AI governance at Clark
AI work your team can review
Ava works within Clark's audit and accounting platform. Your team defines the task and operating boundaries before work begins, while supporting evidence, exceptions, and human review stay connected throughout execution.
Defined scope
Set the boundaries before work begins
Before work begins, Ava shows the task, planned actions, systems involved, permissions, and review requirements. Your team can approve, adjust, or cancel that plan. Ava can then complete the task within those approved boundaries without requiring another instruction at every step.
Connected evidence
Follow conclusions back to evidence
AI-drafted figures and conclusions remain connected to supporting source documents. Reviewers can move from an answer back to its evidence, checking the support alongside the work being reviewed. Clark keeps source evidence, conclusions, and review in the engagement record.
Human judgment
Keep professional judgment with the team
Finished work follows the firm's preparer-to-partner sign-off process. Ava can execute the scoped task autonomously, but exceptions, proposed conclusions, and supporting evidence remain inputs to professional judgment, not a substitute for it.
What to check with your team
- Who reviews proposed actions and resolves exceptions on the engagement?
- What evidence does each reviewer need before accepting the work?
A shared record
Keep approvals in the engagement record
Approvals and sign-off are recorded in the engagement workflow. Current event, retention, and export details are supplied during review, so your firm can assess the record against its own documentation requirements.
What to check with your team
- Which workflow events and approvals are included in the record?
- What retention and export details apply to your firm's deployment?
Data boundaries
Keep client data out of model training
For the Clark product, firm and client data is not used to train models or shared for model learning across firms. Hosting, subprocessors, and deployment-specific data handling are covered separately in the security review.
Explore data securitySee Clark in action
The peer review guarantee, in practice
Watch Ava prepare a workpaper, then trace every figure back to the source document, the change, and the approval.