Perishables inventory health & disposition

The margin you lose to spoilage is a decision problem, not a cost of doing business.

Augmentics builds an AI decision system for regional foodservice and specialty distributors. Every morning it tells your buyers and warehouse team exactly what to do with the lots that won't sell at full margin — ship it, repack it, divert it, mark it down, or donate it — with the dollar value of each option attached.

Built by operators from
Amazon Coupang HelloFresh
The problem we work on

A perishable lot is a clock you can't see, and a decision your team makes under pressure every single day.

In regional perishables distribution, the difference between margin and shrink is a series of small calls made at 6am with incomplete information: which customer gets this lot, what gets marked down, what gets diverted, what to reorder. Today those calls live in spreadsheets and the heads of a few veteran buyers. When they're right, nobody notices. When they're wrong, it shows up as spoilage.

6:47am
A typical morning

Forty-seven lots need a disposition decision before the trucks load. Each one has a shelf-life clock, a set of customers who might take it, a secondary market, and a markdown floor. Augmentics ranks them, recommends an action for each, and shows the dollar consequence — so the decision takes minutes, not the whole morning.

How it works

One decision system. Three specialized agents.

Augmentics reads your lot-level inventory, your incoming orders, and where it can, your cold-chain data — through read-only connectors to the systems you already run. It is not a chatbot. It is a small, governed decision system that reasons over your data and produces recommendations your team can act on.

AGENT 01 · LIVE

The Forecaster

For every active SKU-and-lot, it predicts demand over the next 1–14 days as a probability, not a guess — “a 70% chance this lot clears at full margin in three days.”

reads → inventory + demand
AGENT 02 · LIVE

The Disposition Agent

For each lot at a decision point, it weighs the options — ship first, repack, divert, mark down, donate, scrap — and ranks them by expected dollar outcome.

decides → the best action
AGENT 03 · LIVE

The Buyer Agent

It closes the loop upstream: given what's already at risk in the warehouse, it advises next week's purchasing — don't reorder this, increase that, switch this supplier.

advises → next week's orders
Decision queue · what your buyer actually sees
Lot · Strawberries, 1lb clamshell · 84 casesDay 4 of shelf life
Expected sell-through at full margin33%
Recommended actionShip FIFO → Restaurant Y
Expected outcome vs. doing nothing+$1,840
ConfidenceHigh · 0.87
Next-best alternativeMarkdown 15% today → +$910

The operator clicks approve, modify, or override. Every override teaches the system where your team knows something it doesn't.

Who it's for

Built for the distributors the big platforms can't serve well.

We work with founder-led and family-owned regional distributors in foodservice, specialty produce, and protein — the operators who chose to stay independent, and who'd rather have a sharp technical partner than be rolled up into someone else's playbook.

01
Regional foodservice & specialty distribution$200M–$600M operators where perishables are the hardest part of the catalog, and where a few buyers carry the decisions in their heads.
02
Founder-led, decision close to the topNo procurement committee. The person who feels the shrink is the person who can say yes to fixing it.
03
Real data, honestly messyLot data that's incomplete on the receiving end. Forecasting that lives in Excel and tribal knowledge. We bring the data layer with the agent — we don't ask you to spend 18 months rebuilding your systems first.
The first 90 days

The agent earns autonomy. It is never granted.

This is where most AI deployments fail — and where being precise from day one is the whole game. The system arrives with no authority. It proves itself on small, bounded decisions, and the envelope only widens as the results hold up. You stay in control the entire time.

Days 1–30 · Shadow

It watches and recommends

The agent runs in parallel. Your buyers make every call. Every disagreement between the agent and your team is logged — that's the most valuable thing it learns.

~0% autonomous
Days 31–60 · Conditional

It handles the small stuff

Low-value lots with plenty of shelf life and high confidence can execute without review. Everything bigger, riskier, or customer-sensitive still goes to a human.

~30% autonomous
Days 61–90 · Bounded

You set the envelope

Routine decisions run on their own. You decide how far that goes. The system tracks its own accuracy and dollar impact the whole way, so the value is never a matter of faith.

~60–80% autonomous

Three things the agent never does without a human

  • Divert a key customer's preferred SKU away from them
  • Make any call that touches food safety or donation compliance
  • Amend a purchase order above a dollar threshold you set
The operators behind it

Two founders who have done this work inside the businesses you compete with.

Co-founder · Supply chain

Matt West

Twenty years building and scaling supply chains, most of it on the hardest problem in perishables: matching capacity and inventory to demand that won't sit still. At Coupang — South Korea's largest online retailer — Matt built the company's first capacity-planning function from the ground up, forecasting demand for its dry and cold-chain networks through hypergrowth, and led the build-out that grew inventory from 10 million to 96 million units across twelve new cold-chain sites. He went on to run end-to-end supply chain for Coupang's launch into Taiwan, cutting out-of-stock rates from 36% to 12% while standing up a local team.

Earlier, at Amazon, he managed inventory for the fiercely seasonal US Toys business — forecasting, buying, and cutting purchase orders against demand that spiked and collapsed faster than any system could track — then built the multi-year capacity plan for Amazon's launch into Mexico. He started where the best operators do: on the floor, coming up over six years at UPS.

Matt has built and led supply chain teams across South Korea, Taiwan, China, Mexico, and now Spain, repeatedly delivering in markets where he didn't share the local language. For a company whose customers span Iberia, Northern Europe, and the US, that range isn't a footnote — it's the job.

Coupang cold-chain capacityAmazon demand planningCross-cultural opsMBA, Supply Chain
Co-founder · Technology

Dr. Tim Schoenharl

Twenty years building the systems that decide what to buy, what to hold, and what to let go. At Amazon, Tim spent eight years on the Global Inventory Platform, managing the engineering teams behind its automated buying algorithms and — most relevant here — its Optimal Inventory Health and Removals Planning systems, which recommended what inventory to pull and how to dispose of it across Amazon's worldwide network. The disposition problem Augmentics solves for perishables is one he has already solved at global scale.

He went on to lead supply chain engineering as a Senior Director at Coupang — owning forecasting, procurement, and inventory automation across engineering, data science, and operations for both fresh and full-line retail — and then as VP of Engineering at HelloFresh, where he ran the supply chain technology organization and, latterly, the platform governing how GenAI is deployed safely across the business.

Tim builds the part of Augmentics that turns a promising agent into a trustworthy one: the evaluation harness that scores every recommendation against its real-world outcome, so the system's accuracy is proven decision by decision rather than taken on faith. He holds a Ph.D. in computer science from Notre Dame.

Amazon inventory health & dispositionCoupang SCM automationHelloFresh VP EngineeringPh.D. CS, Notre Dame
Start a conversation

If perishable shrink is a number you'd rather not say out loud, let's talk.

We work with a small number of distributors at a time. A first conversation is exactly that — two operators who understand your business, asking what your hardest disposition decisions actually look like.