Features
Built to replace complexity with clarity
Visual entity design, AI-powered automation, and native GOV.UK compliance, everything you need in one configurable platform.
Entity Designer
Model any operational process. Visually.
Define custom entities, schemas, statuses, and relationships through an intuitive admin interface. Without writing a single line of code.
- Custom fields: text, number, date, select, relation, file, rich text
- Status-driven workflows with guard functions for validation
- Relationship mapping between entities with visual selectors
- Full audit trail on every record change
Contact
Workflow Settings
Define lifecycle states and allowed transitions
Active States
State Transitions
Workflow Diagram
Visual Workflow Builder
Automate routing, notifications, and escalations.
A visual, node-based workflow engine that handles conditional branching, automated notifications, SLA tracking, and webhook integrations.
- Drag-and-drop visual workflow editor
- Conditional routing based on record data or status
- Automated notifications via GOV.UK Notify
- SLA enforcement with escalation rules
Visual Page Builder
70+ components. Zero code required.
Build operational dashboards, forms, record views, and portals using the Puck visual editor. Bind components to your entity data and publish instantly.
AI Chat
Infinitely expandable models. Zero code.
Tell the platform what you need in plain English. AI Chat analyses your request, generates a structured plan with the exact changes, and waits for your approval before applying anything. Full rollback on every change.
- Describe changes in plain English: add fields, reshape models, update menus
- AI generates a plan with full JSON diff for review before applying
- One-click approve or reject with instant rollback capability
- Complete audit trail of every model change, who requested it, and when
Move Operations & Maintenance to the top of our sections.
I can see the "Operations & Maintenance" section is currently the 4th section. I'll move it to the top, right after Dashboard. Here's my plan:
This plan moves the Operations & Maintenance section to be the 3rd item in the sidebar (right after Dashboard).
Would you like me to apply this change?
2 tool calls
Plans
11 pendingDefault Sidebar
{
"data": {
"root": {
"props": {
"title": "Default Sidebar"
AI & Automation
AI that's constrained by design. Not bolted on.
Most platforms let users talk to AI and hope for the best. Jarsis Platform takes a different approach: every AI capability is governed by an ai_definition rule that controls context in, validates output, and writes results to auditable data tables. AI suggests, classifies, and completes. It never freelances.
The Classification Engine
A 10-step pipeline that gets smarter with every correction.
When a request comes in, the classification engine extracts individual issues from free text, generates vector embeddings, narrows candidates by the submitter's location and organisation, then asks the LLM to classify against only the shortlisted options. The result: high-accuracy classification at a fraction of the token cost, with a self-improving feedback loop built in.
Sarah Mitchell
Building A, Floor 3
James Okafor
Building A, Floor 1
Building A scoped · 2 requests · filtering irrelevant service classes
Heating
Confidence: 97%
The radiator in room 4B is not producing any heat ...
Electrical
Confidence: 94%
Two of the ceiling lights in corridor C are flicke...
Heating
Sarah Mitchell97% confidence · Auto-routed
Electrical
James Okafor94% confidence · Auto-routed
~300
Tokens each
95%+
Accuracy
$0.002
Per request
Token cost per request
~3,500 tokens
~90%
Token reduction
~300 tokens per classification vs ~3,500 with a naive approach. Context-aware narrowing means the LLM only sees shortlisted candidates, never your entire service catalogue.
Self-improving
Feedback loop
When a human corrects a classification, that correction is stored and fed back as a few-shot example. The system gets more accurate with every correction, without retraining.
Cache-first
Architecture
Results are cached by vector similarity (98% cosine threshold, 7-day TTL). Repeat and near-duplicate requests return instantly at zero token cost.
Context-aware
Candidate narrowing
A source_context_filter resolves the submitter's building, floor, and organisation to dynamically scope which service classes are even considered. Irrelevant categories never reach the model.
Dual-mode operation
Real-time suggestions as users type via classifyText(), plus background async batch processing via runClassificationPipeline(). Same engine, two speeds.
Provider-agnostic with credential isolation
Supports OpenAI, Anthropic, and AWS Bedrock with per-request credential overrides. Switch providers without changing a line of classification logic.
AI Definitions
Every AI action is governed by a rule you control.
