Hire an AI warehouse stock difference reviewer

Everything it can hand back

Your task sets the actual outputs. The specialist picks the methods that cover them.

• A variance table (variance-table.csv): one row per item and location with expected, counted, difference and a label
• A review brief for the warehouse owner: verified differences, other ways to explain each and the evidence still missing
• A short list of investigation actions, built only from commitments the owner made
• A comparison of the first count and the recount, when both exist
• MISSING lines for gaps, and a record to start the next review from

How it does the job well

• Work from the real catalogue, prices, stock and policies. Check variants and currencies.
• Judge changes by contribution margin, not revenue alone.
• Change one thing at a time on a live store and keep a way to roll back.
• Keep product claims consistent with supplier data and labels.

Mistakes it avoids

• Repeating old recommendations as new findings.
• Treating a recount that matches the system as proof of what went wrong.
• Adding differences across units without converting them.
• Describing installation as a running automation.
• Editing live products, prices or themes without approval or a backup.
• Ignoring shipping, returns and fees when judging profit.

What a good result looks like

An example, not your task: A cycle count or full count shows differences and nobody knows which are real.

• variance-table.csv. One row per item and location: snapshot, movements, expected, counted, difference and a label of timing, unit, count error or unexplained.
• Recount comparison. The first count and the recount side by side, and whether each matches the expected quantity.
• Owner review brief. Verified differences, other ways to explain each, and the evidence still missing. No loss or fraud is stated.
• Investigation actions. Follow-up actions built only from commitments the owner made, with owners and due dates, or MISSING.
• Sheet check notes. Duplicate locations, repeated item IDs and unit mismatches found in the count sheets.
• MISSING lines and CHECKS MISSED. Every unit, time or movement the sources did not give, and every check that could not run.

How the work is checked

• Every timing label shows the movement document behind it.
• Timing, unit, count error and unexplained are never merged into one total.
• No word such as loss, theft or fraud appears, and no adjustment is posted or proposed as done.
• The original count sheets are unchanged.
• States which checks passed, failed or could not run, and names missing inputs.

These checks have not been run against a model yet. Your AI uses them to check its own work.

Capabilities

• Read the last review. Loads the snapshot cutoff, count version, difference by item and unit, movement sources and open commitments. A first run builds a baseline with you.
• Check the sheets. The Spreadsheet checker looks for duplicate locations, repeated item IDs and unit mismatches. It keeps your original counts.
• Work out what to expect. For each item and location, takes the snapshot quantity and adds or subtracts the movements recorded since, in the same unit.
• Label each difference. Marks it timing, unit, count error or unexplained. A recount that differs from the first count is shown beside it, as evidence and not as proof.
• Write the owner brief. The Stakeholder update writer drafts a brief of verified differences, other explanations and missing evidence. No stock adjustment is made.
• Plan the follow-up. The Executive action register keeper lists investigation actions from commitments the owner has made. Then it checks and saves the files and the new record.

How the role works

1. Checks your task, context and available tools.
2. Picks the methods needed and prepares a work plan.
3. Carries out supported tasks, with each method's limits and review gates.
4. Checks the real deliverables and corrects supported faults.
5. Hands back files, evidence and unresolved gaps.

Your specialist's working setup

Download the base specialist without a task form or account. README.md and SETUP.md explain how to use the folder. SOUL.md defines its scope. INTAKE.md asks missing questions inside your AI. PLAYBOOK.md, SKILLS.md, BEST_PRACTICES.md, CHECKS.md and HANDOFF.md guide the work and its checked deliverables.

Complete associated skills, supporting files and licences are included. Choose only the methods needed for the actual task. Keep context, working files and accepted feedback in your AI project.

Use this specialist through MCP or the API

Call prepare_specialist with role_id warehouse-count-discrepancy-review-assistant only. No task is required to load it. Your AI asks what job you need done, what you want back and which relevant context it can use.

For task-led hiring, call find_specialist, explain the suggested specialists and let the user accept or change the team.

Complete ZIP: Download base specialist. API: GET /api/v1/specialists/warehouse-count-discrepancy-review-assistant/pack. The host AI supplies its own model, tools and execution; a pack does not connect accounts or run a task.

Inputs and helpful context

The specialist starts from the task brief. Only active methods require their inputs.

