One connected picture of your capital project, kept honest by AI.
Contarva is an AI-powered project controls platform for owners of major capital projects.
It connects your EPC contractor's reporting with your own cost, commitment and schedule systems. Its agents are designed to flag where the two disagree, with the evidence, so your team reviews exceptions instead of rebuilding the workbook.
Seeking owner-side design partners on major capital projects, starting in energy. Built on FinVorx semantic layers.
What changes when your project controls are connected.
The owner carries the total cost, but the total exists in no single system: the contractor holds one half and your systems hold the other. Connecting them is what lets AI do useful work, because agents can only find disagreements between two views once both sit in one governed layer.
One reconciled picture.
EPC-reported and owner-held cost, commitments and schedule sit in one governed layer, each figure tagged with who reported it and at what level of detail. Total exposure is shown with a confidence level.
Agents that watch the seams.
The costly surprises sit where the two views meet: owner-furnished items, pending changes, accrual cut-off timing. Agents are designed to watch those points and raise a flag with the evidence behind it.
Judgement over rebuilding.
The monthly effort moves from stitching the workbook to reviewing exceptions and testing the forecast. People decide.
| Role on the owner's team | Today | Connected with Contarva |
|---|---|---|
| Project director | Waits for the monthly pack and asks which version of the number is right. | One reconciled view of total exposure, with the open disagreements listed. |
| Project controls lead | Rebuilds the EPC-plus-owner workbook every month. | Reviews flagged exceptions, with the evidence attached. |
| Cost and commercial | Chases pending changes and owner-furnished items by email and spreadsheet. | Sees the items agents flag as unreconciled, by cost code. |
| Finance and accruals | Discovers cut-off timing differences at close. | Sees cut-off and commitment differences as they appear. |
| Executive sponsor and JV partners | Reads a report assembled after the fact. | Sees one snapshot of exposure with confidence shown, and what sits behind it. |
What EPC detail is available depends on the contract type and reporting rights; the product is designed to make those limits visible, not to remove them.
An AI platform that connects, maps, validates and summarizes.
Four steps, from two separate views of your project to one snapshot your team can act on. The concept video shows the idea on a fictional project.
Two views, left as they are.
Your EPC contractor's reporting and your own cost, commitment and schedule systems come in through agreed imports. Nothing is replaced and nothing is rekeyed.
AI maps both views to one structure.
AI is designed to propose how EPC cost codes map to your cost codes, with a confidence score on each match. Low-confidence matches go to your team to confirm.
AI validates the seams.
Designed to check EPC and owner data against each other and flag where they disagree, such as owner-furnished items, pending changes and cut-off timing, with the evidence behind each flag.
AI summarizes it for your team.
A drafted summary of EPC plus owner cost and schedule exposure, with confidence shown and each statement linked to its source records. Your team reviews and decides.
Read a text description of the video
0:00–0:05 The text “Your project’s cost lives in two places. Your decisions need one view.” appears while six sources drift across the screen: EPC progress report, EPC invoices and changes, owner commitments, owner cost ledger, P6 schedules and Excel workbooks. Then: “Contarva uses AI to connect, map, validate and summarize it.”
0:05–0:12 · Connect. “Two views of one project. Keep both as they are.” Two EPC sources and two owner systems feed separate raw landing layers (actuals, estimates, schedule, procurement), each tagged by who reported it and at what level of detail. Agreed imports; nothing replaced, nothing rekeyed.
0:12–0:21 · Map. “AI maps both views to one cost structure.” EPC cost codes are matched to owner cost codes with confidence scores of 97%, 95% and 91%. One low-confidence match, 71%, is flagged for human review and then confirmed by the Project Controls Lead. A counter shows 412 cost accounts mapped and 9 sent for review.
0:21–0:31 · Validate. “AI validates the seams between them.” Checks run across both views for totals, accrual cut-off timing, duplicate invoices and unmapped cost codes: 1,846 checks run and 4 flagged for human review, each with its evidence. The flags are EPC piping hours running 14% over earned value; 7 pending EPC changes worth $12.4M missing from the owner forecast; late compressor drawings threatening start-up by 11 days; and owner-furnished spools slipped six weeks. “Agents show their evidence. People decide.”
0:31–0:39 · Summarize. “AI summarizes it for your team to decide.” Estimate at completion of $2.740B; start-up forecast of 18 May 2027, 11 days later than baseline; change exposure of $86M; moderate forecast confidence; and three items requiring management attention. An AI-drafted summary, marked “draft for review”, cites its sources and is then marked as reviewed by the Project Controls Lead.
0:39–0:45 The Contarva logo, “Project controls. Rebuilt.”, “AI-powered project controls for owners.”, “Connect. Map. Validate. Summarize. People decide.” and “Powered by FinVorx semantic layers.”
Concept visualization of the intended product, using dummy anonymized data for a fictional project. Continuous monitoring, refresh frequency, direct connectors and any write-back are subject to validation and agreed pilot scope; the proposed starting point is controlled file imports, with SAP, Oracle and Primavera P6 remaining the systems of record. What EPC detail is available depends on the contract type and reporting rights; the product is designed to make those limits visible, not to remove them. Agents surface findings for human review—they do not approve or change forecasts. AI-drafted summaries are drafts for human review.
AI that shows its evidence and leaves the decisions to your team.
Evidence with every flag.
Agents show what they found and the records behind it.
People decide.
Agents surface findings for human review. They do not approve or change forecasts.
Your systems stay the record.
SAP, Oracle and Primavera P6 remain the systems of record. The starting point is controlled file imports.
Your data stays yours.
Named, limited access in an agreed project-specific environment, and no cross-customer or shared-model training without your explicit agreement.
Storage location, AI provider handling, retention and deletion terms are confirmed before any data is shared. The full data-handling proposal is below.
Help us build the owner's view.
The agents are only as useful as the process they are configured around. We're looking for owner organizations on major capital projects willing to map a real monthly controls process and test a focused prototype on an agreed set of historical or current project data. Sanitized samples are welcome; a complete dataset is not required to start. What EPC detail we can use depends on your contract and reporting rights, so we scope that together at the start.
Map the process
Founder-led mapping of how EPC reporting and owner costs are combined today: monthly close, handoffs, approvals and recurring bottlenecks.
Configure a prototype
A focused prototype built around your agreed codes, data, EPC reporting and workflows.
Give Contarva Month 6
Can the prototype surface an exposure your reporting only recognized in Months 7–9? We test that against historical records and check false positives too. Earlier detection is a hypothesis to validate, not a promised outcome.
Owners carry the total cost of a major project but rarely see it in one place: the EPC contractor holds one view, the owner's systems hold another, and spreadsheets stitch them together. Replacing that stitching means understanding the controls process in detail.
Contarva is founder-led, by someone with more than 25 years in international capital project finance, including LNG. It starts from the owner's problem: cost-code mapping, accrual cut-offs, commitment exposure and forecast assumptions.
Senior project controls specialists, with more than 50 years of major capital project experience between them, advise on different areas of the product.
Less time rebuilding the workbook.
More time testing the forecast.
If you own a major capital project and combine EPC reporting with your own cost data, we'd like to hear how you do it today. Start with your workflow and the problem you want to solve.
Please don't email project files or EPC contractor data. We'll agree permitted use and a handling plan before any data is shared.