United StatesProfessional services2025 – ongoing

Armanino

AI explainer for rejected audit document requests

Armanino built an AI tool to draft notes on rejected audit documents and expects it to cut manual writing time by 65%

WorkingVendor case studyVendor ↗

Problem

When a client sent a document that did not meet audit needs, an auditor had to write a note explaining why and what to provide instead. Each rejected item took 10 to 15 minutes, and a typical audit had 50 to 75 rejected items.

Approach

Armanino's AI lab built a document request explainer on Claude through Amazon Bedrock. The tool drafts a note of about three to four sentences that says why a document was rejected and what the client must supply. The first tool went from idea to production in two weeks, and the lab is working on similar tools for testing and documentation.

Results

65%expected reduction in time spent on manual writing of rejection notesVendor case study[1]

60%expected decrease in follow-up clarifications from clientsVendor case study[1]

90%share of users who said they felt more confident using the AI-assisted featureVendor case study[1]

Takeaway

A narrow writing task that repeats dozens of times per audit was a practical first target for a firm's AI lab. The firm's time and follow-up savings were still expected figures, while the confidence result came from users.

Sources

Armanino builds AI-powered accounting tools with Claude
Anthropic12 Dec 2025Vendor publication

Source summaries

[1]Anthropic12 Dec 2025

Anthropic's customer story describes how Armanino built a Claude-based explainer that drafts notes on rejected audit documents. It gives the earlier manual time of 10 to 15 minutes per item, expected reductions of 65% in writing time and 60% in follow-up questions, and a 90% user confidence figure. The page carries no date; the Internet Archive first saved it on December 12, 2025.

Full details are in the original sources.

Added 4 Nov 2026 · Updated 4 Nov 2026