Product
One system of record for the land lifecycle
From the day a landowner asks “would you buy my property?” to work logged decades later — and responsible disposition when the mission changes.
The marriage of three things
CRM
Pipelines, approvals, and stakeholder tracking for acquisitions and disposition.
Field service
Work orders, crew scheduling, and task completion for active land management.
GIS
Every record is a place. The map is the primary interface — not a report bolted on later.
Everything else — reporting, dashboards, document generation — orbits around those three cores.
Lifecycle, not point solutions
Application → prospectus → close → manage forever → optional disposition → archive. Documents are outputs. The 3-year plan is a database object. Data cascades away from humans.
- Modules — acquisition, management, disposition
- Personas — day-in-the-life workflows
- Integrations — silos to system of record
From first inquiry through perpetual management — and responsible disposition when the mission no longer fits.
- 1
Landowner / strategic parcel
- 2
Acquisition
- 3
Closing
- 4
Management (perpetual)
- 5
Mission review
- 6
Disposition (if needed)
- 7
Archive with full history
What “good” looks like
Acquisition
“Zero time reformatting data into memos; zero time hunting for parcel boundaries.”
Management
“One map click: last 5 years of work, next 3 years of plan, contracts, and compliance.”
Disposition
“Full acquisition and management history preserved after divestment — institutional memory intact.”
AI assist: drafts, never decisions
Intelligence where re-keying costs staff the most time — governed by rules the product enforces, not a policy page.
Drafts, never decisions
AI output enters the system only through a human approval step, labeled as a draft until a person accepts it. State transitions, scoring rubrics, and finance postings never take AI input directly.
Reads and proposes; code decides
Extraction, summaries, and drafting are AI work. Routing, validation, and the deterministic document pipeline stay in code — a model proposes a query or a draft, the existing system answers it.
Nothing invented
Extractions are schema-constrained: a field the document does not state comes back empty, never guessed. Refusals and truncation surface as “no draft produced,” not a silently wrong record.
Provenance on every acceptance
Accepting a draft writes an append-only record: which task, which model, a content fingerprint, who approved it, and when.
What it drafts today
- Extract intake and case fields from deeds, appraisals, and title work into a review form
- Summarize a document beside the paper — never replacing it
- Draft board briefs; packets wait until every brief is reviewed
- Draft narrative fields, streaming into an editor the person owns
- Translate a typed search into structured tract filters, shown before they apply
- Sweep the received-document backlog through extraction, parking drafts for review
AI assist is a licensed enterprise integration with a provider choice — Anthropic, OpenAI, or Google — configured per deployment. Keys stay server-side; only the record under review is sent, never bulk data. Deployments that do not configure a provider see no AI affordance at all.
See where every record sits
Disposition cases and acquisition files carry a compact stage tracker driven by the record's actual server-side status — “referred to the board” is visible at a glance with its date, not something staff track in a side spreadsheet. Stages the system has not recorded are never invented: the tracker reports honest gaps rather than fabricated progress.
Open core. Commercial enterprise.
Open source
GeoTract Core
The public foundation any organization can run, inspect, and extend — independently useful on its own.
- Tract registry with the map-first application shell
- Work orders and completion tracking
- Three-year planning tied to tracts
- Global search with per-module filters
- Module SDK, identity (OIDC), and the design system
Commercial
GeoTract Enterprise
Licensed modules and adapters composed on top of the open core — the core never depends on them.
- Acquisition pipeline: intake, prospectus, ranking, memos
- Disposition and board portal with packet assembly and decisions
- Documents: versioned templates, generated records, inbound catalog
- Field operations with offline completion sync
- Reporting profiles, finance (AccuFund), GIS (ArcGIS), intake (CivicPlus), signatures (DocuSign)
- AI assist across the modules — always draft-with-review
Enterprise modules are entitled per deployment and consume the core as versioned packages. The dependency is one-way: the open core stays independently deployable, and an enterprise feature that is not licensed or not configured simply is not there — no upsell chrome inside the product.
Not just a map. Not just a CRM. Not just work orders.
GeoTract is the lifecycle system of record — acquisition through perpetual management to responsible disposition — with the map as the interface.
| Pattern organizations have today | What GeoTract is instead |
|---|---|
| Siloed GIS + desktop databases + folders | One system of record where the map, pipeline, and work orders share the same tract truth |
| Generic field service / work-order tools | Land lifecycle from acquisition through perpetual management and disposition — not tickets without place |
| Document-centric ECM as the source of truth | Structured data first; memos and contracts generated as outputs |
Design principles
- 01
The map is the interface, not a feature
Every list view has a map view, every record has a “show on map” button, every dashboard tile is filterable by drawing a polygon.
- 02
Data flows one direction: away from humans
A completed task does not require someone to also update the GIS layer, plan progress, and the annual report. The system does the cascade.
- 03
Documents are outputs, not inputs
Staff work in structured forms; memos and contracts are generated from that data. Nobody edits a Word doc as the source of truth.
- 04
Institutional memory is permanent
A tract divested years later still has its acquisition memo, treatment history, and disposition rationale — one query away, forever.
- 05
The multi-year plan is a database object, not a PDF
Progress against plan is a live metric, not something manually reconciled at year-end.