labs@atheron:~/industries/supply-chain/ai$ train agent --grounded

AI supplier risk watch

On a private model, with every match explained.

Procurement teams hear about a supplier fire or a sanctions listing too late because nobody can read everything every day. An agent can, if it is grounded on your supplier list and tells the buyer why a story matters. Because that list and your spend are sensitive, we usually run it on an open-source model you host.

Start a project See the example projects

// what matters here

What is different in this industry.

Your supplier list stays private

The model runs in your own cloud account, so the names, parts and spend it matches against never go to an outside model provider.

Every match explained

Each alert links to its sources and states which supplier, site and parts it concerns, so a buyer can judge it quickly.

Fewer false alarms over time

Buyers mark alerts as useful or not, and those marks become the evaluation set every later change is measured against.

Reads more than English

Open-source models read news and filings in the languages your suppliers work in, and the brief comes back in yours.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAIOpen-source model

Supplier risk watch agent

Hypothetical. Not a client, not a result.

// The problem

A procurement team learns about a supplier's fire, strike, sanctions listing or insolvency from the news, after the orders that depend on it are already late.

// What we would build

An agent that reads news, sanctions lists and public filings every day, matches them to your suppliers and the parts each one makes, and writes a short brief for the buyer. It runs on an open-source model in your own cloud account, because the supplier list, part numbers and spend are commercially sensitive.

// What is in it

Supplier list and part mapping imported from the ERP
Daily read of news, sanctions lists and public filings in several languages
Each match explained, with links to its sources
Briefs for the buyer on what changed and which orders depend on it
Evaluation set of past events, run before every change
An open-source model served on GPUs in your own cloud account

// Stack

Qwen or Mistral (open-source)
vLLM
Python
PostgreSQL with pgvector
Next.js
AWS

// estimate

Build: ≈ US$113,000 to US$172,000, delivered within 22 weeksCAD 160,200 to 245,000
Hosting: ≈ US$6,710 a monthCAD 9,550 a month
Support: ≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route: Open-source model
Model running cost: ≈ US$6,530 a monthCAD 9,300 a month

Timeline by milestone

Discovery: 2.9 to 3.4 weeks
Specification and evaluation plan: 0.4 weeks
Design approved: 2.4 to 3.9 weeks
Core features: 1.4 to 1.9 weeks
Full build: 0.9 to 1.4 weeks
Working pilot: 3.9 to 6.9 weeks
Testing and fixes: 0.9 to 1.9 weeks
Launch: 0.9 to 1.4 weeks
Production: 0.9 to 1.4 weeks

How it is paid

Deposit 20%: CAD 31,980 to 48,940
Discovery 1.8%: CAD 2,878.20 to 4,404.60
Specification and evaluation plan 10.5%: CAD 16,789.50 to 25,693.50
Design approved 1.8%: CAD 2,878.20 to 4,404.60
Core features 5.6%: CAD 8,954.40 to 13,703.20
Full build 3.7%: CAD 5,916.30 to 9,053.90
Working pilot 31.7%: CAD 50,688.30 to 77,569.90
Testing and fixes 1.8%: CAD 2,878.20 to 4,404.60
Launch 1.8%: CAD 2,878.20 to 4,404.60
Production 11.3%: CAD 18,068.70 to 27,651.10
Holdback, 30 days after launch (10%): CAD 15,990 to 24,470
GPU time for training and testing, at cost: CAD 300. Billed up front, at cost, outside the milestones

Start from this example

// questions

Questions we get about this.

Why not use a frontier model for supplier monitoring?

You can, and it is quicker to start. We suggest an open-source model when the supplier list, part numbers and spend are confidential, because nothing leaves your cloud. Both routes are priced in the estimator.

Which sources does the risk agent read?

Public news feeds, government sanctions lists, company registries and filings, and any paid feed you already license. We agree the list with you before building.

Does the agent decide to drop a supplier?

No. It informs a buyer, who decides. Every brief records what the agent saw and when.

Back to the industry: Supply chain About the service: AI agents and models

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

Start a project Book a call