New: Document Integrations are live. See what your AI can read now

Why good project teams make late decisions, and 3 ways to fix it

Capital projects are run by capable people, yet decisions still arrive weeks after the data. The fix lies in redesigning the system that informs them.

November 18, 2026

4 min read

Sat nav style route across a project schedule, with open RFIs and change orders flagged along the way

Only 0.5% of major projects finish on time, on budget and as promised. Every year, $1.6 trillion is lost to late and over-budget delivery.

It would be easy to blame the people, and it would be wrong. Capital projects are run by experienced, capable teams who know their programs inside out. When results look like this across an entire industry, the cause sits deeper, in the system that informs every decision. If we can’t win, something in the rules is rigged. That is where AI for project controls can help. 

Why project decisions arrive late

Every call on a major project has to be defensible: to leadership, to the client and sometimes to a tribunal. That is how it should be. Accountability demands information that is thorough and auditable.

To understand why decisions arrive late, we traced how a decision actually gets made on a major project, from the moment something happens on site to the moment someone acts on it.

Accountability slows decisions down, for good reason​. It starts with a site visit, which becomes a diary entry. An RFI is logged in one system and a change order in another. Updates are keyed into the schedule, then exported and reconciled. Someone drafts a report, and it is reviewed, rewritten and reviewed again. Only then does the information reach the people who make the call.

Accountability slows decisions down, for good reason​

Every step is there for a reason, and each one is done by capable people. The problem is the sum of them. By the time a decision is made, it is often weeks after the data that should have prompted it, and the project has moved on in the meantime.

Late, incomplete and uncertain information can’t be audited, so people wait. Instead of indecision, it’s accountability.

The solution: redesign the system with AI for project controls

That leaves two options: loosen accountability, or redesign the system that informs it. On a project worth hundreds of millions, only one of those is acceptable. 

Here are three ways to do it.

 

1. Bring all your context into one place

The schedule records what a project intends to do. The reasons dates move live elsewhere, in RFIs, change orders, submittals and correspondence. That's why we've been connecting Nodes & Links to the systems where this context lives, including Aconex, Oracle Unifier and Autodesk Forma. An AI that reads the schedule alongside those records can explain why a date is moving, as well as when.

2. Run it through ground-truth models

Context on its own produces answers that sound right. Running it through proven project controls models, such as quantitative schedule risk analysis, delay analysis and earned value, produces answers you can trace and defend. The expert stays in charge of the decision, and the AI shows its working.

3. Switch on always-on monitoring

Monthly reporting describes what has already happened. Leading indicators, such as how reliably contractors deliver their short-term plans, show where a project is likely to slip next, while it is still cheap to fix.

The result: earlier decisions you can defend

Get all three right and the timing of decisions changes. On a major shut down project, ECS Associates used this approach to trace 42.7% of completion risk to a single work area and surface up to 10 days of delay exposure with two months still to act. The decision came while it still mattered, with the evidence to back it.

The questions change too.

Ask AI Mode “Are we on track?” with the schedule alone and you get a date. Ask with the RFIs, change orders and submittals behind it, and you get the activities at risk, the open items holding them up and what to clear first.

A recovery plan gets tested against what is actually blocking the work.

A claim starts from a cited timeline drawn from the project record.

Reported progress gets checked against inspections and submittals before it reaches the board.

The expert still makes the call, with a fuller picture and more time to act.That is the shift worth aiming for: accountability kept intact, and decisions that arrive in time.

Get the full framework in our free guide

Our new guide, Better decisions: the guide to AI that backs project experts, sets out the three foundations in detail, with 7 questions to ask on your own projects.

To see these foundations on live programs, browse our customer stories or book a demo below.

See it on your projects

Book a demo and we’ll show you what Nodes & Links finds in your own schedules.

Share

Keep reading

More from the blog

Auditable AI - Nodes & Links & The Herne Group
As featured in BD&C: how The Herne Group used Nodes & Links to recover eight weeks and generate £14m in acceleration value across live projects.

November 20, 2026

· 4 min read

new schedule integrity templates - Nodes & Links
Build schedule integrity standards for each type of project in Nodes & Links, apply them across projects and update them in one place.

November 18, 2026

· 4 min read

oracle unifier integration
Nodes & Links now connects to Oracle Unifier, so RFIs and change orders sit beside your Primavera schedule, showing how each change affects project delivery.

November 18, 2026

· 4 min read