User roles: ✔ Project Manager ✔ Project Team Member
Keeping a project current in Fluid means a steady round of small updates: this week's project status report, a new risk, a slipped date, a reassigned schedule item. If you already work in an AI app such as Claude or Microsoft Copilot, you can ask it to make these updates for you in plain language, without opening each screen.
The Project Manager toolset of the Fluid AI connector lets the app read and write project status reports, log and update risks, issues and other impacts, keep the schedule current, and ask FluidAI for an informed view of one project or of all your projects. Every tool runs as you: the app sees only the projects you can see in Fluid, it can change only what your project permission lets you change, and a status report it saves goes through the same rules as the status report screen.
In this article you will learn what you need before you start, how the app finds your project, what each Project Manager tool does, with example prompts, and how to get good results.
Before You Start
MCP access must be on for your site. An Application Administrator turns on Enable MCP Access on MCP Tools and Connected Apps in the Administration Console. See Managing Fluid AI Connectors (MCP) for Administrators.
You need access. Application and Integration Administrators can always connect. Everyone else needs the MCP access role. While Administrators only (evaluation) is on, only Application and Integration Administrators can connect, even people who have the role.
Connect your AI app and sign in to Fluid. On the consent page, reading and Ask FluidAI are ticked for you. Changing status reports, impacts and the schedule, and deleting, stay off unless you tick them. See Connect AI apps to Fluid with MCP.
Your project permission decides what the app can change. You need the Editor permission on the project to change impacts, RAG status updates and schedule items, and Owner to create or update a project status report.
Ask FluidAI needs FluidAI. FluidAI must be switched on for your site.
Once you're connected, you talk to the AI app in plain language. You don't need to know any tool names; the app picks the tools for you. Every tool is listed, with its inputs and the permission it needs, in Fluid MCP (Connectors) Tools Definition.
How the AI app finds your project
Every tool in this set works on a project, and the app finds it with fluid_pm_find_my_projects, which lists the projects you're allocated to in any role (project manager, sponsor or team member). So name your project in the prompt and the app looks it up.
Tip: “In Infrastructure Modernisation, …” is enough. If more than one project matches, the app asks which one you mean.
Finding projects and getting the full picture
Tool | What it does |
|---|---|
fluid_pm_find_my_projects | Lists the projects you're allocated to, optionally narrowed by name. Each result has its portfolio, type, status, and start and end dates. |
fluid_pm_get_projects_brief | Your active projects, the most material first (worst RAG, highest tier, largest estimate at completion), each with its official RAG. Where FluidAI has written a project overview, the brief includes its theme, RAG reason and headline. It makes no AI call. |
fluid_pm_get_project_brief | One project's full reporting context, the same material FluidAI reads: computed signals (overdue milestones, open risks and issues, how long since the last status report, governance result), a project dossier and the headline numbers. Financials are included only if you're allowed to see them. It makes no AI call. |
Example prompts
“List my projects.”
“Which of my projects most need my attention today? Rank them and say why.”
“Give me a one-page brief on Training Portal Implementation: health, overdue milestones and open risks.”
Project status reports
Tool | What it does |
|---|---|
fluid_pm_get_reporting_options | Reads what this project's status reports and impacts accept: status report components, RAG values (and which ones need a path to green), the open risks and issues a component can link to, impact types and sub types, statuses, the Probability and Impact scales, and treatment types. The app reads this first, so it uses your site's own names and values. |
fluid_pm_get_property_definitions | Describes the custom properties a status report and an impact can carry: name, data type, whether it is required, multi-valued or writable through the connector, and its options. The app reads it before it sets a custom property. |
fluid_pm_get_weekly_status | Reads the current project status report, or a specific earlier one: narrative, Overall RAG, each Component RAG with its path to green, and custom properties. |
fluid_pm_update_weekly_status | Creates a new project status report or updates an existing one, including the narrative (strapline, executive summary, achievements, next steps), the Component RAGs and custom properties. Only the fields and components you mention change. You need the Owner permission on the project. |
The tool names keep “weekly” for compatibility; your site sets its own reporting cadence.
The status report screen's rules apply to the AI app too:
A component worse than Green needs a path to green: commentary, a green date after the report date, an owner and, where your site requires it, a linked risk or issue. Where linking is required, a primary component at Red or worse must link to an issue, not a risk.
Where your site calculates the Overall RAG from the components, the app changes the components, not the Overall RAG.
A new report carries the Component RAGs forward from the last published report, with their commentary cleared.
Where your site uses strict status reporting, a report can only be published once achievements and next steps are filled in.
If a request breaks a rule, Fluid refuses it, lists every problem and saves nothing. The app can then ask you for whatever is missing.
Example prompts
“Start this week's status report for Infrastructure Modernisation. Set Timeline to Amber because supplier onboarding slipped two weeks. Jo Bloggs owns the path to green, the green date is 30 November, and link it to the supplier risk.”
“Update the executive summary on the latest report: testing finished, and we're preparing for go-live on 4 December.”
“Set the Confidence custom property on this week's report to Medium.”
“Which components on the current report aren't Green, and what's the path to green for each?”
“Publish the draft status report once achievements and next steps are filled in.”
RAG status updates
Every schedule item, impact and action, and the project itself, keeps its own trail of RAG status updates, separate from the project status report.
Tool | What it does |
|---|---|
fluid_pm_get_status_history | Reads the RAG status updates of a schedule item, impact, action or the project, so you can see how the RAG and its strapline have changed over time. |
fluid_pm_save_status_history | Adds or corrects a single RAG status update. |
fluid_pm_delete_status_history | Deletes a RAG status update. |
Example prompts
“How has the RAG on Training Portal Implementation changed over the last three months?”
