AI Agents help you complete repeatable CRM work with AI assistance. An agent can review relevant Wealthbox data, follow a set of instructions you give it, and produce a result for you to review, so routine work gets done in the background while you focus on your clients.
Introduction to AI Agents
Agents are designed to support your work, not replace your judgment. You should always review important outputs before acting on them.
Unlike AI Assistant, which you actively ask questions or give commands to, Agents run largely on their own, either on a schedule, when triggered by a CRM event, or on demand, so you don't have to remember to prompt them each time.
Availability
AI Agents is available to Organization Admins and Account Owners on eligible plans.
Organization Admins can extend access to other users in their organization by enabling the Create and Use Personal Agents privilege. Once enabled, those users can create and manage their own agents within the workspaces they already have access to.
Availability may vary depending on your firm's plan and settings.
If you don't see AI Agents in your account, check with your Organization Admin or Account Owner to confirm whether access has been enabled.
Accessing AI Agents
To access Agents, click AI Agents from the left-hand navigation. You'll see your existing agent, AI Assistant, alongside a "Create your first agent" option to get started.
Creating an Agent
You can create an agent in one of two ways.
If starting from scratch, describe the goal you want the agent to accomplish, and Wealthbox will create a draft agent for you to review. Examples of useful starting prompts include: "Help me prioritize the upcoming work that is assigned to me," "Identify clients who have not been contacted recently and suggest follow-up," or "When a new client contact record is created, start the onboarding workflow and assign me a task to follow up."
You can also start with an out-of-the-box agent, built for common advisory and operations workflows, including planning, client monitoring, follow-up, outreach, CRM hygiene, reporting, pipeline review, and other common practice activities. Selecting one creates a version you can review and tailor to your own workflow.
Reviewing Your Agent
Before an agent is created, the Review your Agent page gives you a chance to look it over. From here, you can rename the agent, modify its description, change its instructions, or adjust its triggers before enabling it.
Understanding the Agent Page
Once created, the Agent Show page is where you run, review, and manage an individual agent.
Agent Information Panel
The information panel includes the agent's name, its category, a description of what it's intended to do, its instructions, and its current status. The panel can be opened and closed using the icon in the corner.
Instructions
Instructions tell the agent what it should do each time it runs. You can edit an agent's instructions at any time from the dropdown in the Agent Info Panel to better reflect your workflow, preferred output, or business process. When updating instructions, be specific about the outcome you want, the Wealthbox information the agent should consider, the format of the desired result, and any steps the agent should or should not take.
Recent Chats
Each time an agent runs, Wealthbox creates a record of that session. Review an agent’s recent chats to see its prior conversations and results, including whether a run is in progress, completed successfully, or encountered an error.
Use recent chats to understand what the agent did, review its output, and troubleshoot unexpected behavior.
Triggers
Triggers determine when an agent runs. You can configure an agent to run manually, on demand, whenever you choose to run it; on a schedule, at a selected time or cadence; when a CRM event occurs, for supported record changes and activity; or when a date field condition is met, for supported upcoming, due, or overdue date based activity tied to relevant date fields. For supported event and date based triggers, you can also narrow when an agent runs by adding filter criteria.
An agent with an enabled event or date based trigger cannot be run manually, since it requires the context supplied by that trigger.
Human in the Loop
Wealthbox AI Agents are designed to keep users in control of the work they perform.
Human in the Loop is the default approval model. When an agent proposes an action that requires approval, the user reviews it before the action proceeds.
Where organization policies permit, users can grant Standing Approval for supported actions. This allows an individual agent to perform those actions without requesting approval each time.
Human in the Loop: Review Before Action
With Human in the Loop, the agent prepares a proposed action and presents it for review.
You can approve it, decline it, or provide a correction so the agent can revise its proposal.
For example, an agent might prepare a follow-up task with a proposed description, assignee, and due date. You review the proposal before the task is created.
Approval requirements depend on the action. Reading available information, searching records, and preparing responses do not require the same approval steps as changing CRM data.
Organization Administration: Set the Boundaries
Organization approval policies define where users may grant Standing Approval.
