Caseware launches audit agent that flags gaps across engagement files

By Sophie Lindqvist

Early testers of Caseware’s Engagement Review Agent reported savings of roughly three to four hours per audit engagement, CEO David Marquis said. Caseware launched the product on Nov. 14, 2026, initially for customers in the Netherlands. The company describes engagement review as one of the most time-consuming stages of an audit.

Marquis tied the product’s goal to the nature of audit work:

“Audit quality depends on understanding how every part of an engagement fits together.”

He then described what early testers saw:

“Our early testers remarked that the difference was immediate. Caseware Verity already has the engagement context, so they review a purpose-built agent's work instead of prompting from scratch. It flags conflicts across workpapers and missed items, saving roughly three to four hours per engagement.”

Caseware’s announcement attributes that figure to early testers. The company also places the agent in a supporting role: auditors assess its findings and retain every decision and approval.

What does the Engagement Review Agent check?

Caseware says the agent sits within an audit engagement and examines linked documents across workpapers, checklists and financial statements. In that setup, it surfaces inconsistencies, missing information and other issues rather than changing the underlying record.

The agent can review a whole engagement, selected phases or individual workpapers. Caseware says it flags:

• Conflicts between planned procedures and documented conclusions
• Checklist responses that do not match supporting workpapers
• Missing references
• Incomplete conclusions
• Possible discrepancies in materiality calculations
• Absent sign-offs
• Exposure of sensitive personal information

The list combines substantive conflicts with process controls. Caseware presents these findings as potential issues for auditors to assess, not as final conclusions about the engagement.

The basic division is simple: the agent finds possible problems, and auditors decide what they mean.

Why is engagement review so hard to check?

Caseware says reviewers must test whether risk assessments match planned procedures, whether materiality calculations are supported and whether conclusions remain consistent across connected files. Reviewers also need to determine whether later work supports earlier judgments.

The release targets those relationships across files rather than the contents of a single document. Cross-file consistency is the stated problem, and the agent reports mismatches for an auditor to investigate.

Caseware says late changes can create gaps or contradictions that are difficult to trace across dozens of documents. That can leave engagement teams spending significant time reconstructing which version supports a conclusion and which version no longer does.

According to Marquis, embedded engagement context is what separates Verity from a system prompted from scratch. Auditors receive a purpose-built agent’s findings rather than starting with an empty prompt, which he linked to the reported time savings.

For a late-changing audit, the hard part is linking one document to another; that is the job Caseware assigned to the agent.

How does Caseware filter out resolved findings?

Before reporting an issue, Caseware says the agent searches related workpapers, later conclusions and existing review notes. That check can find evidence that an apparent problem was resolved elsewhere in the engagement file.

Findings appear in a consolidated report that points users to relevant workpapers and supporting evidence. Auditors can move directly to the underlying documents to verify each observation and decide whether further investigation is required.

The workflow keeps the source documents in view and asks the engagement team to decide what still needs attention. It does not treat every detected inconsistency as an unresolved finding.

The consolidated report points to evidence; it does not replace the reviewer’s judgment.

Where does the agent run, and who approves?

The agent runs in read-only mode within Caseware Cloud. Users can start it through Verity Chat or from an engagement phase. Caseware says it does not alter workpapers, approve documentation or make audit decisions.

It can also work alongside other Verity functions. Caseware says the agent can turn review questions into client queries and convert items for the engagement team into issues in the engagement file.

Those capabilities narrow the agent’s role to finding and organizing possible problems. The engagement team retains professional judgment, approval authority and responsibility for the final documentation.

Marquis described the intended balance:

“With Engagement Review Agent, we're helping auditors identify issues that might otherwise be difficult or time-consuming to uncover, giving them a more complete view of their work while keeping professional judgment and accountability firmly in their hands.”

AI can point to a missing sign-off, but only the engagement team can approve the engagement.

How does the launch fit Caseware’s AI strategy?

The release extends Caseware’s Verity AI agent suite, which operates inside assurance and financial reporting engagements. It follows the earlier introduction of the Risk Suggestion Agent, broadening Caseware’s use of AI from identifying risk areas to assisting with engagement review.

Caseware serves more than 23,000 firms in over 130 countries. The company positions the Verity line as part of a broader move to embed AI tools into accounting and audit workflows.

The sequence matters for the product strategy. Risk Suggestion Agent focuses on identifying risk areas, while Engagement Review Agent examines the work produced across an engagement. Both operate within Verity rather than as separate tools.

Marquis tied the release to that platform strategy:

“Caseware Verity and the Caseware AI platform bring together our decades of deep domain expertise with the latest agentic technology to create powerful new efficiencies, while maintaining governance and compliance.”

He continued:

“The Engagement Review Agent is another step in our vision to enable a fully agentic audit and financial reporting workflow. Together with our customers, we are transforming how audits get done.”

Caseware is extending Verity from risk identification into cross-file review without giving up human sign-off.

What has to happen next?

Caseware’s announcement gives no tester count, measurement method or independent validation for the three-to-four-hour figure. The reported saving therefore remains an early result supplied by the company, rather than a published performance benchmark.

The next proof point is whether that saving holds beyond the initial Netherlands rollout and across varied engagements. For audit teams, every flag needs to remain traceable to source documents while the engagement team keeps the final say.

For Caseware, the bet is that embedded agents can save review time without weakening accountability; for audit firms, every flag still has to connect back to evidence.