Useful guidance, drafts or instructions. Your team still has to turn the response into software.
See how a Roth conversion tool
takes shape.
A chat can answer a question. Studio goes further: its agentic harness coordinates the work required to build and refine a working application around your firm’s preferences, process and approval requirements.
Specialized agents use approved tools and project context to plan, build, check and refine the application for your review.
The difference is what happens
behind the conversation.
Studio does not stop at a text response. It carries your firm’s preferences and workflow into a structured build process, keeps the project context for follow-up requests, and returns a working application your team can inspect and refine.
“Add a required ‘Planning priority’ field to our intake form, and display it prominently at the beginning of the client report.”
A chat-only experience can be useful for explanations and drafts. In the example below, that experience ends with text. Studio’s agentic harness coordinates the work behind the request and returns the changed application for review.
Useful guidance.
Your tool stays the same.
The reply explains the change but does not make it.
Sent to the language model as a prompt.
Suggests how the field and report could be changed.
“Add a required field to the intake and place its value beneath the report masthead.”
Someone still has to build and check the change.
Your process, carried into
a working application.
Studio breaks the request into tasks and assigns them to specialized agents. The plan carries forward your firm’s preferences, workflow and review requirements.
Studio reviews the current intake, report and requirements.
Finds where the change belongsStudio adds the Planning priority field and connects its saved value to the report.
Makes the requested changeStudio checks the changed form and report against your request.
Checks the required field and report placementChanges that need correction return for refinement.
Studio makes changes through approved tools and security rules. Each tool receives only the access needed for its job.
A required intake field. A connected report section. Your team inspects, refines and tests the result against the way your firm works.
Review, Build and Check illustrate the roles used for this request. The exact path can vary by the application and the change you request.
See the tool.
See what your team reviews.
This fictional Northstar example shows Studio adding a planning field, preparing a suggested goal from relevant client information, and carrying the saved wording into a client report.
“Add a required ‘Planning priority’ field to our intake form, and display it prominently at the beginning of the client report.”
“Add a button to draft or refine the goal using the selected client’s records.”
Prepare a Roth conversion analysis.
Select a client with records to suggest a goal. No client is selected while you shape the tool.
Empty until wording is saved. The saved value supplies this section.
Setup: Elliot Marlowe owns the application; Naomi Calder is the required launch reviewer in this example. Selection is not approval.
Interactive illustration. Northstar, its people, and its records are fictional. Numerical outputs are a historical sample, not new calculations. Technical annotations explain the workflow; they are not a production execution log.
Approve the tool.
Put it to work.
Launch approval makes the tool available for the team to use. From there, each person reviews the reports they prepare.
The tool is approved for use.
The designated reviewers approve the tool version—what it does, how its calculations work, its report format and its workflow.
Both launch decisions are required in this Roth example. Reviewer selection during setup is not approval.
The user reviews each report.
The person using the approved tool reviews the inputs, assumptions and resulting report before using it with a client.
The report does not return to the tool’s launch reviewers each time it is created.
Ask to see how it works.
If your team wants to evaluate whether a product is truly agentic, ask for evidence of the work behind the chat—not only a polished conversation.
01What happened beyond the prompt?
Follow a request through orchestration and tool actions. Match the resulting application changes to the intake field and report section.
02How did the firm’s process shape the application?
Compare the firm’s preferences, workflow and review requirements with the application Studio prepared. Check that the result reflects the practice rather than a generic template.
03What could the model access?
Inspect tool permissions, the returned client context and the model request. Check how sensitive identifiers are excluded and access is scoped to the selected client.
04What persists, what is checked, and who approves?
Follow a second request against the existing application. Inspect saved project context, relevant checks, correction paths and the designated reviewers who authorized release.
Chat is the interface. The harness is the operating structure.
A chat window is one way to make a request. What makes Studio agentic is the coordinated work behind it: Studio keeps the project context, assigns specialized work, uses approved tools to build and check the application, and returns the result for human review.