“Show me the trend for the Go-live milestone's RAG.”
Risks, issues and other impacts
Impacts are Fluid's RAID log: risks and issues, plus the changes, assumptions, dependencies, decisions and other impact types your site enables.
Tool | What it does |
|---|---|
fluid_pm_get_impact_list | Lists a project's impacts, optionally filtered by type and status, and optionally including child projects. |
fluid_pm_get_impact | Reads one impact in full, including its custom properties. |
fluid_pm_get_impact_summary | A quick roll-up for status reporting: total, open, open risks, open issues, and open impacts at a Red or Amber RAG. |
fluid_pm_get_impact_counts | Impact counts and weightings for a project or portfolio, optionally filtered the way the steering committee pack is. |
fluid_pm_save_impact | Creates an impact, or updates one. An update changes only the fields you mention. Type, status, RAG, Probability, Impact and custom property values are checked against your site's options and refused with the valid choices rather than ignored. Risk Rating is worked out from Impact and Probability. A RAG change is recorded as a RAG status update on the impact, as it is on the impact screen. Closing an impact sets its RAG to Completed. |
fluid_pm_delete_impact | Deletes an impact. |
Example prompts
“Log a risk on Infrastructure Modernisation: the data-centre migration window may slip if the network vendor misses the 15 November cut-over. Probability Likely, impact Major, mitigation: book a fallback window with the vendor.”
“Raise an issue: the UAT environment has been down since Monday. Set it Red, due Friday, and set Workstream to Infrastructure.”
“Move the supplier onboarding risk to Amber with the strapline ‘Contract signed, onboarding under way’.”
“Close the UAT environment issue: it was fixed on Wednesday.”
“What are the open high-severity risks across this programme, including child projects?”
Deleting: a deletion made through an AI app is as real as one you make by hand. Close or archive a risk rather than deleting it, and check carefully before asking the app to delete anything.
Schedule
Tool | What it does |
|---|---|
fluid_pm_get_schedule | Reads the project's schedule items (phases, workstreams, deliverables, actions and milestones) and the dependencies between them. |
fluid_pm_update_task_dates | Changes a schedule item's start date, end date or both. |
fluid_pm_update_task_effort | Changes a schedule item's effort. As in the schedule, the end date is recalculated from the start date plus the new effort. It is not for milestones, or for parents whose dates are calculated. |
fluid_pm_update_task_description | Replaces a schedule item's description. |
fluid_pm_update_task_assignees | Replaces a schedule item's Assigned To list. An action must keep at least one assignee. |
Example prompts
“Which milestones on Training Portal Implementation are overdue, or due in the next 30 days?”
“Move the end date of High Level Plan Signoff to 21 November.”
“Assign Sam Patel and Alex Chen to the Data migration dry run schedule item.”
“Add to the description of the cut-over item: ‘Requires change board approval two days before.’”
Ask FluidAI
Some questions don't map to a single record, such as “Is this project on track?” or “Where's the biggest risk exposure across my projects?”. For these the AI app can pass the question to FluidAI, the AI assistant built into Fluid. FluidAI already knows how to gather and summarise your project data, so the app doesn't have to read and analyse every record itself.
Tool | What it does |
|---|---|
fluid_pm_ai_ask_project | Asks FluidAI's project assistant about one project. It answers from the project dossier, and passes risk and schedule questions to its risk and schedule assistants. |
fluid_pm_ai_ask_projects | Asks a FluidAI portfolio assistant about all your projects. It can focus on an overview (health, status, themes, recommendations), on risk (risks and issues) or on the schedule (milestones, slippage, what's due). The assistant first picks the projects or items the question is about, then answers using just those. |
Good to know
Read only. FluidAI won't change anything. If you ask it for a change, it says so, and the AI app makes the change with the tools above.
Answers show their sources. Each answer lists the projects or items FluidAI used, so the app can check them.
Follow-up questions continue the same conversation. A question can be up to 2,000 characters.
It costs more. Each question uses AI and typically takes 10 to 60 seconds, and there's a per-person limit on questions per minute. For simple facts, the briefs and read tools are faster.
It is on unless you untick it. Ask FluidAI is ticked with the read permissions on the consent page. Untick it when you connect if you don't want the app to use FluidAI.
FluidAI must be enabled on your site. The risk and schedule focus needs the portfolio risk and schedule assistants switched on as well.
Example prompts
“Ask FluidAI whether Infrastructure Modernisation is on track, and what I should focus on this week.”
“Follow up: what's driving the first of those?”
“Across my projects, what are the highest-scoring open risks?”
“Across my projects, which milestones are overdue?”
“Using FluidAI, draft the executive summary for this week's report on Training Portal Implementation, then show it to me before saving it.” (FluidAI drafts, you review, and the app saves the result with fluid_pm_update_weekly_status.)
Good to Know
Ask the app to show you before it saves. For example: “Draft the update and show it to me first.”
Use your site's own words for components, RAG values and custom properties. The app checks them against fluid_pm_get_reporting_options and fluid_pm_get_property_definitions, and if a value doesn't exist it's told the valid choices.
Name people clearly. If an owner's name matches more than one person, Fluid refuses it and lists the candidates rather than guessing.
Check AI answers. FluidAI answers are AI-generated. Check anything important against the project before acting on it.
A missing tool usually means you did not tick its permission when you connected, or your administrator has switched it off.