Agents in the Trades
Finance

Dead stock found before it eats the warehouse

It is 8:20 a.m. on the first Monday of the month and the warehouse manager at a ten-truck HVAC shop is staring at a shelf of filters and motors bought for jobs that already closed — while the system says another restock order is ready to send.

Source: ACHR News, "What AI-Driven HVAC Demand Planning Means Downstream" — https://www.achrnews.com/articles/166777-what-ai-driven-hvac-demand-planning-means-downstream — the article describes tools that track live inventory across every truck and warehouse, send usage-based reorder alerts and compare vendor pricing, and reports an example where that visibility found over a million dollars in excess inventory. That figure is the source’s reported example, not a promised result.

The problem

A contractor is a small distributor whether it plans to be or not: parts on the trucks, parts in the warehouse, parts staged for a job. Each pile gets counted separately, if it gets counted at all, so the shop can be long on a part in one place and reordering it in another.

Dead stock does not announce itself. It accumulates one sensible purchase at a time — a spare for a job that changed, a bulk buy, a return nobody logged — until cash sits on shelves instead of in the account, and the next order adds more of what is already there.

Get one workflow like this, every day.

The free Agents in the Trades newsletter — each workflow broken down: what it does, what it costs, and where it breaks.

How the agent workflow works

  1. Every location reconciles into one count. Once a month the agent pulls live truck and warehouse inventory into a single view, by part — not by shelf — so duplicates stop hiding from each other.
  2. Usage decides what “slow” means. The agent compares that count against recent job usage and flags what has not moved, what is stocked past its recent use, and what it is worth at cost.
  3. Restock orders get checked before they go out. Any pending reorder for a flagged part is held for review, with the on-hand total across all locations shown next to it.
  4. Vendor pricing gets compared on the rest. For parts that do need reordering, the agent compares current vendor pricing before the order is placed, so the restock does not quietly overpay out of habit.
  5. A person decides what happens to the excess. The sweep ends in choices a manager approves: return it, transfer it, discount it into upcoming work, or keep it deliberately as a spare — nothing is disposed of by the agent.

What you need to set it up

Where it breaks

Messy data makes confident, wrong answers. The source makes this point plainly: if counts, part numbers and usage logs are messy, the sweep will flag parts that are actually gone and clear parts that are stuck. The first sweep is really an audit of your own records.

Slow is not the same as useless. A rarely used part can be the one that saves a no-heat call in January. Seasonal and emergency stock looks dead for most of the year on purpose. The flag starts a conversation with someone who knows the work — it is not a verdict to clear the shelf.

Finding excess is easier than recovering its value. Returns have windows, restocking fees and vendors who will not take opened stock. The dollar figure on the report is what is tied up, not what you will get back, and treating it as recoverable cash overstates the win.

The bottom line

Run this monthly if cash keeps disappearing into shelves. Reconcile every location into one count, let usage — not habit — drive the next order, and make a person decide what the excess is for, before it eats the warehouse.

Related use cases

← All use cases

One working agent workflow, every day.

Real AI agent use cases from solar, construction, and field operations — what they do, what they cost, and where they break. No hype, no theory.