voice-ai

Voice AI Tutorial: How to Use Granular Permissions

Charles Higgins

Charles Higgins

— 12 min read
Featured
Featured

We watched a clear, hands-on walkthrough by Andrew George that explains how to use granular permissions for Voice AI inside the platform. In this article we’ll take his step-by-step approach further: we’ll explain exactly what each permission does, show practical examples for different team roles, outline best practices, and give a checklist you can use to set up secure, efficient access for your team.

Our goal is to help small teams and busy business owners get the most value from Voice AI while keeping sensitive information protected and workflows simple. We’ll explain the four core permission types, show how they interact, offer recommended permission sets for typical roles, and walk through testing and troubleshooting so you can deploy this without guesswork.

Table of Contents

Why granular permissions for Voice AI matter

Giving everyone full access feels easier at first but quickly creates problems: accidental configuration changes, exposed goals, inconsistent bot behavior, and security risks. Granular permissions let us control who can see and change what, so we can keep operations running smoothly while minimizing headaches.

  • Protect sensitive data: Not every team member needs to see agent goals or dashboard analytics. Limiting access reduces the risk of leaks or accidental edits.
  • Reduce errors: When only the right people can edit agents, we avoid broken bots, overwritten goals, or misconfigured call handling.
  • Improve accountability: Clear roles mean it’s obvious who is responsible for creating, updating, or troubleshooting an agent.
  • Simplify onboarding: New users see a clean interface with only the tools they need. That speeds training and reduces frustration.
  • Streamline work: Teams can focus on results—writing good goals, monitoring performance, or handling calls—without fighting permissions issues.

Overview: The four permission types

There are four core permission toggles that control access to Voice AI capabilities. Understanding these is the first step to setting up secure, efficient workflows.

  • View and manage voice AI agents: Allows users to see the list of agents, open individual agents, and perform management actions—edit, duplicate, and delete. It gives access to agent details and phone availability settings.
  • View voice AI agent goals: Grants read-only access to the goals set for each agent. Users can see what the agent is trying to accomplish but cannot change those goals.
  • View and manage voice AI agent goals: A higher level of access that allows users not only to view goals but also to create, edit, and delete them. This permission is often reserved for managers or specialists who design bot behavior.
  • View voice AI dashboard: Permits users to access the dashboard that shows analytics and high-level performance metrics for agents. Dashboards often contain behavioral and performance signals we may want to keep limited to analysts or managers.

Step-by-step: Setting granular permissions

Here’s a practical walkthrough for setting permissions, modeled on the demo in the original tutorial. We’ll keep each step clear so you can follow it on the platform without missing anything.

  1. Open Roles and Permissions: From your admin area, navigate to the Roles and Permissions section. This is where you assign access levels for each role or user.
  2. Find AI Agents Permissions: In the permissions list, look for the section related to Voice AI agents. The four toggles appear near the top of the section.
  3. Toggle the permissions you want: Click the checkboxes to enable or disable each capability for the role you are editing. Typical options include the four described above.
  4. Save your changes: Always press Save after changing permissions. Forgetting to save is a surprisingly common source of confusion.
  5. Test as the user: Use the "log in as" or "impersonate" feature (if your platform has it) to see exactly what that role can access. If you don’t have impersonation, create a test user with that role and log in as them.
  6. Adjust and repeat: If the test account can see or do too much (or too little), return to Roles and Permissions, make small changes, save, and test again.

That’s the basic flow. Now let’s walk through the most common permission combinations and what you’ll actually see when you test them.

What each permission combination looks like in practice

Testing permissions helps ensure they behave the way we expect. Here are the real outcomes you’ll observe for a few common combinations.

Only "View and manage voice AI agents" enabled

What the user sees:

  • The full list of voice AI agents.
  • Three-dot menu actions on each agent that allow Edit, Duplicate, and Delete.
  • Ability to open an agent and change agent details like phone availability.
  • No access to agent goals or dashboard analytics.