Creating CRM records and updating CRM records are controlled separately.
The available organization-level approval settings include:
| Capability | Approval Setting |
|---|---|
| Creating CRM records | Ask every time or allow Standing Approval |
| Updating CRM records | Ask every time or allow Standing Approval |
| Sending emails | Blocked |
| Deleting CRM records | Ask every time |
Note: Sending emails and deleting CRM records are not configurable at this time. They will always show as Blocked and Ask every time, respectively.
Because the Create and Update sections are separate, an organization can allow users to delegate record creation while continuing to require review for record updates.
Additional restrictions from an affiliated parent organization may also apply. The highest level of restriction will cascade down to the organization and then to the user.
Important: Allowing Standing Approval does not automatically enable it for users. It makes the choice available within the organization's boundaries.
User Controls: Choose What Each Agent May Do
Within those boundaries, an agent's creator can grant or revoke Standing Approval separately for:
- Creating CRM records
- Updating CRM records
These permissions belong to the individual agent. Granting Standing Approval to one agent does not grant it to every agent in the user's account.
For example, a user could allow a meeting follow-up agent to create tasks without repeated approval while keeping approval required for changes to existing records.
Another agent could continue to require review for both categories.
Users can revoke a grant to return that category to approval before action. Standing Approval also remains subject to applicable organization restrictions, even if the user granted it before those restrictions were introduced.
How the Controls Work Together
Standing Approval depends on both organizational permission and the user's choice.
It is limited to supported action categories. A grant for creating or updating CRM records does not authorize other categories, such as deleting records or sending email, and does not expand the user's underlying access permissions.
Organizations set the boundaries, and users choose how much to delegate within them.
Agent Status
| Status | Description |
|---|---|
| Draft | The agent has been created but is not active. Draft agents do not run automatically. |
| Active | The agent is available to run according to its configuration. |
| Paused | The agent is essentially disabled. It cannot be triggered in this state, but can be edited and enabled again. |
| Deleted | The agent is no longer in active use, but can be found in your account's Trash Can if it needs to be revived. |
Explore Out-of-the-Box Agents
Wealthbox offers a growing library of out-of-the-box agents designed for common advisory and operations workflows.
These agents can help with activities such as client monitoring, follow-up, CRM organization, pipeline management, and other practice management tasks.
To explore the available agents and learn more about what each one can do, visit:
Note: Available agents and capabilities may change over time.
Frequently Asked Questions
What is an AI Agent?
An AI Agent is a configurable Wealthbox assistant designed to help with repeatable CRM work. It uses instructions, available tools, and, when configured, triggers to complete a task and provide a result for you to review.
Can I create my own agent?
Yes. You can create an agent by describing the work you want it to perform, then reviewing and editing the resulting configuration.
Can I start with a pre-built agent instead?
Yes. You can select a pre-built agent for a common use case, then review and tailor it to your workflow.
What is a trigger?
A trigger determines when an agent runs. Depending on the agent's configuration, it may run manually, on a schedule, after a supported CRM event, or when a supported date condition is met.
Why can't I run my agent manually?
Agents with enabled event or date based triggers require the context of the triggering record or date condition, so they cannot be run manually without it.
Does an agent replace compliance review or professional judgment?
No. Agents support your work but do not replace your firm's policies, supervisory process, professional judgment, or compliance review. You should review outputs before acting on them.
Can every customer use AI Agents?
Not yet. AI Agents is currently in early access, so availability depends on your firm's plan and whether access has been enabled for you.
Troubleshooting
I can't find AI Agents in Wealthbox.
AI Agents may not be enabled for your user profile or firm. Check with your Organization Admin or Account Owner to request access.
My agent did not run.
Check the agent's status and trigger configuration. Draft agents will not run automatically, and triggers must be created and enabled to fire.
My agent's result was not what I expected.
Review the agent's instructions, available tools, trigger conditions, and prior run history. Update the configuration as needed, add more specificity where helpful, and test again.
My agent is asking for approval before taking action.
Review the agent's approval permissions and your organization's approval policy. Certain actions always require approval or may be blocked.