In brief: Your connected AI client reads your Jobs, effort, time, contacts, tasks, and lists, and can create or change records your role allows.

Applies to: Anyone with an MCP connection on an active trial or a paid plan. Answers are bounded by your own role.

See also: Connect Your Own AI to Net Net with the MCP Server covers why the connection works this way, Setting Up the MCP Server in Claude, ChatGPT, and More covers making it, and User Roles Explained: Owner, Admin, Lead, and Member covers what limits the answers.

Once your assistant is connected, you can ask it about your Jobs, Deliverables, planned effort, logged time, contacts, tasks, and lists, and it can create and change records that your role already allows. The useful questions are not the ones you would type into a search box. They are the ones you would ask a colleague who had read everything and had no reason to soften the answer. This article groups real examples by what you are actually trying to find out, and it is honest about the places your AI cannot reach yet.

Everything below assumes a working connection. If you do not have one, start at Setting Up the MCP Server in Claude, ChatGPT, and More. For why the connection works the way it does, see Connect Your Own AI to Net Net with the MCP Server.

You do not need to phrase these precisely. Your assistant works out which parts of your workspace to look at. Ask the way you would speak.

Reading your Jobs

This is where most people start, and it is where the difference between a dashboard and an assistant shows up fastest. A dashboard shows you a number. An assistant tells you which number should worry you.

• "List my active Jobs and tell me which ones have the least planned effort left against them."
• "Summarize where the Montage video platform Job stands right now."
• "Pull the net net summary for this Job and tell me where the margin is going."
• "Which of my Jobs have Deliverables due in the next two weeks?"
• "Compare planned Programming effort against what has actually been logged on this Job, and tell me whether the gap is normal for this stage."
• "Which service types are consistently running over their planned effort across all my active Jobs?"

That last one is the sort of question that is genuinely hard to answer by clicking. It requires reading across every Job, grouping by service type, and comparing plan to actual. Your assistant does it in one pass.

Each of those answers has a home in the product too, and it helps to know which one your assistant is reading from. The single Job view is Reading the Job Performance Tab: Plan, Actual, Drift, the across all Jobs view is the Jobs Report, and due dates and confidence come from Deliverables: Status, Due Dates, and Confidence Explained. Retainers answer these questions period by period rather than across a whole plan, which Running Retainers: Monthly Cycles, Tasks, and Actuals explains.

Time and the team

Effort is the unit Net Net speaks in, so this is the richest seam.

• "Who logged time this week, and against which service types?"
• "Show me every time entry on this Job in July, grouped by person."
• "Has anyone logged time against a Deliverable that is already closed?"
• "Which of my people have logged the most effort against Jobs that are drifting?"
• "Log two hours to my design task from yesterday afternoon."
• "I worked on the Acme rebuild for most of Tuesday. Put that in for me."

That last pair matters more than it looks. The reason time tracking fails in most shops is not that people object to it. It is that logging time means stopping, switching context, and finding the right row. Telling your assistant what you did in one sentence removes that.

More on how effort and drift work together is in How Performance Works: Effort, Timeline, and Drift. To see the same numbers where they live, My Performance is your own effort in one place, and the Time Report is the workspace wide version your assistant is reading when you ask who logged what.

Contacts

Your assistant can read and maintain the Companies and People in your workspace.

• "Find the company Montage and show me everyone we have on file there."
• "Do we already have a contact at Acme, and who is it?"
• "Add Priya Raman as a new person at Acme, title Marketing Director."
• "Update that person's email address to the new one."
• "Export our people list so I can review it in a spreadsheet."

Contacts in Net Net separate Companies, People, and Individuals in a specific way, and knowing that shape makes your questions land better. It is covered in Companies, People, and Individuals: Contact Records.

Capturing work

Quick Tasks and My Lists are the two places small things go so they stop living in your head. Both are reachable.

• "Add a Quick Task to chase the missing brand assets from Acme."
• "Turn everything I just listed into Quick Tasks."
• "Put that idea in my lists so I do not lose it."
• "What is sitting in my lists at the moment?"
• "Create the build tasks for this Deliverable and assign them to me."

The difference between these three homes for a task is worth understanding before you start creating them in bulk. See Quick Tasks vs Job Tasks vs My Lists: Which to Use. Tasks that sit inside a Job follow their own rules, set out in Job Tasks: Assignees, Allocations, and Statuses, and the private capture side is My Lists.

