Structured project and evidence record
Maintains project entities, requirements, constraints, alternatives, supplier records, scenarios, decisions, review states, output versions and audit events.
AI and cloud architecture for traceable CEA project decisions.
The Crop Urbanis Intelligence Layer powers the Platform’s internal MVP. It combines document intelligence, structured project data, deterministic scenario services and qualified review so organisations can move from fragmented evidence to controlled project decisions. Selected private-beta deployments are considered by contact.
Architecture for controlled project decisions
The internal MVP connects project evidence, source-aware AI, deterministic services and qualified review into a controlled output path. Product hardening extends this architecture to selected organisational beta under agreed conditions.
How the internal MVP architecture works
01 · Intake
Project and evidence intake
Project brief, organisation and decision context, source documents, supplier files, agronomic references, constraints, assumptions, versions and permissions
02 · Documents
Document processing and normalisation
Document processing, extraction, classification, schema mapping, terminology and unit normalisation
03 · Data layer
Structured project data layer
Designed for project files, protocols, trial notes and validated agronomic references so teams can compare recommendations against source context.
04 · Evidence
Retrieval and source-aware evidence
AI information workflows, deterministic scenario services and explicit assumptions
05 · AI
AI information workflows
Qualified review, corrections, decision and approval state
06 · Scenarios
Deterministic scenario services
Controlled project, comparison, scenario, decision, delivery and handover outputs
07 · Review
Expert review and approval
Selected-beta evaluation and later governed operating-feedback roadmap Qualified review
08 · Outputs
Controlled outputs and delivery pack
Inputs and outputs designed for real CEA projects
The product is structured around the information teams already use during feasibility, design, commissioning, operation and research. The goal is reusable decision support, not generic chat output.
Input
Inputs the system is designed to structure
- Project brief, organisation and decision context
- Source documents and supplier files
- Agronomic references, requirements and constraints
- Site, utility, crop and market assumptions
- Alternative configurations and supplier options
- Units, capacities, CAPEX/OPEX and sensitivities
- Versions, permissions and review questions
- Permitted structured files, images and text
Output
Outputs for operators and project teams
- Structured project brief and evidence register
- Requirements and constraint model
- Source-aware comparison matrix
- Deterministic scenario record with explicit assumptions
- Risk, dependency and clarification log
- Review, approval and decision record
- Controlled delivery pack and handover inputs
- SOP or training drafts for qualified review
The Crop Urbanis Intelligence Layer is the technical core of the Platform’s operational internal MVP.
- Human-in-the-loop
- Internal MVP architecture
- EU-region deployment priority
- Expert-reviewed outputs
Current internal MVP components
Document processing and normalisation
Processes permitted PDFs, spreadsheets, text, images and structured files for extraction, classification, schema mapping, terminology and unit normalisation, source-location capture, missing-field detection and contradiction flags.
Scenario support keeps assumptions explicit so feasibility work can be reviewed by investors, operators and agronomy leads.
Retrieval and evidence layer
Connects AI outputs to permitted source material and project context through source-aware search, scoped knowledge access, citation retention, retrieval evaluation and separation of project-specific from reusable knowledge.
AI information workflows
Supports supplier-document extraction, technical terminology mapping, evidence retrieval, comparison assistance, missing-information and contradiction detection, controlled drafting, multilingual transformation and model evaluation.
Any future outputs must be treated as drafts: wording, crop assumptions and operational steps require checking before teams use them in training or production.
Deterministic scenario services
Handles units, capacity, crop cycles, production assumptions, utilities, compatibility, CAPEX/OPEX, sensitivities and project gates. Formulas and assumptions remain inspectable and versioned.
Dashboard logic focuses on operational indicators such as yield, quality, cycle length, labour, energy and recurring issue patterns.
Expert review and approval
Records reviewer, scope, comments, corrections, decision, approval state, unresolved issues and version. AI does not become the accountable decision owner.
Computer-vision and sensing work remains validation-led: use depends on data quality, crop context and clear acceptance criteria.
Controlled outputs
Assembles project briefs, evidence registers, requirements tables, comparison matrices, scenario records, risk and dependency logs, decision registers, delivery packs, SOP/training drafts and handover inputs.
Guardrails are being specified around measurable acceptance criteria, expert corrections and traceable feedback for future validation.
Product hardening and roadmap
The next stage covers authentication, organisation controls, tenant separation, workload isolation, monitoring, data lifecycle controls, deployment automation, cost optimisation and support operations; operating feedback, image/sensor analysis and advanced analytics remain roadmap work.
The next stage requires application services, APIs, document workers, background jobs, model evaluation, relational project data, object storage, document versioning, vector retrieval, queues, identity and access, tenant separation, monitoring, logs, audit events, backups, recovery, EU-region deployment and model routing with token and cost controls.
Business model
The model is built around recurring Platform access, private organisational deployment, fixed-scope onboarding and optional expert support.
- During private beta, access, support and commercial terms are defined individually; standard plans follow validation of initial organisational workflows.
Current internal MVP and beta deployment stage
The status separates operational internal-MVP components, selectively scoped beta work, product hardening and later roadmap capability.
-
01
Operational in internal MVP Operational in internal MVP
Structured project record, evidence/source register, document extraction pipeline, terminology/schema mapping, retrieval, AI-assisted comparison, deterministic scenario service, review and decision record, and controlled output generation.
-
02
Selective private beta Available within selected beta scope
Organisation-specific workspace, permissioned data, role and review configuration, scoped output package and evaluation against agreed criteria are considered under separate written terms.
-
03
In validation Product hardening
Authentication, tenant separation, workload isolation, monitoring, data lifecycle controls, deployment automation, cost optimisation and support operations are the next hardening priorities for selected-beta deployment.
-
04
Roadmap Roadmap
Later governed work includes integrations, operating feedback, image and sensor workflows, advanced analytics and cross-project learning under appropriate data conditions.
-
05
June 2026 Operational internal-MVP architecture
The Platform connects project records, evidence, document processing, AI information workflows, deterministic scenarios, review and controlled outputs.
-
06
Current status Selected private-beta deployment stage
Selected organisations may be considered for separately governed access after user, data, confidentiality, responsibility and support conditions are agreed.
-
07
Next validation track Sensing, image and diagnostic modules
Computer-vision, sensing integrations and automated alerts remain planned validation tracks tied to data quality and expert-review requirements.
Technical review routes
The technical record is split across product overview, MVP & Private Beta, security and the downloadable brief. Together they distinguish operational internal-MVP workflow, selected beta scope, product hardening and roadmap work.
Technical brief for cloud and AI partners
A downloadable brief summarises the internal MVP architecture, data flow, AI role, deterministic services, review controls, cloud workload, security boundary and private-beta deployment stage.
Selected organisations, cloud providers, AI partners and research teams can share high-level context. A controlled information route is agreed before confidential data or technical evidence is exchanged.