Solutions built around your ERP.

The problem

Orders, stock, customers and invoices live in separate systems. Teams re-type data between them, records drift apart, and nobody notices a failed sync until a customer does.

What I build

An integration hub that sits next to your existing ERP, keeps it as the source of truth, and connects it to your other tools through their official APIs: field mapping, duplicate prevention, retries for failed jobs and a full audit history. No ERP replacement.

How it runs

1. Map systems, fields and manual hand-offs
2. Connect each system through its official API
3. Run alongside the manual process, then switch over
4. Monitor sync status and exceptions on one dashboard

What you get

• Connectors with field mapping and validation
• Retry queue, duplicate protection and audit log
• Sync status dashboard and alerts
• Runbook and handover documentation

The problem

Supplier invoices arrive as PDFs, scans and email attachments and are typed into the ERP line by line. Duplicates slip through, mismatches with purchase orders surface late, and e-invoicing rules now require structured data.

What I build

Invoices come in by email or upload, OCR and AI extract every field with a confidence score, duplicates are caught, lines are matched to purchase orders, and anything uncertain goes to a person for approval before a draft vendor bill is created in your ERP.

How it runs

1. Collect sample invoices and your validation rules
2. Tune extraction, duplicate checks and PO matching
3. Add the approval screen for your team
4. Post approved bills to your ERP through its API

What you get

• Extraction with confidence scores
• Duplicate detection and purchase-order matching
• Approval screen with exceptions
• Draft vendor bills in your ERP, with audit trail

The problem

Trip income, fuel, driver pay, tolls and waiting time sit in different files, so nobody knows which routes or customers make money, invoices wait on proof of delivery, and customers call to ask where their order is.

What I build

One operations view that combines trip revenue and costs, tracks proof of delivery and invoice readiness, sends customers automatic status updates, and flags exceptions before they turn into losses.

How it runs

1. Connect trip, cost and delivery data sources
2. Define cost and margin rules with your team
3. Build dashboards and exception alerts
4. Add proof of delivery and customer notifications

What you get

• Trip and customer profitability
• Proof-of-delivery and invoice-readiness workflow
• Exceptions dashboard and alerts
• Automatic customer status updates

The problem

Staff lose hours searching shared drives and asking colleagues, and the answer depends on who you ask. New hires take months to learn where everything is.

What I build

An AI assistant connected to your own documents. It answers in French, Arabic or English, shows the exact source for every answer, respects who is allowed to see what, and says clearly when the documents do not cover a question.

How it runs

1. Collect and OCR your documents, including scans
2. Index them with access rules per team
3. Deploy the assistant on your intranet or web app
4. Measure answer quality with a test set and monitor usage

What you get

• Working assistant with cited answers
• Document ingestion pipeline you can re-run
• Admin view of sources and questions
• Hosting, logs and handover documentation

The problem

Most support tickets are repeat questions about orders, returns and policies, so the team has little time left for the cases that need a person.

What I build

An AI agent that answers from your policies and live order data, hands the conversation to a person with full context when needed, and logs every action. Secured against prompt injection and limited to the data each customer may see.

How it runs

1. Collect policies, FAQs and order data access
2. Build the agent with guardrails and hand-off rules
3. Pilot on a share of conversations
4. Review transcripts and tune before full launch

What you get

• Support agent on your site or help desk
• Hand-off to your team with context
• Audit log and conversation review
• Security review of data access

The problem

Each client environment was set up by hand, deploys depend on one person, and nobody knows which dependencies are vulnerable.

What I build

One reproducible platform: every client isolated, GitOps deployments, scanned and signed images, secrets in a vault, backups, and monitoring that alerts before clients notice.

How it runs

1. Audit the current environments
2. Build the platform and pipelines
3. Migrate clients in batches
4. Hand over runbooks and monitoring

What you get

• Kubernetes platform with per-client isolation
• CI/CD and GitOps pipelines
• Image scanning, signing and SBOMs
• Monitoring, backups and runbooks