Why EPC margins are squeezed
EPC contracts price risk poorly. The lowest bid wins, contingency is whatever survived the BD pressure, and schedule assumptions get inherited from whoever produced the original tender pack. Two jobs go sideways out of every ten and the firm is fighting for breakeven on the portfolio. The data to do better — historical productivity, schedule slip patterns, vendor delivery realism — exists in every EPC. It just isn't sitting next to the bid spreadsheet.
Capital Project AI sits where bid evaluation, schedule modeling, and resource planning meet. It uses your firm's historical performance as the calibration baseline and applies probabilistic models to the inputs that matter most: schedule slip, productivity variance, materials lead time, and rework. The output is a quantified bid recommendation, not a gut call.
How AI helps EPCs win the right work
Capital Project AI applies four engines specifically tuned to the EPC bid-and-execute cycle. Each one answers a question the bid team currently answers by feel.
Bid evaluation with portfolio context
Not every winnable job is worth winning. The platform evaluates each prospective bid against your active backlog: does this job add diversification or concentrate risk in one client, one geography, one critical resource? The output is a bid-or-pass recommendation with explicit reasoning.
Schedule realism check
The proposed schedule from the client gets stress-tested against your firm's historical productivity for similar scope. The platform flags activities where your actual hours-per-unit run materially higher than the bid assumption, and quantifies the schedule risk in days and dollars.
Resource bottleneck detection
Your engineering pool, fabrication shop, or specialty crane fleet is finite. Capital Project AI runs the bid through your active backlog to identify where the new job consumes constrained resources at problematic times. The recommendation includes an explicit "if you take this, defer that" trade-off.
Contingency that's defensible
Contingency stops being a number you defend in the bid review meeting and starts being a number that ties to specific risk drivers. P50 contingency, P80 contingency, and the cost of buying the remaining risk down — all quantified, all auditable.
Why Capital Project AI
- Calibrated to your firm. The models train on your historical projects, not generic industry averages. The forecasts reflect your actual productivity, cost, and slip patterns.
- Built around the bid review meeting. Outputs are the bid recommendation, the schedule risk summary, and the resource trade-off — not a back-office analytics dashboard.
- Fast enough for live bids. Run a new bid through the model in under a minute. Re-run with revised assumptions during the review meeting itself.
- Built by an ex-Shell capital owner. Founded by an engineer who has been on both sides of EPC contracts at $800M megaproject scale.
Run your next bid through proper risk math
Upload the bid pack and your active backlog — get a quantified bid recommendation with schedule and resource risk in minutes.
Open the Dashboard →What it looks like in practice
A mid-sized EPC is invited to bid on a $120M brownfield revamp. The bid pack assumes a 22-month schedule with standard productivity. The BD team is enthusiastic — the client is a strategic account and the geography is a current strength.
Capital Project AI runs the bid against the firm's last six brownfield revamps in similar scope. Three findings emerge. First, the schedule assumption for stainless welding is 18% optimistic — historical productivity puts the activity 6-8 weeks longer. Second, the engineering pool is already 85% loaded through Q3 of next year; this job consumes the remaining headroom and pushes a different active project off-schedule. Third, the proposed contingency at 12% covers the schedule risk at P50 but breaks at P75 — meaning a 1-in-4 chance of margin erosion. The recommendation: bid, but at $128M with revised contingency, and explicitly defer a small advisory engagement that competes for the same engineering capacity.
The same probabilistic engine powers probabilistic project planning on the owner side, and the underlying risk math is detailed in AI capital project risk management. For firms making the broader portfolio question — which mix of jobs to pursue — see project portfolio optimization with AI.
Frequently asked questions
How does AI help an EPC contractor specifically?
Three places: bid evaluation (which jobs are actually attractive after schedule risk and resource constraints), schedule realism (does the proposed schedule fit your historical productivity), and execution monitoring (which projects are about to consume disproportionate engineering or fabrication capacity).
Does this replace our project controls team?
No. It gives them a quantified second opinion. Project controls runs the schedule and earned value; Capital Project AI runs the probabilistic model that tells you which schedule risks will actually bite and which contingency cost is defensible.
Can we run this against historical projects to validate it?
Yes — and you should. Back-testing the model against your last 10 to 20 completed projects calibrates it to your firm's specific productivity, cost, and schedule patterns. The forecasts then reflect your actual performance, not industry-average data.