. When it makes targeted edits, it keeps unrelated settings as they are. These improvements are available for Managed Agents in Agent Studio with no extra setup.
Key Benefits of the Guided Managed Agent Builder
The builder makes creating and editing agents simpler, while giving you more control over configuration changes.
Targeted Edits: Change specific agent settings without affecting unrelated ones.
Guided Clarification: Sort out unclear requests by choosing the right resources, rather than relying on guesses.
Intent Recognition: Ask questions or request changes without accidentally modifying your agent.
Live Configuration Updates: Watch instructions, triggers, tools, skills and Knowledge Bases update in real time.
Inline Editing: Fine-tune supported fields directly in the configuration panel.
Smart Cascades: Related instructions are synchronised automatically when triggers, plugins or Knowledge Bases change.
Account-Grounded Resources: Set up agents using the forms, calendars, pipelines, Knowledge Bases and connected apps already in your account.
Targeted Configuration Edits
Targeted editing lets you make precise changes to an existing Managed Agent without regenerating its whole configuration. This keeps your working instructions, triggers and connected resources in place while you refine specific behaviours.
For example, if you want an existing agent to sound more conversational, enter:
"Make this agent's tone friendlier without changing its existing triggers or Knowledge Base."
The builder updates the relevant instructions and leaves unrelated settings alone.
Other examples include:
"Add a trigger for the Contact Us form."
"Make the agent's responses shorter."
"Update the instructions to ask for an email address before creating a task."
When an edit affects a related configuration element, the builder may also synchronise the matching instructions.
Guided Clarifying Questions and Intent Recognition
The builder works out whether you're asking a question, requesting a change or giving an unclear instruction. This helps prevent unintended edits and makes sure the right resources are chosen. If a request is missing information or is ambiguous, the builder asks a focused clarifying question instead of guessing.
For example, if your account has more than one form named "Contact Us", entering:
"Add a trigger for the Contact Us form."
prompts the builder to ask which form to use.
The builder also uses resources available in your ZipLeads account, which helps prevent invalid selections.
Request | Builder Behaviour |
"What can this agent do?" | Explains the agent's capabilities without changing its configuration. |
"Make the tone friendlier." | Updates the relevant agent instructions. |
"Improve this agent." | Asks for clarification before making changes. |
"Add a trigger for the Contact Us form." | Selects the matching form, or asks which one to use if there are several matches. |
"Use the Support Knowledge Base." | Uses the available resource, or asks for clarification if it can't be identified. |
The same intent recognition applies when you create a new agent. If you type a question instead of a creation instruction, the builder answers the question rather than creating an agent you didn't want.
Live Editable Configuration Panel
The live configuration panel shows your agent's settings as the builder works through your requests. You can see changes straight away and review or adjust supported fields before publishing.
The panel shows key agent settings, including:
Instructions: Define the agent's behaviour and communication guidelines.
Triggers: Decide which events start the agent.
Tools: Give access to supported actions and integrations.
Skills: Let the agent perform specific tasks.
Knowledge Bases: Provide information the agent can refer to.
As the builder processes a request, instructions stream into the panel and configuration elements fill in as they're resolved. You can also edit supported fields directly instead of sending another request.
Smart Cascades and Instruction Synchronisation
Smart cascades keep an agent's instructions in line with its configuration automatically. When you change a trigger, plugin or Knowledge Base, the builder updates related instructions where needed and keeps unrelated settings as they are.
For example, enter:
"Replace the current Knowledge Base with the Support Procedures Knowledge Base."
The builder updates the selected Knowledge Base and adjusts related instructions to match. In the same way, changing a trigger can update the instructions about when and why the agent runs.
Smart cascades apply to:
Triggers: Keep instructions in line with the events that start the agent.
Plugins: Reflect changes to available integrations and capabilities.
Knowledge Bases: Keep instructions consistent with the agent's information sources.
Review the updated configuration before publishing to make sure the changes match the behaviour you want.
How to Use the Guided Managed Agent Builder
The builder lets you create or edit an agent, review changes as they happen, and test its behaviour before publishing. Following these steps helps make sure your agent is set up correctly while keeping existing settings in place.
Step 1: Open the Managed Agent Builder
Log in to your ZipLeads account.
Go to AI Agents → Agent Studio.
Open the Managed Agents tab.
Create a new Managed Agent, or open an existing one to edit.
Step 2: Enter Your Instructions
In the builder's conversational composer, describe what you want to create or change.
For example:
"Add a form submission trigger for the Demo Request form without changing the existing tools or Knowledge Base."
Submit your request.
Step 3: Answer Clarifying Questions
If the builder asks for clarification, choose the right resource or give the missing information. The builder then carries on applying your changes.
Step 4: Review and Adjust the Configuration
Watch the changes appear in the live configuration panel. Check the updated instructions, triggers, tools, skills and Knowledge Bases.
If needed, make more requests or edit supported fields inline.
Step 5: Test and Publish
Use the Test option to check the agent's behaviour.
Confirm your requested changes work as expected.
Check that unrelated settings and capabilities haven't changed.
Make any adjustments you need.
Select Publish when the configuration is ready.
Frequently Asked Questions
Q: Do I need to turn on the guided-builder improvements?
No. They're available for Managed Agents in Agent Studio with no extra setup.
Q: Can I ask questions without changing my agent?
Yes. The builder tells the difference between questions and editing requests, and answers questions without intentionally changing the configuration.
Q: What happens if the form or resource I asked for can't be found?
The builder asks for clarification rather than using an invalid or made-up resource. Check that the resource exists in your account.
Q: Can I edit settings directly without using prompts?
Yes. The live configuration panel supports inline editing of available fields.
Q: Will changing a trigger or Knowledge Base overwrite all my instructions?
No. Smart cascades update related instructions as needed and keep unrelated settings as they are.
Q: Does using account-grounded resources guarantee my agent will work correctly?
No. Account-grounded selection helps prevent invalid resource references, but you should still review and test your agent before publishing.


