AI advisory for commodities and recycling
AI advisory and implementation for physical commodity and recycling professionals.
We help traders, yards, mills and recyclers put AI to work on deals, compliance, quality and claim management, logistics, manufacturing data and supply and demand analysis. Before your team relies on it, we test every result systematically on your own cases, so you know where the AI is right and where it is not.
- Materials
- Recovered paper · Plastics · Metal scrap
- Clients
- Traders · Importers & exporters · Logistics · Yards & sorting plants · Plastic recyclers · Paper and steel mills
- Markets
- Germany & EU · Türkiye · Saudi Arabia · India
Advisory, implementation and evaluation of AI for commodity trading, recycling and mills.
Physical commodity businesses run on documents, judgement and thin margins: contracts and confirmations, inspection reports, shipping papers, quality claims and correspondence with suppliers and customers. Our AI offerings are designed for yards, mills and trading desks.
We come from the business side. We have bought, sorted, baled, shipped and sold secondary raw materials, and we understand where a wrong grade, an error in a document or a late shipment costs real money. We use that knowledge to decide where AI is worth using, to build it properly into your systems and to test its output against the standard an experienced trader or plant manager would apply.
01 AI impact
Results, not technology.
Lower costs, faster turnaround, fewer errors, and teams that get more done without working longer. AI makes this possible in commodity trading and recycling today.
What decides the outcome is not the tool but the approach: choosing the right processes, knowing the material and the market, and implementing carefully until the system holds up in daily operations.
AI agents now take over whole sequences of work: reading documents, checking them against each other, preparing decisions. That changes which tasks people do, not how your business works. We fit agents into the workflows your team already uses, decide with you which steps stay with people, and train your staff to work with them confidently.
The aim is not only more output. It is a business that adapts faster to markets and regulation, and in which experienced people make the decisions.
02 Services
Nine services, one standard
AI Adoption Strategy & Advisory
Decide where AI pays off in your business, and where it does not.
We sit down with your trading, operations, quality and back-office teams, go through their workflows with them and advise where AI saves time or reduces risk. We rank the opportunities by value, feasibility and data readiness. You get a prioritised roadmap, a realistic business case per use case and clear build, buy or wait decisions. We also cover governance: tool usage, protection of contract and pricing data, and what the EU AI Act and local regulation mean for you.
Deliverables
- Process review
- Use-case portfolio
- Roadmap and business case
- AI usage and data policy
- Vendor shortlist where relevant
AI Model Selection
The right model for each task: quality, cost, speed and where your data goes.
There is no single best model. Commercial models from Anthropic, OpenAI, Google and Mistral and open-weight models differ in accuracy on your documents, cost per task, speed, languages and where data is processed. We test the candidates on your own cases, such as contracts, grades and multilingual correspondence, and recommend which model to use for which task, including EU or on-premise hosting where your data policy requires it. When new models are released, we re-run the comparison, so you are not locked into last year's choice.
Deliverables
- Shortlist of candidate models
- Benchmark on your own cases
- Cost per task and per month
- Hosting and data-location options
- Recommendation per use case
- Re-check when new models are released
EU AI Act & GDPR Compliance
For European clients: AI use organised and documented in line with the EU AI Act and the GDPR.
We take stock of the AI systems you use and plan, classify them with you against the risk categories of the EU AI Act and put the resulting obligations in place: AI literacy of staff, transparency notices, documentation and human oversight. On the data protection side, we prepare records of processing, data protection impact assessments and technical and organisational measures for AI use, and check data flows to model providers. We advise on implementation and organisation; binding legal assessments of individual cases are made by your legal counsel or a specialised law firm, with whom we coordinate.
Deliverables
- Inventory of AI systems in use and planned
- Risk classification prepared for legal review
- AI literacy training (Art. 4 AI Act)
- Records of processing and DPIA preparation
- Technical and organisational measures for AI use
- Provider and data-flow check (hosting location, sub-processors)
AI Literacy Training
Teams that know what AI can do with their material, and where it fails.
We train your teams on their own documents and processes, not on generic examples. The training covers how language and vision models work and where they fail, working with AI on contracts, shipping papers and grading photos, checking AI output, data protection and confidentiality, what the EU AI Act means in daily work, and using prompts and agents safely.