When to use this: Ideal for technical operators who need to maintain agents—duplicate templates, update availability, or fix agent settings—without changing high-level goals or viewing analytics.

Only "View voice AI agent goals" enabled

What the user sees:

  • Read-only access to the goals assigned to agents.
  • No ability to edit or delete goals, and no access to agent editing tools.
  • No dashboard access.

When to use this: Useful for team members who need context about what each agent is trying to achieve (for example, sales reps or customer success teams) but should not change the behavior of the bot.

"View and manage voice AI agent goals" enabled (with or without other toggles)

What the user sees and can do:

  • Create, edit, and delete agent goals.
  • If they also have "view and manage agents," they can modify the agent details and how those goals are applied.
  • If they also have dashboard access, they can link performance data back to goals and iterate improvements.

When to use this: Reserved for AI managers, bot designers, or senior team members who should own the logic and outcomes of agents.

"View voice AI dashboard" enabled

What the user sees:

  • Analytics like call volume, success metrics, conversion signals, or other performance indicators shown on the dashboard.
  • Depending on other toggles, they may be able to drill into specific agents or just consume top-level data.

When to use this: Great for analysts, executives, or team leads who need to monitor performance without editing agents or goals.

Here are practical permission templates you can copy. Use these as starting points and adjust them to your team’s needs.

System Administrator (full control)

  • View and manage voice AI agents: On
  • View voice AI agent goals: On
  • View and manage voice AI agent goals: On
  • View voice AI dashboard: On

Who this fits: Owners and lead admins who set policy, oversee billing, and resolve platform-level issues.

AI Manager / Bot Designer

  • View and manage voice AI agents: On
  • View voice AI agent goals: On
  • View and manage voice AI agent goals: On
  • View voice AI dashboard: Optional (On recommended)

Who this fits: People responsible for writing goals, testing bots, iterating behavior, and linking tactics to outcomes.

Support Agent / Call Handler

  • View and manage voice AI agents: Optional (Recommended: limited to assigned agents only if your platform supports that granularity; otherwise Off)
  • View voice AI agent goals: On
  • View and manage voice AI agent goals: Off
  • View voice AI dashboard: Off or Read-only if available

Who this fits: Frontline employees who need to understand agent intent and assist leads, but should not change the agent’s behavior.

Analyst / Manager

  • View and manage voice AI agents: Off
  • View voice AI agent goals: On
  • View and manage voice AI agent goals: Off (unless they also create experiments)
  • View voice AI dashboard: On

Who this fits: Data-focused roles that track performance and make recommendations to the AI Manager or System Admin.

Contractor / Outsourced Specialist

  • View and manage voice AI agents: Off
  • View voice AI agent goals: Off or On depending on need
  • View and manage voice AI agent goals: Off
  • View voice AI dashboard: Off

Who this fits: External contractors who shouldn’t have any platform control unless explicitly engaged for a specific task.

How to test permissions safely

Testing permissions prevents surprises. Here’s a short checklist to verify that roles are configured properly:

  1. Create a test user or use an impersonation utility to assume the role you just edited.
  2. Confirm what the user can see on the Agents page: agent list, three-dot menus, and agent details like phone availability.
  3. Open an agent to see whether goals are visible and whether they can be edited.
  4. Check the dashboard view to ensure it displays analytics only if intended.
  5. Try actions like Edit, Duplicate, and Delete if the role should have those permissions. If they can perform an action they shouldn’t, return to Role settings and disable the toggle.
  6. Log out and log back in as the test user after any permission change to ensure the system applies updates.

We recommend running tests on real tasks—duplicating a template, editing phone availability, or opening goals—because UI labels can be ambiguous until you try them.

Best practices when managing Voice AI permissions

Follow these practical guidelines to keep things controlled and efficient.