Orienting the connection

Short, boring, and the first thing to ask when something looks wrong.

• "Which Net Net workspace am I connected to right now?"
• "Which workspaces can you see?"
• "What is my role in this workspace?"

If your login belongs to more than one workspace, ask this before you trust an answer. Name the workspace explicitly in your questions rather than assuming your assistant chose the one you had in mind. What a given answer means for what you can then do is in User Roles Explained: Owner, Admin, Lead, and Member.

The questions worth connecting for

Everything above could, with enough patience, be answered by clicking around. These cannot, because they need your workspace data and your assistant's judgment at the same time.

• "Read where this Job actually stands and draft a status email to the client that is honest without being alarming."
• "Based on what has been logged so far, is our original planned effort for this service type systematically wrong, or was this Job unusual?"
• "Here is the brief for a new project. Compare it to the last three Jobs of this shape and tell me what our planned effort should realistically be."
• "Write me a short Monday summary of everything that moved last week and everything that did not."
• "Which client asks for the most unplanned work, and can you show me the evidence for that?"

This is the reason bring your own AI exists rather than a built in chat panel. Your assistant is holding your workspace alongside the proposal you were drafting, the email thread you pasted this morning, and everything else in that conversation. Net Net supplies the truth about your work. Your assistant supplies the reading of it.

If you want the wider version of that argument, written for agencies rather than for this help center, how ChatGPT can help automate your agency workflow makes the case.

What your AI cannot do

Being straight about the edges saves you a frustrating afternoon.

There is nothing to ask about Sales yet. Sales is coming to Net Net in Q3 2026 and is not in the app, so there is no pipeline and there are no Opportunities anywhere for your assistant to read. This is not an MCP gap, it is a module that has not shipped. The preview of how it will work is in Sales Opportunities, Not a CRM: A Q3 2026 Preview.

Billing and payment are not exposed. Your assistant cannot see your invoices, change your plan, or read your payment details, and it cannot be talked into doing so.

It cannot exceed your role. If something is closed to you in the product, it is closed to your assistant, because the connection carries your permissions. That also means it cannot do things on behalf of a colleague with wider access than yours.

It is not a second opinion on its own output. Your assistant reports what the data says. Whether the data is right is still your call, and if your service types or planned effort were set up loosely, the answers inherit that.

Frequently Asked Questions

Do I have to phrase questions in a special way?

No. Plain language works, and vague is usually fine. Your assistant works out which parts of your workspace to look at. The one thing worth being specific about is which Job or which workspace you mean, because guessing wrong there produces a confident answer about the wrong thing.

Can my AI log time for me?

Yes. Describing what you worked on in a sentence is enough. This is one of the more useful habits to build, because the reason effort goes unlogged is almost never unwillingness. It is the interruption of stopping to find the right row.

Can it see other people's time and performance?

Only if your role can. A Member sees their own. Admins and Owners see across the team. The connection inherits whatever you already have, so the answer is the same as it would be if you looked yourself. See User Roles Explained: Owner, Admin, Lead, and Member.

Can I ask it about my sales pipeline?

Not yet, because Sales has not shipped. It is coming in Q3 2026. There are no Opportunities in the product today, so no amount of rewording will produce an answer. Everything your assistant can reach is listed above.

What if it gives me a number I do not believe?

Ask it to show its working, then check that one record in the product. Usually the number is right and the surprise is real, which is the entire point. Occasionally it has read across a boundary you did not intend, such as including archived work, and asking it to narrow the range settles it quickly.

Can I stop it from changing anything and let it only read?

Not today. The connection carries your role whole, so an assistant connected as you can do what you can do. If read only access matters for how you want to work, connect using a login whose role is already limited to what you are comfortable with rather than expecting the connection to narrow it for you.

Related articles

If your assistant is being refused rather than answering badly, AI Access and Your Plan: Why MCP Needs a Paid Plan is almost always the reason. When the answers point at a Job that is slipping, Responding to Drift: What You Can Actually Do covers the moves available to you, and Which Tasks Can Accept Time (and Which Cannot) explains why a request to log effort against a particular task is sometimes turned down. For the bigger picture of what this connection is and what it inherits from you, go back to Connect Your Own AI to Net Net with the MCP Server.

Updated on: 08/04/2026