Deliverables
- Training on your own documents and cases
- Modules: foundations, checking AI output, data protection, EU AI Act, prompts and agents
- Management briefing and team workshops
- Participation documented per person
AI Projects & Implementation
From selected use case to a system your team uses every day.
We run AI projects end to end: scoping, data preparation, solution design, build, testing on your own documents, rollout and handover. Projects are small enough to finish and specific enough to measure, for example contract review for the purchasing desk or automated reporting for a sorting plant. We work with your IT team or your technology partners and deploy where you prefer: on your own servers or in the cloud.
Deliverables
- Working solution in production
- Test results on real cases
- Documentation and training
- Support plan
Workflow Design & Automation
Workflows designed and automated around your daily work.
Your team keeps working the way it does today. We look at the daily work as it really runs (inbox, spreadsheet, phone call, ERP entry) and build workflows that take over the repetitive parts: reading, sorting, drafting and checking. Decisions that carry commercial risk stay with your people. Typical cases: order intake from emails and PDFs, supplier onboarding, claims handling, daily position and stock reporting.
Deliverables
- Map of your daily workflows
- Automations built into your existing tools
- Human review points
- Measured time per task, before and after
Connecting AI to Your Systems
AI that works inside the systems you already run.
Your data sits in ERP, trading and contract systems, document archives, email and spreadsheets. We connect AI to these sources securely and with the right permissions, so answers are based on your actual contracts, stock and shipments. We work with Microsoft 365 and SharePoint as well as the industry-specific and in-house systems typical for traders and recyclers.
Deliverables
- Integration architecture
- Secure data connections
- Access and permission model
- Monitoring
Model Fine-Tuning & Domain Adaptation
Teach the model your grades, your terms and your languages.
General-purpose models do not know the difference between two specific commodity grades, a local yard term or how your contracts are structured. Where it is justified, we adapt models through curated examples, retrieval from your own documents or fine-tuning on your historical data, and compare the result against the general model so you only pay for improvement you can see. Deployment in the cloud, in a European data centre or on your own infrastructure.
Deliverables
- Curated training and test data
- Adapted model
- Comparative evaluation
- Deployment that fits your data policy
AI Output Evaluation by Industry Experts
Already using AI? Have its output checked by us in a standardised, repeatable way.
Many companies have introduced AI tools without a reliable way to judge whether the answers are correct. We evaluate what your AI produces with a repeatable, neutral evaluation process that also tests random cases and includes security checks. We build test sets from real cases, review outputs for accuracy, completeness and commercial risk, find where the system invents information or misreads a grade, a clause or a figure, and tell you where it can be trusted, where it needs a human check and where it should not be used.
What we evaluate
- Grade and specification classification
- Contract and document extraction
- Price and market summaries
- Answers to customers and suppliers
- Compliance and waste-shipment guidance
- Reporting from plant data
- Security checks such as manipulated inputs and data leaks
Deliverables
- Domain test set from your own cases
- Scored evaluation report
- Risk register of errors with commercial or compliance impact
- Concrete fixes
- Optional recurring review
03 Use cases
Where AI earns its place
01 Trade documents
Export documentation check
Every shipping document checked against contract, loading photos, permits and letter of credit, before anything goes to the bank or the authority.
An agent reads bills of lading, proforma and commercial invoices, packing lists, weight notes and certificates, extracts every relevant field and cross-checks them against each other, against the contract and against the loading photos. Discrepancies with the letter of credit, the waste-shipment notification or the export permit are flagged before documents go to the bank, the carrier or the authority.
- Bill of lading, proforma and commercial invoice, packing list, certificates
- Container and seal numbers, weights and grades matched to loading photos
- Letter of credit terms checked for discrepancies before presentation
- Conformity with permits, notifications and Annex VII documents
02 Trade documents
Contract and confirmation check
Terms extracted and compared across purchase and sales contracts and confirmations, with deviations flagged before they become disputes.
The AI extracts the terms of purchase and sales contracts and confirmations (grade, tolerance, price basis, payment, delivery window, claims clause) and compares them with each other and with your archive of previously closed contracts. Deviations are flagged before they become disputes, which matters most for long-term contracts.
- Grade, tolerance, price basis, payment and delivery window
- Purchase and sales side compared
- Checked against your archive of closed contracts
- Deviations flagged before they become disputes
03 Quality & claims
Claim management with visual evidence
A trained visual model compares loading photos with the claimant's photos and checks the claim against contract and documents.
