Finding the real critical path on a capital project
A late construction program, a schedule with 4,000 lines, and the AI-assisted review that found the three constraints nobody was talking about.
Capital program, engineering and construction
ThemesThe situation
A multi-contractor construction program was reporting a six-month slip and asking for more money. The integrated schedule had grown to thousands of activities across several contractor files. Nobody in the program office could say with confidence what was really driving the date.
What changed
An independent assessment combined classic schedule quality checks with AI-assisted reading of contractor narratives, meeting minutes and change logs. The machine work was to cross-reference thousands of pages for the same three questions: what is constrained, what is assumed, and what has changed. The human work was to sit with the planners and test what the analysis suggested.
Three constraints emerged that were not on any risk register: a permit sequence that forced a winter outage, a long-lead item ordered against an outdated specification, and a shared crane between two contractors with no agreed schedule.
The result
The re-plan recovered part of the slip by re-sequencing around the permit window and resolving the crane conflict in one meeting. The remaining delay was presented to the board with a quantified risk range instead of a single date, which the board accepted.
What it teaches
- AI is excellent at reading everything. People are needed to decide what it means.
- Most schedule risk lives in documents, not in the schedule file.
- Boards accept ranges when they come with a plan.
A composite drawn from our engagement work and published practice. Details are generalized and no client is identified.
More case studies
The PMO that stopped writing status reports
How a portfolio office replaced two days of weekly copy-and-paste with an agent that drafts, a person who checks, and a leadership team that finally trusts the numbers.
Meetings that produce decisions, not minutes
A transformation program drowning in meetings used AI note-taking with a strict governance brief, and found that the real gain was in what it chose not to automate.
Writing an AI policy project teams actually follow
Why a two-page policy with examples beat a forty-page one, and how a delivery organization moved from “AI is banned” to “AI is governed” in a quarter.
Tell us about the project, the PMO or the decision in front of you.
A 30-minute call. We listen, we ask questions, and we tell you honestly whether we can help.