Project margin is rarely where you expect it. The small low-pressure job turns out loss-making, the prestigious big project barely returns its costs, and that one routine engagement is suddenly a gold mine. The problem is not that data is missing — Exact Online holds all hours, costs and invoices neatly registered. The problem is that you only realise it once you have actually run the report.
With an AI assistant via MCP you pull these insights out in seconds. In this article: how to analyse project costs and margins, spot overruns, and write better quotes based on your own historical data.
What Exact Online knows about your projects
Per project Exact Online holds:
- Hours per employee, linked to a cost rate;
- Direct costs (materials, purchases, subcontracting);
- Invoiced revenue;
- Planned versus actual — both in hours and in money.
All of this is queryable through MCP tools (projects_get, projects_timecost_search, projects_planning_*). Your AI combines them into usable insights.
Three questions worth asking every quarter
1. Which projects are running over budget?
"Give all ongoing projects where actual hours exceed the plan by more than 10%."
The AI compares planned and actual, filters above your threshold and presents a ranking with absolute and relative deviation. Distinguish overruns you can still correct from projects where the loss is already locked in.
2. What is the actual margin per project?
"For all closed projects in Q2, compute the margin: invoiced revenue minus hour costs minus direct costs."
The AI pulls three figures per project and presents the net result. Sort by margin and you immediately see which engagement types actually pay off.
3. Which customer relationships are profitable?
"Which five customers delivered the highest project margin in 2025, and which five the lowest?"
Customer selection becomes data-driven. Loss-making customers are a commercial choice: either raise pricing or wind down the relationship.
Catching overruns in time
The biggest gain is not in reporting after the fact, but in correcting in time. Set up a weekly routine: every Monday ask the same question:
"Which projects have hit 80% of budget this week while the plan says they should be at 60%?"
An early signal — before the gap grows. Combine with our tips in automating workflows for a complete management routine.
Pricing quotes based on real history
An often overlooked application: let the AI use your historical data for sharper quotes. Ask for example:
"What was the average lead time and margin of comparable projects in the installation sector between €25,000 and €75,000 in 2025?"
The AI pulls comparable projects, computes averages and median, and gives you a data point for your next quote. No more gut feeling, just your own data.
Employee cost per project
Per employee an average cost rate versus actual billable utilisation:
"Which employees had the highest billable utilisation in Q2, and on which projects?"
Result: a utilisation dashboard. Critical for consultancy, IT and construction-type businesses where people are your main resource.
Tips for reliable numbers
- Make sure every employee posts hours in time on the right project — read our time tracking with AI guide for a faster workflow.
- Keep direct costs consistently at project level, not on a catch-all account.
- Close projects formally in Exact Online — only then is the margin definitive and useful for analysis.
Getting started
Project management with AI is not magic — it only works when your administration is in order. But when it is, the combination delivers minutes of insight instead of hours. Check the MCP setup guide and the prompt library for more examples. Want to move into cashflow analysis next? Read our blog on cashflow forecasting with AI.