A Personal AI Agent for Every Business Role: ProjectVIEW AI

Personal AI agents such as Claude Duet and ChatGPT dots work for one person: they read that person’s email, calendar and apps, and they take care of personal tasks. ProjectVIEW AI gives every business user a personal AI agent, because it inherits each user’s profile and role from ProjectVIEW ERP. So a project manager, a cost controller and a CFO each get their own AI assistant, working on the company’s records within their own role, approvals and a private environment.

What are Claude Duet and ChatGPT dots?

Both are always-on AI agents launched in autumn 2026. They run in the background on their own cloud computer and act for one user.

• Claude Duet (Anthropic, research preview since October 2026) sends a short morning brief, checks the user’s connected apps every few hours and carries out tasks. It asks before it sends, pays, signs or deletes anything.
• ChatGPT dots (OpenAI, launched in September 2026) gives each user a personal agent, called a “dot”. It works across apps such as Gmail, Google Calendar, Slack and Teams.

In short, these agents are personal assistants with good judgment. They answer a simple question very well: “What do I need to do today, and can you do some of it for me?”

Why can’t a personal AI agent run business operations?

A personal agent sees what one person sees: an inbox, a calendar, some files. However, business operations do not live in an inbox. They live in structured records that many people create over months: bills of quantities, schedules, cost codes, contracts, certificates and approvals.

Consequently, a personal agent faces three limits in project-based companies:

1. It does not know your operating model. It sees emails about a project, not the cost structure behind it.
2. It works with one person’s access. Business decisions need company-wide data, filtered by role, not one user’s mailbox.
3. It sits outside your governance. Approvals, segregation of duties and audit trails belong in the business system, not in a personal assistant.

The table below sums up the split.

	Personal AI agent (Claude Duet, ChatGPT dots)	ProjectVIEW AI
Works for	One user, in private life and work	Each business user, in his or her business role
Personalised by	The user’s own apps and accounts	The user and role profile inherited from ProjectVIEW ERP
Reads	The user’s email, calendar and connected apps	The company’s ERP records: BoQ, WBS, cost codes, contracts, cash
Typical task	“Reply to the client and book the site meeting”	“Why is Project X losing margin, and what should we do?”
Knows your operating model	No	Yes — from your own ERP structure
Access control	The user’s own accounts	Role-based, at module, function, field and record level
Approvals and audit	Asks the user before key actions	Company approval workflows and audit trail
Runs where	The vendor’s cloud	Private environment on Microsoft Azure or on-premises

How does ProjectVIEW AI become personal for each business user?

ProjectVIEW AI does not create a second set of users. Instead, it inherits each user’s profile from ProjectVIEW ERP. As a result, the AI knows who is asking and what that person is allowed to see and do, from the first day.

The AI inherits five things from the ERP profile:

1. The user account: who is asking.
2. The role: for example project manager, cost controller, procurement manager or CFO.
3. The assignments: the companies, joint ventures and projects the user works on.
4. The permissions: access at module, function, field and record level.
5. The place in workflows: which documents and decisions the user prepares, reviews or approves.

Therefore, the same question gets a different, role-correct answer. For example, ask “What needs my attention this week?”:

• A project manager sees slipping activities and open variations on his own projects.
• A cost controller sees cost codes trending over budget across her portfolio.
• A CFO sees cash exposure by company and joint venture, without line-level site detail.

This role-based assistant is live in ProjectVIEW ERP today, and you can see it in a demo on request.

In short, Claude Duet and ChatGPT dots personalise AI around a person’s private apps. ProjectVIEW AI personalises AI around a person’s business role. And because the profile comes from the ERP, there is no separate permission set to maintain, and nobody sees more through the AI than they see in the ERP.

Why does business AI need to understand each company separately?

Two contractors in the same market rarely run the same way. One self-performs civil works; the other subcontracts almost everything. One splits a project into 300 cost codes; the other uses 3,000. Their approval chains, pricing logic and joint-venture structures differ as well.

For example, both companies ask the same question: “Is the tower project overrunning?”

• Contractor A self-performs the frame. Its overrun sits in labour cost codes, because productivity on site is below the rate in the budget.
• Contractor B subcontracts the frame. Its overrun comes from variations that the site approved but nobody certified, so the client has not been charged.

A generic AI gives both companies the same, average answer. In a business with thin margins, an average answer is often a wrong answer. ProjectVIEW AI gives each company a different answer, because it traces the cause in each company’s own records.

How does ProjectVIEW ERP give the AI an accurate model of your business?