AI Definition: Triage Service Request
Configure AI tasks, prompts, and model settings
How an AI Definition works
1
Define inputs
Select which entity fields are sent as context to the model
2
Set the prompt
Write a system prompt with the task, constraints, and expected format
3
Validate output
Define output fields and schema. Malformed responses are rejected
4
Write to data
Valid output is written to auditable entity records, never surfaced raw
AI Definition Rules
Every AI capability is governed by an ai_definition: a rule that controls exactly what context is sent to the model, which input fields are included, and how the output is validated before it touches your data.
Constrained, Not Conversational
Users never talk to AI directly. Instead, AI runs behind your components to suggest, auto-complete, and classify via transactional, auditable data tables. No open-ended chat risk. No hallucination in your workflows.
Controlled Inputs & Validated Outputs
Each definition specifies which entity fields feed into the prompt and which output fields the model must return. Output is validated against your schema before being written. Malformed responses are rejected automatically.
Document Intelligence
AI ingests documents, extracts structured metadata, redacts sensitive content, and stores searchable versions. All governed by the same definition rules with full audit trails.
RAG-Powered Search
Natural language search across all documents and records using retrieval-augmented generation, constrained to your organisation's data and access policies.
Full Transparency & ATRS
Every AI invocation is logged: input sent, output received, model used, cost, latency, and review status. ATRS records are auto-generated from these audit trails for UK government compliance.
Powered by AWS Bedrock in EU-WEST-2
All AI processing runs within UK AWS infrastructure. Your data never leaves your governance boundary, reducing your risk profile and simplifying compliance.
Report a Problem
Tell us what's wrong and we'll get it sorted.
Building
Floor
Self-Service Portal
Your people raise requests. AI does the routing.
Give staff, tenants, or customers a clean branded portal to submit requests directly. Each submission is scoped to their building and organisation, auto-classified by AI, and routed to the right team without manual triage.
- Branded request portal scoped to the user's location and organisation
- AI auto-classifies by service class and category at submission
- Follow-up questions configured per service type for richer triage
- Instant confirmation with reference number via GOV.UK Notify
Data Ingestion
Bring your data in. From anywhere.
Built-in pipeline builders for importing data from legacy systems, spreadsheets, and third-party APIs. Map columns, resolve dependencies, and track every import with full error detail.
- CSV and Excel upload with column mapping and preview
- API webhooks: accept authenticated POST requests from external systems
- SFTP sync: auto-process file drops on a schedule
- Ingestion Plans: orchestrate multi-source syncs with dependency resolution
Data Ingestion
Import data via CSV upload, API webhooks, or SFTP file drops
CSV Import
Upload CSV or Excel files, map columns, and bulk import into entities.
3 configurations
API Webhooks
Accept data from external systems via authenticated POST requests.
5 configurations
SFTP Sync
Auto-process file drops from SFTP watch folders on a schedule.
2 configurations
Ingestion Plans
Multi-source orchestration: group endpoints, resolve dependencies, sync.
1 configuration
Outbound Integrations
Outbound Integrations
Sync changes to external systems. In real time.
When data changes inside Jarsis Platform, push those updates to your existing systems automatically. Configure targets, define sync rules per entity, and monitor every outbound request in the sync log.
- Configure webhook targets with OAuth2, API key, or basic auth
- Sync rules define which entity changes trigger outbound pushes
- Full sync log with timestamps, HTTP status, response time, and retry count
- Test connections before going live, pause and resume at any time
Compliance
GOV.UK-ready by design.
The only low-code platform in the UK market with native ATRS, GOV.UK One Login, GOV.UK Notify, and WCAG 2.2 AA as integrated capabilities.
WCAG 2.2 AA
Accessibility enforced at the component level. Every element in the 70+ component library meets WCAG 2.2 AA standards.
ATRS Auto-Generation
AI transparency records auto-generated from audit trails, meeting the UK government mandate effective May 2025.
GOV.UK One Login
Native OIDC integration with GOV.UK One Login for secure, standards-compliant user authentication.
GOV.UK Notify
Send transactional emails and SMS through the GOV.UK Notify service, the approved channel for government communications.
Data Sovereignty
Your data stays in the UK. Full stop.
AWS EU-WEST-2
Entire platform hosted within UK AWS infrastructure. No data egress to third-party providers.
Bedrock AI
AI processing via AWS Bedrock. Your data stays within your governance boundary, reducing risk profile.
See it in action.
Book a personalised 30-minute demo and see how Jarsis Platform fits your operations.