Why traditional risk management fails on capital projects
The numbers are remarkably consistent across studies of large capital projects: roughly two thirds miss budget, four out of five miss schedule, and more than half deliver less than planned scope. McKinsey, IPA, and the Oxford Saïd Business School have all measured the same outcome from different angles for over twenty years. The risk-register-and-RAG-status approach most teams still rely on isn't failing because people aren't trying — it's failing because deterministic single-point estimates can't capture how real projects unfold.
A traditional risk register lists risks. It rarely tells you which risks actually move the finish date, which ones are correlated, or which mitigation has the highest expected return on management time. By the time the warning signs are obvious enough to make it to the steering-committee deck, there are usually months of compressed float, eroded contingency, and re-baselined milestones standing between you and the original commitment.
How AI risk management actually works
Capital Project AI runs four engines on your project data, each answering a different risk question. You upload a schedule once; the engines do the simulation, the sensitivity analysis, and the surfacing of unfair bets — automatically, in seconds.
Monte Carlo schedule simulation
The schedule engine treats every activity duration as a probability distribution rather than a fixed number. It runs thousands of full-project simulations and returns a P50 / P80 finish date with the activities that contribute most to slippage. You see which paths are actually critical, not just the ones the planner drew with the longest bars.
Cost overrun forecasting
The financial engine combines schedule risk with cost loadings to forecast a probability distribution over total installed cost. It separates known risks (commodity exposure, currency, escalation) from unknown unknowns (estimating bias, late scope changes) and tells you where contingency is mathematically defensible.
Sensitivity and tornado analysis
The Attention engine ranks every input by its marginal impact on the outcome. You learn whether your finish date is more sensitive to vendor lead times, weather windows, or commissioning duration — and you focus management attention accordingly.
Portfolio-level risk view
When you load multiple projects, the Kelly engine computes the joint risk of the portfolio: which projects are correlated, where you have concentration risk, and which moves would actually shift the portfolio's risk-adjusted return.
Why Capital Project AI
- Built by capital project operators, not finance MBAs. Founded by an engineer who managed $800M in delivered megaprojects at Shell and partnered with a Penn State researcher to formalize the math.
- Fast enough to be usable. Thousands of scenarios in seconds. Re-run risk analysis after every steering committee, not once a quarter.
- Schedule-native. Reads P6, MS Project, CSV, and Excel directly. No data-warehouse project required first.
- Quantified output. P50 and P80 numbers, contribution by activity, and a clear ranking of where contingency belongs. Not a heat map.
See your real risk profile in 60 seconds
Upload a schedule and get a probability distribution over your finish date — for free.
Open the Dashboard →What it looks like in practice
A typical pattern: an EPC delivery team uploads the current P6 schedule for a $300M brownfield expansion. The schedule shows a 22-month duration with a 4-week management reserve. Capital Project AI runs Monte Carlo on the activity durations and reports a P50 finish at 24 months and a P80 at 28 months, with the dominant slippage drivers being long-lead heat exchanger fabrication and FAT durations on a tagging package the planner had assumed would compress.
The team uses the output two ways. They renegotiate the long-lead delivery to start the fabrication clock earlier, which compresses the P80 by six weeks. And they stop arguing about whether the schedule is achievable — the conversation becomes whether to commit to the P50 with a real buffer or the P80 with a smaller one. That's the conversation a capital committee actually wants.
Closely related: see how probabilistic project planning handles the same problem from the planning side, and how the same approach scales to EPC contractors bidding fixed-price work.
Frequently asked questions
What is AI capital project risk management?
It uses simulation, machine learning, and probabilistic forecasting to identify schedule, cost, and execution risks across a capital project portfolio — before those risks turn into overruns. Capital Project AI runs thousands of Monte Carlo scenarios on your live schedule and cost data to surface what could go wrong, what's worth the risk, and how much capital to commit.
Does it integrate with Primavera P6 or MS Project?
Yes. Capital Project AI accepts schedules exported from Primavera P6, MS Project, and other planners as XER, XML, CSV, or Excel. The risk engines run on your data without requiring you to migrate platforms.
How is this different from a traditional risk register?
Risk registers list risks; Capital Project AI quantifies them. Each risk gets a probability distribution, an impact estimate, and a contribution to your P50 / P80 finish date and cost. You see which risks actually move the needle on outcomes — not just which ones look scary.