For the primary method, when used

• Count sheets with item ID, location, unit and counted quantity (from your count app or paper sheets)
• The inventory snapshot: system quantity for each item, and the exact time it was taken (from your warehouse or stock system)
• Movements recorded between the snapshot and the count, such as receipts, picks and transfers, with document IDs (from a system export)
• The count scope: which locations and items were counted
• Your tolerance for a difference, meaning the size that counts as a variance (optional)
• Earlier explanations for known differences, a recount sheet and last review's record (optional)

Helpful context (optional)

• A stock system export with the snapshot time in it.
• Receipts, picks and transfer documents for the period.
• Your count procedure and tolerance rules.

Tools and access

• A capable chat AI with file support: Reads count files and keeps the record between reviews. In a plain chat tool, paste tables as CSV.
• A spreadsheet-capable AI: Formulas are only recalculated if your AI has a spreadsheet engine. In plain chat, it does the sums in text and says so.

Functions and deliverables

• Warehouse stock difference reviewer: Compare stock counts with dated movements and units. Get explained differences, open questions and investigation actions to approve.
  • A variance table (variance-table.csv): one row per item and location with expected, counted, difference and a label
  • A review brief for the warehouse owner: verified differences, other ways to explain each and the evidence still missing
  • A short list of investigation actions, built only from commitments the owner made
  • A comparison of the first count and the recount, when both exist
  • MISSING lines for gaps, and a record to start the next review from
• Warehouse stock variance reviewer: Compare stock counts with the recorded snapshot and movements. Get explained differences, source gaps and questions for the warehouse owner.
  • A variance table: snapshot, movements, expected, counted and variance for each item
  • Each gap in a class: timing, unit, duplicate, count error or unexplained
  • A list of duplicate locations, duplicate item IDs and unit mismatches
  • A review brief for the warehouse owner with missing evidence and decisions needed
  • A note that no stock adjustment was made
• Action item tracker: Share meeting notes or email threads. Get action items with owners, deadlines and unresolved commitments.
  • action-register.csv: id, task, owner, due_at, timezone, status, depends_on, source_id
  • clarifications.md
  • proposed-task-changes.md
• One-time stock quantity picker: Find the best one-time stock quantity for a short-life item. Give it cost, price and demand. You get the order size and the risk of other choices.
  • The best order quantity
  • Expected profit
  • The chance of running out
  • Profit and risk at other order sizes
  Source: Kishor Kukreja.
• Brand profile builder: Share your website. Get a reusable profile covering your products, customers, voice and approved claims, with sources.
  • A folder of eight plain files, from a start file for your AI to a claims list and a change log
  • A tag on every fact, and every claim marked approved, pending or banned
  • A two-sentence diagnosis of your biggest gap, and a list of what's missing and who can answer it
  • A short slice for your next step, if you name one

Tasks you can use it for

• A cycle count or full count shows differences and nobody knows which are real.
• A recount was done and you need to compare it with the first count.
• Goods moved in or out between the system snapshot and the count.
• The owner wants a review brief before anything is adjusted.

Limits to check before starting

• You want to adjust stock in your system. This agent reads and explains. You make the adjustment.
• You suspect theft or fraud. Take it to the person who handles loss investigations. This agent does not infer it.
• Your count sheet is messy and has no snapshot time. Clean it first with the Spreadsheet checker.
• You only want a status note for the owner. Use the Stakeholder update writer.

Best way to use this role

Give it one clear outcome, the actual files and a definition of done. Keep the work in one AI project so it can reuse checked context and earlier files.

• Add the agent to Claude, ChatGPT or Gemini with its helpers, the Explain a warehouse stock count variance workflow and the Executive action register keeper. Or download the pack and attach it.
• Paste your count sheets, the snapshot with its time, and the movements since. Paste them as CSV or a table. Say the count scope. From the second review, paste last review's record.
• Read the variance table. Check that each timing label shows the movement record behind it.
• Fill in each MISSING line. A missing snapshot time or unit cannot be guessed.
• Make any stock adjustment in your own system. The agent never does.
• In a chat-only tool, copy the record it gives you and paste it in next time. Installing sets up no schedule, so you start each run.

Read it as data

Read the complete role setup as JSON. Connect your AI to the AI Blocks MCP library and read aiblocks://roles/warehouse-count-discrepancy-review-assistant for its role instructions and pinned sources.

This is a prepared AI role. No task has run. Your AI carries out authorized work using its available tools and reports the actual deliverables and checks.