AI Audit Assistant – Image-Based Counting of Granite Blocks for Stock Verification
Problem
In large construction, mining, or export units, granite blocks are stored in open yards in uneven, multi-layer stacks. During statutory audits, internal audits, or management verification, auditors face several challenges: Irregular stacking makes manual counting inaccurate. Front-facing visibility varies due to angles, overlapping, shadows, and weather. Human estimation errors occur, especially when the yard contains multiple similar stacks. No uniform labeling or tagging is available on raw granite, unlike finished goods. Time constraints during physical verification increase the risk of mismatch with stock records. This prompt solves these challenges by allowing the auditor to upload a high-resolution photo of a granite block stack and get an instant, AI-assisted count of only the front-visible blocks, removing the need for manual estimation. It also enhances audit documentation quality by offering consistent, unbiased, image-based stock proof.
Prompt Input
Assume you are conducting a stock audit at a granite storage yard where raw granite blocks are kept in open areas. Using the uploaded photograph of the granite stack, perform a detailed AI-based evaluation as follows: 1. Counting Logic Count only the granite blocks whose front faces are clearly visible. Ignore: Blocks partially hidden behind others Blocks visible only 20–30% from the side Back-row blocks Top-layer blocks unless fully front-facing Any item that is not granite (wood, soil, machinery, vegetation) 2. Visual Analysis Requirements Identify granite block boundaries using shape, texture, and color segmentation. Apply perspective correction to avoid counting duplicate or distorted edges. Remove noise elements such as shadows, reflections, or similar-looking objects. 3. Audit-Specific Checks Provide: Total count of clearly identifiable front-row granite blocks. Mention if the image has: Poor lighting Shadows affecting visibility Occlusions (trees, machines, workers) Angle distortion requiring caution Highlight limitations, if any, in interpreting the image. 4. Deliverable Format Your final answer should include: Exact front-row granite block count Short explanation of the counting logic used Disclaimer if certain blocks are partially visible or doubtful Audit note that can be pasted into working papers
Prompt Output
Your output should contain the following: 1. AI-Derived Count “Visible front-row granite blocks: XX” 2. Short Technical Explanation Example: “Counted only distinct, front-facing rectangular granite faces. Excluded shadow-affected edges and partially visible top-row blocks.” 3. Limitation (if any) Example: “Some blocks on the right side are partially occluded by a tree; hence count is based on fully visible faces only.” 4. Audit Working Note (Auto-Generated) Sample format: “Physical verification performed using AI-based image analysis. Only fully visible front-face granite blocks were considered. The count derived is reliable for front-row verification but does not represent total yard stock.”