When a buyer claims on quality or contamination, the model compares the photos taken at loading with the photos submitted with the claim and assesses whether they plausibly show the same material in the condition shipped. In parallel, the claim is checked against the quality terms, tolerances and claim deadlines of the contract and against the shipping documents. Your team receives an evidence file and a draft position within minutes.
- Visual comparison of loading and arrival photos
- Check against quality specification, tolerances and deadlines
- Consistency with weights, container and seal numbers
- Evidence file and draft response for the claims desk
04 Quality & claims
Visual grading of material
Grade recognition from photos, trained on your own images and data or pre-trained for recovered paper, plastics and metal scrap.
A visual model classifies material by grade and flags visible contamination and prohibitive materials, with a confidence score for every decision. It is trained on your historical photos and grading decisions, or starts from a model pre-trained for the recycling industry and is adapted to your grades. Uncertain cases go to your graders.
- Recycling commodities
- Trained on your images and grading history, or pre-trained for recycling
- Confidence score per decision, human review below threshold
- Used at intake, at loading and in claims
05 Compliance
Environmental and waste-shipment compliance
An agent that checks every movement of waste and secondary raw materials against the applicable regulation, permit and notification.
Waste shipments depend on classifications, permits, notifications and quantity limits, and the rules differ by country of dispatch, transit and destination. The agent classifies the material, checks it against the EU Waste Shipment Regulation and the relevant national rules, tracks notified and remaining quantities and prepares the required documents, such as Annex VII. Exceptions go to your compliance officer together with the rule that triggered them.
- Waste codes (EWC, Basel) and classification of the material
- Permits, notifications and remaining approved quantities
- Annex VII and movement documents prepared and checked
- Rules for country of dispatch, transit and destination
06 Compliance
KYC compliance agent
Onboarding and periodic review of counterparties: identity, ownership, sanctions and documents in one auditable workflow.
The agent collects and checks company documents, register extracts and beneficial owners, screens counterparties and their owners against sanctions and PEP lists, monitors document expiry and triggers periodic reviews. Every step is logged, so the file is ready when a bank, an auditor or an authority asks.
- Register extracts, beneficial owners and company documents
- Sanctions and PEP screening, repeated on every change
- Expiry monitoring and periodic review
- Complete audit trail per counterparty
07 Compliance
Audit preparation and follow-up
An agent that assembles the evidence before an audit and tracks every finding to closure afterwards.
Before certification, authority or customer audits, the agent maps the requirements to your documents and records, compiles the evidence and lists the gaps. After the audit, it turns findings into corrective actions with owners and deadlines, follows up and documents closure.
- Requirements mapped to documents and records
- Evidence pack and gap list before the audit
- Corrective actions with owners and deadlines
- Waste-management certification, ISO 9001 and 14001, customer audits
08 Planning & risk
Supply chain risk management
Counterparty, country, logistics and price risk along the chain, analysed and built into daily decisions.
We analyse where your supply chain is exposed, from supplier concentration and counterparty credit to country, freight and price risk on open positions, and implement the controls: limits, early-warning signals and alerts that reach the right person. The analysis becomes a running system, not a one-off report.
- Counterparty and supplier concentration
- Country, sanctions and logistics risk per trade lane
- Price exposure on open positions
- Limits, early warnings and alerts in daily operations
09 Planning & risk
Supply and demand forecasting
Models fine-tuned on your trading history and market data that forecast supply and demand by grade and region.
Fine-tuned on your purchase and sales history, mill demand, seasonal patterns and market indicators, the model forecasts available supply and expected demand by grade and region. Forecasts come as ranges, not single numbers, and are back-tested on past periods before your traders rely on them.
- Trained on your contracts, volumes and prices
- Market indicators and seasonality included
- Forecast ranges by grade and region
- Back-tested before use
10 Planning & risk
Plant utilisation and maintenance planning
Throughput and maintenance planning for sorting lines, balers and shredders, based on your operating data.
AI matches expected inbound tonnage with the capacity of your lines and machines to plan shifts and throughput, and uses operating hours, fault logs and, where available, sensor data to schedule maintenance before a breakdown stops the plant.
- Inbound tonnage against line capacity
- Shift and throughput planning
- Maintenance by operating hours and fault history
- Planned downtime instead of unplanned stoppages
11 Data & systems
Agent-run business reports
Daily and weekly reports produced by agents directly from your systems: positions, margins, stock and shipments.
