ProjectVIEW AI Agents

DANAOS ProjectVIEW AI Agents are AI agents for ERP in construction, shipbuilding and other project-based industries. Each agent watches one process, for example procurement or cost control, on live ProjectVIEW ERP data. They detect deviations, explain them and recommend an action. Then they request human approval before anything runs. In addition, each client can configure its agents, which are part of ProjectVIEW AI, the AI add-on to ProjectVIEW ERP.

An AI agent in ProjectVIEW is a piece of software with a defined job, access to the ERP data that job needs, and a fixed set of actions it may propose. However, it does not chat for the sake of it. Instead, it monitors a process continuously and reports when something needs a decision.

Example: the procurement agent compares a new quotation with the budget rate on the same BoQ item and cost code. If the quotation is above the budget rate, the agent flags it and explains the gap. Then it recommends a next step and asks the procurement officer to approve or reject.

Agents work on the BoQ ↔ WBS ↔ Cost Codes structure of ProjectVIEW ERP. As a result, an agent can compare an estimate, a purchase order, a subcontract and an actual cost on the same line.

Every agent follows the same seven-step cycle:

1. Detect a deviation in its process.
2. Investigate the related records next.
3. Explain in plain language what is off and why it matters.
4. Recommend an action.
5. Request approval from the person whose role owns the decision.
6. Execute the approved action in the ERP.
7. Audit: finally, log every step, input and decision.

For the technical detail, see how ProjectVIEW AI works.

There are eight purpose-specific agents. Each one maps to a ProjectVIEW ERP module and supports the roles that own each decision. Also, all agents are available with ProjectVIEW v12, and we configure each one per client.

• Bidding & Estimation: comparable items and Cost Recipes from past work; unusual assumptions in a new estimate.
• Cost Control & Forecast: forecast cost at completion, margin erosion, probability of overrun, earned value.
• Subcontractor & Certification: delayed certifications, retention exposure, subcontractors with high commercial risk.
• Procurement: quotation comparison, anomalies, prices above budget rates.
• Materials & Production: material flow and production progress against the WBS.
• Machinery & Maintenance: abnormal fuel use, idle time, CO₂ per activity.
• Cash Flow & Finance: best, expected and worst-case cash flow forecasts.
• Claims & Variations: unapproved variations and the evidence needed for claims.

As a result, each agent aims to raise weak signals early, before they show up in the month-end report.

We write each agent alert for the person who has to decide, not for an analyst. For this reason, each alert aims to contain:

• What the agent found: the deviation, with the project, BoQ item, WBS package and cost code.
• The evidence: the ERP records the agent used, such as the budget line, the quotation and past rates.
• The explanation: why the deviation matters, in plain language.
• The recommendation: the proposed next step and its expected effect on cost or cash.
• The approval request: who must decide, with the option to approve, reject or ask for more detail.

No. Agents work with human approval. They detect, investigate, explain and recommend without waiting. However, every consequential action needs approval from the responsible person, for example committing cost, changing a contract value or posting a financial entry. Also, agents follow the same governance chain as users: Identity → Role → Permission → Data → AI → Action.

We configure the agents per client. During implementation, we set the thresholds, approval routes, projects in scope and alert recipients to match your own processes. However, AI will not fix broken processes. Instead, it will amplify them. That is why configuration starts from your approval workflow, not from the agent.

Yes. The ProjectVIEW MCP connector lets your teams build their own AI agents and assistants on ProjectVIEW ERP data, for example with Claude or ChatGPT. MCP is an open standard for connecting AI assistants to business systems. Also, agents you build through MCP use the same role permissions as the user who runs them. See the ProjectVIEW MCP server.