Can AI-enabled project management rewrite a project schedule the moment a problem occurs? Researchers at the University of East London (UEL) believe so and have proposed a system architecture that detects emerging risks and automatically adjusts project plans before delays spread across a site.
Rather than proposing a single new tool, the UEL research outlines how existing technologies could be connected in ways they are not currently. Typically, safety monitoring systems, digital risk registers and scheduling platforms operate in isolation.
As a result, risks are identified, but the project timetable often continues unchanged. The research proposes a framework showing how risk warnings could trigger immediate, machine‑readable planning decisions.
Lead author Dr Jawed Qureshi, senior lecturer in structural engineering at UEL, said construction projects already generate the information needed to prevent delays, just not in a form that scheduling systems can use.
“Projects generate enormous amounts of warning data every day – safety alerts, design clashes, supply delays and contractual risks – but nothing in the schedule actually changes when these signals appear,” Qureshi explained. “Our work shows how those signals can be converted into scheduling constraints so the plan adapts before delays escalate.”
The research focuses on connecting two areas of AI: systems that predict risks; and systems that optimise schedules. At present, they function like separate dashboards – one forecasting problems, the other planning work – with no automated link between them, Qureshi added.
The ‘risk-to-constraint translation engine’
The proposed solution is a mechanism that he and co-author Kiran Rai call a “risk‑to‑constraint translation engine”. Instead of simply logging an issue for later review, the system would translate a detected risk into a practical project constraint that scheduling software can act on.
Examples cited include:
- a safety hazard detected by computer vision could temporarily halt specific tasks;
- a predicted material delay could automatically resequence dependent activities; and
- a contractual risk identified through natural language processing could introduce additional time allowances.
These changes would then be tested inside a digital twin of the project, allowing managers to review the consequences and approve the most effective option before the real schedule is affected.
The authors emphasise that their aim is not to remove human oversight, but to close the gap between early warning and actionable response. “At the moment, we manage projects like we drive while looking in the rear‑view mirror,” Qureshi said. “This approach creates a forward‑looking process where risk detection and planning become one continuous activity.”
