From a mapped site to a defensible, auditable project

A restoration project touches a dozen disciplines and produces a paper trail that a funder, a regulator, or an agency reviewer eventually has to trust. Orenna keeps that work in one model, where every number traces back to its source.

You start from a project's Area of Interest, the single mapped geometry that is also the project boundary. From there the platform pulls public data, turns ecological outcomes into modeled economics, tells you which regulatory frameworks apply, drafts the deliverables, and supports the pursuit work that sits on top. The mission behind all of it is to accelerate ecosystem restoration by giving practitioners faster, more defensible analysis.

Two principles hold throughout. Everything is grounded: claims cite their sources, and the assistant flags anything it could not ground. The human decides: the AI drafts, retrieves, and analyzes, but the engineer of record signs the work, and counsel signs the regulatory calls.

Create a project at New project: name it, name the sponsor, and upload the project boundary (GeoJSON, KML/KMZ, or a zipped shapefile) as the Area of Interest. County and state fill in from the geometry. The platform covers the contiguous United States.

The characterization run

A site characterization pulls roughly three dozen public datasets against the AOI: soils (SSURGO), land cover (NLCD), terrain (3DEP), hydrography (NHDPlus HR), watershed and flood-frequency (StreamStats), flood hazard (FEMA), wetlands (NWI), streamflow (NWIS), water quality (Water Quality Portal and ATTAINS), climate (PRISM), wildfire, species and critical habitat, dams, aquatic barriers, levees, water rights, contamination, cultural resources, and more. The contributing watershed auto-delineates from the AOI, and findings appear by category as the stages complete.

A synthesis pass produces a historical narrative, a triage of candidate intervention types, and an executive summary. These are assistive analysis built from the retrieved evidence, not autonomous design.

Map, Data Room, and downloads

The run feeds the interactive Map (the AOI, the delineated watershed, per-type layers, and live reference overlays) and auto-files its geospatial layers into a lifecycle-ordered Data Room. You can download a summary briefing, a full report, a Map Atlas, a geospatial ZIP (GeoJSON per data type, findings as JSON and CSV, and an ArcGIS GeoPackage), raster GeoTIFFs, the watershed, and the raw bytes of every dataset queried.

This is the original ROI workbench, now one area among several. It compares restoration alternatives by the dollar value they can actually realize.

Outcomes

On the Outcomes tab, quantify each ecological result the project delivers. For a standard outcome you set the quantity at full uplift, a ramp (linear, sigmoid, or step), and an attainment probability as a mean and a standard deviation. Habitat outcomes derive the change in weighted usable area from a baseline and target HSI. Water outcomes use a published VWBA method and can pre-fill their observable inputs from your latest characterization run. Stating attainment honestly up front (say "85% plus or minus 8%") is what makes the downstream NPV defensible.

Eligibility and the value claim

The Programs tab evaluates every funding program in the catalog against the project and sorts them into eligible, ambiguous, and ineligible, showing which rules passed and failed. For each eligible outcome-and-program pair the engine computes a value claim as quantity at full uplift times dollars-per-unit times attainment, then applies stacking rules that prevent two programs from paying twice for the same benefit. You can override eligibility or a rate with a written justification, which is recorded in the stacking ledger.

Scenarios and NPV

A scenario lays the value claims out as cashflows on the right schedule per program, adds the cost streams, and discounts to NPV, with IRR and payback alongside. Stewardship costs run across the stewardship horizon, or in perpetuity capitalized as a terminal value, so a scenario reflects the full long-term cost of keeping outcomes intact rather than just the cost of construction.

Uncertainty and total economic value

The single-point NPV is a screening estimate until you run Monte Carlo, which samples rates, attainment, and program persistence to produce confidence bands, a probability that NPV is above zero, and a tornado view of what drives the variance. A separate total economic value panel surfaces literature-based societal value as a transparency range, grouped by valuation method. It is never summed across methods and never folded into the realizable NPV, because the two answer different questions: what the project will be paid, and what public-good value it produces that no program reliably pays for.

Every project has a built-in assistant that answers from the project's own evidence. It retrieves from the project corpus (canon documents, metadata, outcomes, scenarios, decisions, and characterization findings) and cites each claim inline. Any clause it could not ground is flagged as unverified, so you can challenge it rather than trust it.

The assistant works as a project lead assistant with the full corpus and toolset across the major restoration-delivery activities: project management, site assessment, modeling and analysis, design, permitting, implementation, monitoring, and pre-award capture. It keeps the state of play coherent, surfaces cross-discipline implications, and never self-certifies. Permit and regulatory questions are answered from an encoded regulatory matrix, not from general knowledge.

The assistant only proposes memories, each with a citation; nothing is saved until you confirm it. The Decisions tab is your formal record of what was decided and why, indexed so the assistant can cite the reasoning behind a choice later.

The Checklist tab evaluates which regulatory frameworks apply to your project and surfaces the gaps: requirements that apply but are not yet satisfied. Answering a few intake context questions refines which frameworks apply.

The Reports tab generates the deliverable documents from a template catalog filtered to your project's state: the CEQA and NEPA series, state little-NEPA documents, permit and consultation packets, and pursuit deliverables. Each report builds section by section. Deterministic sections are pure functions over the project record and make no model calls; prompt sections run a grounded, cited Claude call. A section that fails grounding lands as needs review, and the way out is a human one: approve it as-is or replace the text. Finalizing seals the document into a signed, hashed manifest and locks it; to change a finalized report you create a new one.

For firms responding to public solicitations, the Pursuit workstream reads an RFQ or RFP and extracts the scope, the weighted evaluation criteria, the submission rules, and the individual requirements. From there you generate a draft Statement of Qualifications, a Technical and Cost Proposal, or an SF330, all grounded in the project record and your organization's Assets — past projects, personnel, certifications, and references.

A compliance matcher maps every requirement to where it is addressed and flags the gaps, and an RFP gets a labor fee estimate priced against your billing-rate schedule. Assets are org-scoped and kept off the project corpus for tenant isolation, so qualifications never leak between organizations, and the proposal generator is instructed never to invent people, projects, or credentials it cannot cite.

The platform is built to be reviewed. Chat answers cite their sources and flag what they cannot ground. Report sections record their citations and retrieval trace, and grounding-failed sections require human approval. Every chat turn produces a signed audit manifest, and finalized reports and checklists seal a signed, hashed manifest. Each project has an Activity tab: a field-level log of who changed what.

A few things to respect as you read the outputs. The single-point NPV is a screening estimate, not a guarantee. A Find Credits match is eligibility to evaluate, not confirmed inventory. And throughout, the AI assists your human practitioners. The engineer of record decides.