Procurement AI Agent

The Procurement AI Agent is an AI procurement agent for construction and project-based industries, part of DANAOS ProjectVIEW AI. It aims to compare supplier quotations and check prices against the budget rate for each BoQ item. Then it flags anomalies before anyone places an order. The procurement officer approves every award and purchase order. It also aims to evaluate suppliers and subcontractors on quantitative and qualitative criteria (price, technical capability, quality and past performance). As a result, it recommends the best-fit supplier, not only the lowest price.

The agent uses the data in the Procurement and Purchasing module of ProjectVIEW ERP, linked to the budget through BoQ, WBS and Cost Codes.

• Purchase requisitions and need dates from the WBS
• Supplier quotations and price history
• Budget rates by BoQ item and Cost Code
• Purchase orders, deliveries and supplier performance
• Committed cost against budget

The agent follows the ProjectVIEW AI agent cycle. Applied to procurement:

1. Detect: flag a quotation above the budget rate, an unusual price change or a missing item in a supplier offer.
2. Investigate: compare all quotations for the requisition with price history and the budget.
3. Explain: state what is different and what it means for committed cost.
4. Recommend: propose an award, a negotiation point or a request for a revised offer.
5. Request approval: send the recommendation to the procurement officer.
6. Execute: after approval, create or update the purchase order in ProjectVIEW ERP.
7. Audit: record the comparison, the recommendation and the decision.

The agent works within the approval limits set for each role. It does not place orders without a person’s approval.

Main inputs, outputs and decision points:

The Procurement AI Agent works inside ProjectVIEW ERP governance, following the chain Identity → Role → Permission → Data → AI → Action. It sees only the data the user’s role allows and proposes only actions that role could take. AI should operate inside enterprise governance, not around it.

• Consequential actions always need human approval.
• Also, the system records every recommendation, approval and change for audit.
• Data stays in the client’s private environment and never goes to public consumer AI services.

Illustrative example. A contractor needs reinforcement steel for Project D. First, the buyer enters three quotations against the requisition in ProjectVIEW ERP.

The agent compares them and flags that the lowest offer excludes delivery to site and that all three are above the budget rate for the item. It explains the effect on committed cost for the concrete works cost code.

It recommends asking the lowest bidder for a delivered price and alerting the cost controller to the budget gap. The procurement officer approves the request, and the agent compares the revised offer. Then the authorised approver releases the purchase order. Finally, the system logs each step.

The agent serves buyers and the people who control spend:

• Procurement officers who compare offers and issue purchase orders
• Cost controllers and project managers who need to see committed cost before invoices arrive

It applies to construction, shipbuilding and repair and other project-based industries that buy against a BoQ.