In project-based firms, margin is won or lost while the work is happening. Yet most firms find out a project is in trouble at month-end, when the budget is already spent and the moment to act has passed. The data to warn them usually exists. What’s missing is someone seeing it early enough, at the right level of detail, with clear ownership of what happens next.
Does this sound familiar?
Projects look fine until late, then go over budget.
Change orders get raised after the cost is already gone.
Reports get produced, but nobody acts on them.
A few people hold the real picture in their heads, and everyone depends on them.
If two or more of these are true, the problem is rarely effort. It’s timing.
Why problems show up late
The facts are scattered. The numbers live in the ERP. The context lives in emails, documents and people’s heads. No one view brings them together while the work is underway.
Reports run on the wrong clock. Most project reports arrive monthly and look at the whole project. A problem in one part of the work can hide inside a healthy-looking total for weeks.
Nobody owns the response. Even when a warning sign appears, it often lands in a report everyone reads and no one acts on. Without a named owner, a deadline and a record of the decision, the same problems repeat.
The window closes quietly. A change order, a staffing shift or a conversation with the client can protect margin, but only before the cost is locked in. By month-end, that window has usually closed.
What an earlier warning looks like
An illustrative example with made-up numbers.
On Monday morning, a project manager gets an alert: one part of a project has used 70% of its budget but is only 40% complete. The alert shows where the numbers came from and when they were last updated. She sees the cause, a client request that was never written up as a change, and raises a change order that week. The decision is recorded. Had she not responded, the alert would have gone to her manager. A month later, the firm checks whether the change order recovered the cost.
Nothing in that example needs a new ERP. It needs the right signal, at the right time, in front of the person who can act.
What makes an early warning worth trusting
A warning sign is only useful if it arrives while there’s still time to act and if people believe it. In practice, that means:
Spot the issue as soon as it appears, with the evidence behind it.
Explain it in plain terms, including what isn’t known yet.
Assign it to a named owner, who decides what to do.
Follow up, and escalate if no one responds or things get worse.
Record the decision and what happened.
Improve the signal based on the result.
Two alert levels help keep people responding. A heads-up flags early signs and asks for nothing. A confirmed alert asks for an owner, an action and a date. The firm signs off its own definitions and thresholds, and every number shows when it was produced and the period it covers.
AI can help spot patterns, explain them in plain language and route them to the right person. The numbers themselves should come from the firm’s own approved calculations. For more on why projects like this stall between pilot and everyday use, see Why Most AI Projects Get Stuck in Pilot Mode.
Where to start
You don’t need to fix everything at once. Start with one decision that tends to go wrong late, and test whether an earlier warning would have changed it, using your own project history.
Which decision in your firm tends to go wrong late?
Book a 15-min AI Readiness Call
A brief conversation to understand your goals and determine the right starting point.
Or explore the AI Workflow Readiness Audit.
Frequently asked questions
Why do project overruns show up so late?
Because most project reporting is monthly and summarized at the project level. A problem in one part of the work can hide inside a healthy total until the budget is spent and the chance to recover it has passed.
What is the earliest sign a project is losing margin?
It varies by firm, but common early signs are budget being used faster than work is completed, client requests that haven’t been written up as changes, and time logged to the wrong cost code. The useful test is which sign would have warned you early on your own past projects.
Do we need new software to catch problems earlier?
Not necessarily. Most firms already hold the data in their ERP and project records. The gap is usually how often it’s looked at, at what level of detail, and who is accountable for acting on it.
Why do reports get produced but not acted on?
Usually because no one owns the response. An early warning works when it goes to a named person, asks for a decision by a date, and records what happened.
Where should a firm start?
With one decision that tends to go wrong late. Test whether an earlier warning would have changed it on your own project history, before building anything bigger.