  • Apply the principle of least privilege: Start with the minimal permissions a person needs and add access as required.
  • Document role definitions: Write a short one-line purpose for each role so everyone knows why that access exists.
  • Use naming conventions: Label agents and goals clearly so limited-access users can quickly identify what belongs to them.
  • Review permissions periodically: Quarterly audits help catch orphaned accounts and outdated access.
  • Train users: Short training sessions reduce accidental edits and improve collaboration between managers and operators.
  • Keep one logging/accountability path: If the platform provides logs, make sure managers review changes to agent goals and configuration.
  • Use templates and cloning: Allow operators to duplicate standard agent templates rather than building from scratch—this reduces configuration drift.
  • Separate edit and publish roles when possible: Let certain users draft changes while another person approves and publishes them into production.

Troubleshooting common permission issues

When things don’t behave as expected, these quick checks usually resolve the problem.

  • Changes not applied: Confirm you pressed Save and then log out/in as the affected user. Some platforms require a fresh session.
  • User can see but not edit goals: That’s expected when "view goals" is enabled but "manage goals" is not. Decide whether they should be allowed to edit.
  • User can edit agents but can’t access goal analytics: If they need analytics, enable dashboard access. If not, leave it off.
  • Actions missing from three-dot menu: Check the "view and manage agents" toggle. If it’s off, those options will not appear.
  • Dashboard shows empty data: Make sure the user has access to the specific agents or data groups feeding the dashboard. Permission gaps can block analytics.
  • Unexpected errors after cloning: Sometimes cloned agents carry over settings or connections that the cloning user doesn’t have access to. Check integrations, phone numbers, or linked resources.

Building bots and updating permissions

As we build new agents and experiment with goals, permissions will likely need to change temporarily. Here’s how to manage that process without risk.

  1. Create a sandbox environment: If possible, use a test account or sandbox workspace to prototype bots and goals.
  2. Assign temporary permissions: Give collaborators the minimum extra access they need, and set a calendar reminder to revoke it after the task is complete.
  3. Use templates: Build new agents by duplicating a standard template so settings remain consistent.
  4. Review goals before publishing: Have a manager or owner review goal changes before they go live to prevent unintended behavior.
  5. Monitor after launch: Use dashboard metrics to validate that the new or updated agent behaves as intended.

By controlling who can create or modify goals, we avoid misaligned objectives and preserve a single source of truth for agent behavior.

Security and compliance considerations

Permissions are a key part of a broader security posture. Here are practical ways to minimize risk:

  • Limit access to goals and dashboards: Goals can reveal business strategy. Restrict write access to a few trusted team members.
  • Monitor account activity: Regularly review audit logs to spot unusual activity or unapproved changes.
  • Use multi-factor authentication: Enable it for accounts that have management access to reduce the risk of compromise.
  • Rotate contractor access: Treat external contractors as temporary roles; revoke their permissions immediately after the project ends.
  • Archive old agents: Instead of deleting, consider archiving agents. That preserves history without exposing live behavior to everyone.

Real-world examples: Using permissions to solve common problems

These short scenarios show how granular permissions make life easier for busy teams.

Scenario 1: Preventing accidental goal changes

Problem: A front-line staff member accidentally edited an agent goal and reduced call routing efficiency.

Solution: We disabled "manage agent goals" for front-line staff while leaving "view goals" on. Only the AI Manager retained goal-editing rights. Result: fewer accidental changes and clearer ownership of agent behavior.

Scenario 2: Giving contractors controlled access

Problem: A contractor needed to duplicate and adjust agent scripts but shouldn’t change live goals or see analytics.

Solution: We enabled "view and manage agents" so the contractor could duplicate templates and adjust phone availability. We kept both "manage goals" and "view dashboard" disabled. Result: the contractor completed their task without access to sensitive analytics or strategic goals.

Scenario 3: Creating an iterative workflow

Problem: Designers were making changes directly in production and it caused regressions.

Solution: We introduced a draft-and-publish workflow: designers had "manage goals" but only in a staging environment; only the AI Manager had publish rights. Result: safer rollouts and higher confidence in changes.

"When we tightened permissions, we stopped seeing the ‘mystery’ edits that used to break our bots. It saved hours every week." — a team member

Checklist: Quick setup for the first day

Use this checklist to get permissions set up correctly the first time.

