Ariath glossary: Agentic AI
AI systems that work toward a goal across several steps, deciding which actions to take and which tools to use along the way, rather than producing a single response to a single prompt.
In brief
- What it does
- Works toward a goal across several steps, choosing actions and tools along the way.
- How it differs from generative AI
- Generative AI responds to a prompt. Agentic AI pursues a goal through a series of actions.
- What it needs in a wealth management firm
- A defined scope, limits on actions, human review for sensitive steps, and records of what it did.
How it works
How agentic AI works
An agentic system typically works toward a goal in a loop:
Set the goal
The system receives a goal from a person or a process, along with the limits it must work within.
Plan the steps
It breaks the goal into steps and chooses which tools or data to use for each.
Act and use tools
It carries out the steps, such as searching records, drafting content or passing a request to another system.
Review and adjust
It checks the results, changes the plan if needed, or stops and asks a person.
Because the system chooses its own steps, the boundaries around it matter as much as the model behind it.
The more autonomy a system has, the more important clear limits, monitoring and human review become.
Comparison
Agentic AI vs. generative AI
The two are related, and they describe different things.
Generative AI
Produces content, such as text or a summary, in response to a prompt. A person decides what to do with the output.
For example
Drafts a summary of a meeting transcript when asked.
Agentic AI
Works toward a goal across several steps, choosing actions and tools along the way, with limited human input.
For example
Gathers records from several systems, drafts the summary, and queues it for advisor review.
They are not opposites. Agentic systems often use generative AI to reason and write, with added components for planning, tool use and oversight.
In practice
Agentic AI examples in wealth management
In a wealth management firm, an agentic workflow might gather information from several systems and prepare work for a person to review. These examples are illustrative.
Annual review preparation
Collecting account, planning and CRM information into a single brief for the advisor.
Account opening support
Tracking required forms across steps and flagging what is still outstanding.
Document intake
Reading incoming documents, extracting key fields and proposing where they belong for staff to confirm.
Reconciliation support
Comparing records across systems and listing differences for staff to resolve.
Task coordination
Breaking a request into steps and routing each one to the right person or system.
Because the system takes actions, firms typically define what it is allowed to do, which data it can reach, what gets recorded, and where a person must review or approve the result.
What it needs
What agentic AI needs to work with firm data
A system that acts on client data needs limits around it. Safeguards often include:
- A defined scope for each workflow
- Role-based access to data
- Limits on which actions it can take
- Human approval for sensitive steps
- Records of actions and decisions
- The ability to pause or stop it
These are firm decisions, set by the firm's own policies, and they sit between the firm's data and the AI model or tool.
Common questions
Frequently asked questions about agentic AI
How is agentic AI different from generative AI?
Generative AI produces content, such as text, in response to a prompt. Agentic AI pursues a goal through several steps and actions, and it often uses generative AI as one part of that process.
Does agentic AI work without human oversight?
It can be designed to act with limited human input, but the firm chooses how much autonomy to allow and where a person reviews or approves the result.
How is agentic AI related to an AI agent?
An AI agent is the component that performs a task. Agentic AI describes the broader behavior of pursuing goals with limited human input.
This entry is educational. It is not legal or compliance advice and does not guarantee regulatory compliance. Confirm the requirements that apply to your firm with your compliance lead or counsel.
Give AI the context it needs to act
Ariath connects your CRM, custodian and planning tools into one governed context, with your firm in control of how data is accessed and used.