The decision layer above your planning stack
Where enterprise data becomes investment decisions
Primavera schedules the work. ERP records the spend. Capvium sits above both — turning planning, execution, market, and risk data into sized, stress-tested recommendations for the board, capital committee, and executive team.
Governance
Recommendations: fund · delay · accelerate · resize · stop
Capvium — AI capital decision layer
Schedules · budgets · actuals · resources · dependencies · scenarios
Planning and execution systems
Every claim on your capex, sized and stress-tested in one portfolio view.
Total Capex
$121.5M
Blended Return
+20%
Blended Downside
-9%
Stress Scenario
| Claim | Vertical | Budget | Exp. Return | Downside | Decision |
|---|---|---|---|---|---|
| Annual Capex Baseline | Oil & Gas | $42.0M | +18% | -6% | Proceed |
| Maintenance vs Growth Allocation | Infrastructure | $28.5M | +13% | -9% | Stage |
| Fleet Renewal Replacement | Oil & Gas | $16.2M | +22% | -13% | Hedge |
| Energy Transition Transformation | Data Centers | $34.8M | +26% | -10% | Stage |
The step every capital process skips
Walk through any capital approval process and you'll find the same shape: business cases are built one at a time, each with its own optimistic base case; they're ranked by IRR or NPV; and the budget is handed down the list until it runs out. Three things are missing from that picture.
Sizing
A ranking says project A beats project B. It cannot say whether A deserves $40M or $90M — and the difference between those two numbers is often the difference between a resilient plan and a fragile one.
The portfolio view
Each business case is evaluated as if it lives alone. But your projects share suppliers, labor pools, regulators, and commodity exposure. When one goes wrong, its siblings tend to go wrong with it.
The bad year
Approval documents are built on base cases. Almost none state the probability that the plan produces a loss the company can't comfortably absorb — the one number that should govern how much you commit.
| Project | IRR | NPV | Funding |
|---|---|---|---|
| Project A | 18.2% | $210M | $90MSized how? |
| Project BThese two share a supplier ↗ | 16.7% | $180M | $90M |
| Project C | 14.9%Base case only | $140M | $90M |
| Project D | 11.3% | $95M | $90M |
One decision, three moments
Moment 01
Before you commit
See every option as a distribution, not a promise
Upload the schedule and the business case, and every option becomes a probabilistic picture: simulated completion dates and cost outcomes (Monte Carlo simulation over the real schedule logic, not a rule of thumb), the activities most likely to blow the date, and the resource bottlenecks that quietly serialize “parallel” work. Single-point estimates go in; honest ranges come out.
Completion date
fan chart The plan says June.
Simulation
says October.
Moment 02
When you commit
Size every commitment against the whole book
This is the heart of the platform. Give it the full set of options and the real budget, and the engine recommends a dollar commitment for each — including zero for options that shouldn't make the cut. Recommendations account for how options move together, are stress-tested across base, downside, and upside scenarios, and are explicitly guarded against over-commitment: the plan is sized so that one bad year doesn't take the portfolio down. And because no committee should trust a black box, every recommendation ships with the conventional alternatives — ranked NPV, ranked IRR, equal split — computed on the same data, so the room can see precisely where the engine disagrees with the default playbook and interrogate why.
Budget line
Moment 03
After you commit
Track the plan you actually sanctioned
Sanction isn't the finish line — it's the baseline. As execution data comes in, the platform scores reality against the sanctioned plan and tells you what to do about it in words, not dashboards: Proceed, Watch, or Do not allocate further. Portfolio-level verdicts flag where you've drifted Material Over or Material Under the optimal book, so reallocation is a standing agenda item instead of an annual scramble.
| Project | Drift | Verdict |
|---|---|---|
| Delaware P3 | On plan | Optimal |
| Eagle Ford | -14% vs. sanctioned | Material Under |
| Midland Loop | +9% vs. sanctioned | Do not allocate |
“Eagle Ford is tracking 14% under its sanctioned commitment as supplier delays push completion right — reallocating the difference to Delaware P3 keeps the portfolio inside its stress-tested loss band.”
Built to be interrogated
A capital recommendation is only useful if it survives the room. Four design choices make ours hold up.
Scorecard.pdf
Every input is visible.
Success probabilities, durations, downside severity, upside — the assumptions behind every option are on screen, editable, and logged. Change one and watch the recommendation move. No hidden weights.
Every recommendation has a control group.
The conventional methods run alongside the engine on identical data, every time. If we disagree with rank-by-IRR, you see the disagreement, its size, and its reason.
Every claim is backtestable.
Where realized outcomes exist, the platform scores its own history — recommended commitments against what actually happened, with “actual” defined purely from realized cash flows. We keep score in public.
Every run leaves a trail.
Inputs, assumptions, scenario settings, outputs and verdicts are logged per run and exportable as a committee-ready scorecard. “Why did we fund this?” has a documented answer.
No integration project required
the files you already have — Primavera and MS Project schedules, Excel budgets and options lists. Useful output on day one; deeper integrations when you're ready, not as a precondition.
your existing project controls and ERP stay the systems of record for execution and spend. We read from them; we don't replace them.
committee-ready scorecards (PDF), portfolio exports, and a live dashboard for the standing review.
Enterprise-grade by default
Data is encrypted in transit and at rest; access is role-gated; analysis runs are isolated per customer.
Security pageBring us one real decision
The fastest evaluation isn't a feature tour — it's your portfolio, our engine, one afternoon. Bring a decision you have to make this quarter and we'll run it together.