August 7, 2025

|

Case Study

How 2 global recycling firms benefit from Safi AI

AI improves site quality workflows, automates quality claims, powers data-driven purchasing and stops non-target material hitting production lines (5 min read)

Key Impacts from Safi AI

  • $5k Claim: Identified quality deviation and auto-generated claim report for $5,000
  • 10 min triage: ‘98/2’ load analysed in 10 minutes by non-specialist - 20% of load flagged as below spec. AI triggered instructions for what to do, preset by Quality Lead
  • Purchasing: Analysed high variation within a load, checked historic data on supplier and re-classified them as high-risk for Purchasing team moving forward
  • Safety: Detected a load contained 1% of undeclared hazardous material, preventing a potential incident in processing
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Common Challenges for Companies

Solution with Safi AI

Reliable visual inspections require high-skilled staff who may not be available or have limited time
Aids inspectors with reliable, accurate data. Computer Vision can spot objects that average human eye can miss
Quality data about material is generated slowly or not at all
Reports with data & images are generated instantly. Real-time data feed for quality, purchasing and production teams
Site team lacks time and skillset to effectively triage material when it arrives
Non-specialists using AI can triage a load in 10 minutes, get results and take material to the right place on site
Quality defects are not reported - Owed refunds not claimed and Purchasing team limited in improving strategy
Automates claim refund reports and performance tracking on sellers, enabling Purchasing Team to take action quickly, act with data and improve financial outcomes
Material can be mis-categorised, misplaced, or quality specialists become overwhelmed with requests
Quality specialists can preset instructions for colleagues to take based on AI results, creating more effective site workflow
Hazardous, non-target or unexpected material goes into production
AI identifies over 40+ common contaminants, enabling production teams to protect quality, yield and people

Action

Safi AI was used on 3 purchases by 2 leading recyclers of Aluminium UBC. The process involved two main steps:

  1. Rapid Load Scanning: Using Safi AI Bale, loads were analysed in minutes to determine average composition, analyse variation and identify any high-risk bales or briquettes
  1. Rapid Sample Analysis: Using Safi AI Loose, samples of loose material were rapidly analysed using Safi AI Loose for a more granular composition breakdown

‍

Interactive, digital reports were auto-generated with composition breakdown and photographic evidence.

Instantly accessible for all the members of purchasing, production and quality teams. Claim refund request was auto-generated for the team.

‍

Results

Analysis of 3 purchases demonstrates Safi AI impact in various scenarios.

Purchase 1: Calculates quality deviation and refund value

  • Scenario: An 80-tonne load of 95% UBC was purchased for $142,400
  • Analysis: Safi AI identified the actual composition as 92.7% UBC. 
  • Outcome: A claim refund value of $4,272 USD

‍

‍

Purchase 2: Identifies material that requires closer inspection

  • Scenario: 48 tons of 98% UBC worth $84,000.
  • Analysis: Safi AI's variation analysis identified 20% of the briquettes within the load were below 96% UBC. It also flagged that 1% of the material was potentially hazardous material.
  • Outcome: Team could set aside lower quality briquettes for pre-production checks. Purchasing team could send data-driven feedback to the seller.

‍

Purchase 3: Flags a supplier as high-risk

  • Scenario: A 25-ton, $50,000 load, expected to be 96% UBC
  • Analysis: Safi AI detected high variability throughout the load (UBC content ranged from 88-96%). Checked historic data on the seller, identified this variation was common.
  • Outcome: Re-classified seller as high-risk. Purchasing team could change their pricing and work with the supplier to understand if quality can be stabilised more.

‍

4. Conclusion

By implementing Safi AI, purchasing, site and quality teams were able to collaborate more easily, prevent errors, save time, claim higher refund value and make data-driven decisions.

‍

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.