Skip to content
English
  • There are no suggestions because the search field is empty.

Recipe: Which projects need attention, and why?

Use Agentic AI Insights to find all projects at risk, ranked by the money at stake

Published: 2 November 2026

Audience

  • Operations Leads

Objectives

This article explains how to find projects at risk with Claude or ChatGPT. The recipe scans all active customer projects for trouble and ranks them by the money at stake, using the alert thresholds you set yourself in PSOhub.

Prerequisites

The MCP connector is active on your organization. 
        An Admin has enabled the MCP connector for your organization.

Your AI assistant is connected.
        Your AI assistant (Claude or ChatGPT) should be connected to PSOhub.
        See How to Connect Claude to PSOhub or How to Connect ChatGPT to PSOhub.

Steps

  1. Copy the prompt exactly as written:
  2.    Which of our active customer projects are in trouble?

       A project is at risk when a contract line is past the alert threshold we set on it or over its budget, when its WIP is more than 25% of the contract value, or when its expected end date has passed. A project without an end date is not on time: say how many had none.

       Rank the projects at risk by the money at stake. For each: customer, project manager, business unit, what put it on the list, and one sentence on why. List the rest as watch or fine, one line each.

       End with the pattern: do the projects at risk share a customer, project type, business unit or project manager?

       Show this as a dashboard: a header with the scope and date, a one-line verdict with three to five key figures, the findings ranked worst first with the reason and the amount, a table with the rest, what this answer could not see, and three actions with an owner. Use a single HTML page if you can show one, otherwise the same blocks as formatted text. Link every project, customer, person and invoice to PSOhub.

     

  3. Paste the prompt into Claude or ChatGPT and send it.

📌 Note: This works best when projects have end dates and contract lines have budgets and alert thresholds. Run Recipe: Is our portfolio ready for deeper analysis? first.

Result

You get a dashboard with the projects at risk ranked by the money at stake, with the reason for each and the rest grouped as watch or fine. It ends with the pattern the projects at risk share.

💡 Tip: Ask a follow-up question, for example:

  • Show only the fixed fee projects.
  • Which of these were already at risk a month ago?
  • What is the total exposure if they all keep going like this?

Related articles