Agents pull data from ERP, trading, weighbridge and logistics systems, reconcile it and write the report your management actually needs, with every figure traceable to its source. Delivered on schedule by email or Teams, with follow-up questions answered on the spot.
- Open positions, margins and stock
- Shipments, delays and open documents
- Every figure linked to its source
- Delivered by email or Microsoft Teams
12 Data & systems
Connecting business systems to AI
ERP, trading and document systems connected to AI infrastructure, securely and with the right permissions.
Every use case above depends on access to your data. We connect ERP systems, trading and contract systems and document archives to the AI infrastructure through secure interfaces, with role-based permissions and logging. Data stays where your policy requires.
- ERP, trading and document systems
- Secure interfaces with role-based permissions
- Every access logged
- Hosting in the EU or on your own infrastructure
04 Who we serve
From the yard to the mill, and the traders in between
- 01
Collect and sort
Intake, weighing, grading, sorting and baling.
- Waste yards and sorting plants
- Plastic scrap traders and recyclers
Where AI helps
- 02
Trade
Buying, selling and moving material across borders.
- Waste paper traders
- Metal scrap traders
- Physical commodity traders
- Importers and exporters
Where AI helps
- 03
Ship
Bookings, transport and documents.
- Logistics companies in commodity trade
Where AI helps
- 04
Process
Purchasing, inspection and supplier management at the mill.
- Paper mills
- Steel mills
Where AI helps
05 How we work
Confidential, and under your control.
- 01
Confidentiality under NDA
Every engagement starts with a signed non-disclosure agreement. You decide which data you share with us, and all of it stays 100% confidential.
- 02
Your data, your knowledge, your models
Data, documents, knowledge bases and adapted models belong to you and stay under your control.
- 03
You decide where it runs
Workflows and models run where you want them: in your own cloud environment or locally on your own servers.
06 About
Commodity professionals advising on AI.
QimaFlow AI advises and supports physical commodity traders, recyclers, yards and mills on the adoption of artificial intelligence.
Our team comes from the trade itself, from waste paper and recycled plastics trading and sorting, and combines that experience with hands-on AI implementation and expert evaluation. Based in Germany, we work with clients across Europe, Türkiye, Saudi Arabia and India.
- Based in
- Germany
- Markets
- EU · Türkiye · Saudi Arabia · India
- Languages
- German · English
07 Why us
We know the commodity business. That is the difference.
There is no shortage of AI consultancies. Very few of them have stood on a yard, argued a quality claim with a mill or supplier, or chased a missing document before a deal closed. We have.
Independent of vendors
We are not tied to any model provider, platform or software vendor. We recommend what fits your case and your data policy, and we tell you when an off-the-shelf tool is enough.
Industry professionals, not generalists
Our background is physical commodity trading and recycling: waste paper, plastics and metal scrap, yards, sorting plants and the mills that buy the material. We speak your team's language and understand the commercial consequences of a wrong answer.
Evaluation is built in
Every system is tested on your own cases in a repeatable way before it goes live, and we offer the same review for AI you already use.
Production, not pilots
Success is what your team uses every day, not the number of proofs of concept.
European standards, working internationally
Data protection, confidentiality of every piece of data you entrust us with and compliance with applicable regulation are part of the design from day one, whether you are in Hamburg, Istanbul, Riyadh or Mumbai.
08 Questions
Frequently asked
We already use ChatGPT or Copilot. Why do we need you?
General tools are useful, but they do not know your grades, contracts or systems, and nobody checks their answers systematically. We connect them to your data where it makes sense, and we evaluate whether their output is reliable for your use.
Where is our data processed?
Wherever your policy requires, including European data centres or your own infrastructure. Confidential pricing and counterparty data is never used to train third-party models without your consent.
How long until we see results?
A focused pilot on one process typically runs a few weeks. We agree success measures before we start.
09 Contact
Start with one process.
Tell us where your team spends most of its time: contracts, grading, claims, documents or reporting. In a first conversation we will tell you honestly whether AI is the right answer, what it would take and what it is likely to deliver.
- Direct
- contact@qimaflow.ai
- Contact
- QimaFlow AI
A business unit of
RSO Entsorgungs- und Beratungs GmbH
Forellenweg 2
87776 Sontheim, Germany