  1. Define roles and responsibilities for every person who will touch Voice AI.
  2. Start with the least privilege for each role and add only what is necessary.
  3. Enable impersonation or create test users to validate each role’s view and capabilities.
  4. Document any temporary permissions and set reminders to remove them.
  5. Train users on where to find goals, how to duplicate agents, and what to do if they need a change.
  6. Set an audit cadence (monthly or quarterly) to review who has access.

Conclusion

Granular permissions for Voice AI give us control, clarity, and safety. By carefully choosing who can view and manage agents, who can see or edit goals, and who can access dashboards, we reduce risk and let our teams focus on what matters: building better customer experiences and improving outcomes.

We started with the simple four toggles—manage agents, view goals, manage goals, and view dashboard—and used them to create real-world role templates, test procedures, and best practices. With a few minutes of setup and a short testing cycle, you can make sure your team has just the right level of access.

We recommend starting by defining three or four roles for your organization—Admin, AI Manager, Agent, and Analyst—then applying the templates above. Test each role in a sandbox environment if possible, and keep a routine review to ensure permissions evolve with your business.

Thanks to Andrew George for a clear walkthrough that inspired this practical guide. Use the tips here to protect your data, reduce accidental errors, and help your team move faster and more confidently.

FAQ

Q: What’s the minimum permission set a basic support agent needs?

A: Typically, a basic support agent only needs "view voice AI agent goals" so they understand the bot’s objectives. We usually keep "manage agents" and any goal-management toggles off to avoid accidental edits. If they need to update phone availability or work with assigned agents, consider enabling "view and manage voice AI agents" for those specific agents only, if your platform allows per-agent controls.

Q: Can we allow an employee to edit agents but not delete them?

A: Many platforms group edit, duplicate, and delete under the same manage permission. If your platform does not separate those actions, the safest approach is to give edit/duplicate rights only to trusted roles and reserve delete rights for admins. Alternatively, use confirmation workflows or an approval-required process for deletions.

Q: How do we handle contractors who need temporary access?

A: Create temporary roles or give contractors the minimum necessary toggles and set calendar reminders to revoke access when their work is complete. If possible, use a sandbox environment so they can do work without touching production agents.

Q: Why can a user see agents but not goals after we enabled permissions?

A: That’s usually because "view and manage voice AI agents" was enabled but "view voice AI agent goals" or "view and manage voice AI agent goals" was left off. If they need to read goals, enable the read-only goals toggle. If they also need to edit goals, enable the manage-goals toggle.

Q: Should analysts have edit access to goals?

A: Not by default. Analysts should typically have dashboard access and read-only view of goals so they can report on performance. If analysts are running controlled experiments, then temporary edit rights in a staging environment are appropriate, with publishing reserved for a manager.

Q: How often should we review permissions?

A: A quarterly review is a good baseline. If your team changes quickly or you have many contractors, monthly reviews might be necessary. Include permission audits as part of your routine security checks.

Q: What if we accidentally delete an agent?

A: Check whether the platform has an archive or trash feature. If it does, restore from archive. If not, we recommend applying an additional safeguard: only allow a small number of trusted admins to delete agents and require confirmation steps or an approval workflow.

Q: Is there a standard way to name agents and goals to keep things organized?

A: Yes. Use a consistent naming convention that includes purpose, target audience, and version (for example: "Sales-LeadQual-v2"). Include the date or owner in metadata fields when possible. This makes it easier for people with limited access to identify what belongs to them.

Q: How do we ensure changes are safe before going live?

A: Use a staging or draft environment, run tests with mock calls, and require a peer or manager to approve changes before publishing. Monitor the dashboard closely after launch to detect regressions quickly.

Q: Who should own the "manage goals" permission?

A: Ideally, a small team of AI Managers or Bot Designers should own this permission. These are the people who understand business objectives, can write clear goals, and can interpret performance data to make iterative improvements.

Read more