The AI is only as good as the structure it reads. That is why ProjectVIEW AI works on top of ProjectVIEW ERP, the integrated ERP built on the BoQ ↔ WBS ↔ Cost Codes framework.

• BoQ (Bill of Quantities): what you sell and get paid for.
• WBS (Work Breakdown Structure): how you plan and build it.
• Cost codes: what it costs, line by line.

ProjectVIEW ERP links these three structures in one record. In addition, its system configurator and workflow engine shape processes and approvals to the way each company works. So the AI does not guess your operating model; it reads it.

ProjectVIEW AI combines domain-specific small language models with leading large language models. The models retrieve ERP context at the moment you ask, so no retraining on client data is needed. As a result, answers reflect your latest records. Read more about ProjectVIEW AI.

Once the AI reads your structure, its agents are designed to flag cost overruns per cost code, compare subcontractor claims with measured progress, collect evidence for claims and variations, and compare new bid rates with past projects. Agents are rolled out by module and confirmed per client. See how this connects to cost control in ProjectVIEW ERP.

What happens to your data when you ask ProjectVIEW AI a question?

Operational data holds bid prices, margins, subcontract terms, payroll and claims strategy. Therefore, security has to be part of the design. Each question follows six steps:

1. You ask a question in the ProjectVIEW AI assistant.
2. ProjectVIEW AI applies your ERP user and role profile. You only reach the modules, fields and records your role allows.
3. The AI retrieves the relevant ERP records at that moment.
4. The model runs in a private environment on Microsoft Azure or on your own servers. Data is not sent to public consumer AI services.
5. You get an answer and, where useful, a proposed action. A person approves every consequential action.
6. The decision stays on record in the audit trail.

On Microsoft Azure, prompts are not used for model training. Data residency options, encryption at rest and customer-managed keys are available. DANAOS Projects holds ISO 27001, ISO 9001, ISO 14001 and ISO 20000-1. For details, see ProjectVIEW AI data security and governance.

How do personal agents and ProjectVIEW AI work together?

The two layers complement each other. A project director might use both in one morning:

1. Her personal agent prepares a morning brief, drafts replies and books a meeting with the client.
2. ProjectVIEW AI flags that two subcontract packages are trending over budget and shows why.
3. She approves the corrective action in ProjectVIEW ERP, and the decision goes on record.

In addition, ProjectVIEW ERP v12 includes an MCP connector for Claude and ChatGPT. MCP (Model Context Protocol) is an open standard that lets AI assistants connect to business systems in a controlled way. Companies that use it should connect through business or enterprise AI workspaces and check each plan’s data terms, because data leaves the private environment when an external assistant reads it.

What mistakes should companies avoid?

• Pasting operational data into personal AI tools. It feels fast, but bid prices and margins leave company control.
• Running AI on exports instead of live records. Exports go stale within days, so the answers do too.
• Starting AI before the cost structure is clean. If BoQ, WBS and cost codes do not connect, no AI can explain a margin.
• Skipping approvals. AI should prepare decisions; people should make them.

How should executives decide which AI to use for what?

Use one simple rule. If the task belongs to one person — email, calendar, reminders, personal research — a personal agent is a good fit. If the task belongs to the business — cost, schedule, contracts, cash, risk — it needs AI on your system of record, inside your governance.

A full ProjectVIEW ERP implementation takes 5 to 7 months. ProjectVIEW AI can also run on top of existing systems such as SAP, Oracle or Primavera P6, so you can start with the area that hurts most. Pricing depends on the modules, the number of users and the deployment option.

FAQ

Frequently asked questions

Is Claude Duet or ChatGPT dots a replacement for an ERP AI assistant?

No. They are personal agents that act for one user across that user’s apps. Business operations need AI that reads the company’s system of record within roles, approvals and audit.

Is ProjectVIEW AI personalised for each user?

Yes. ProjectVIEW AI inherits each user’s profile, role, project assignments and permissions from ProjectVIEW ERP. So every business user gets answers and proposed actions that fit his or her role, and never sees more than the ERP allows.

Can ProjectVIEW AI work with Claude or ChatGPT?

Yes. ProjectVIEW ERP v12 includes an MCP connector for Claude and ChatGPT. Use business or enterprise workspaces and check each plan’s data terms.

Is our data used to train AI models?

On Microsoft Azure, prompts are not used for model training. ProjectVIEW AI retrieves ERP context when you ask, so no retraining on client data is needed.

Does ProjectVIEW AI make decisions on its own?

No. It prepares analysis and proposed actions. A person approves every consequential action, and the decision stays on record.