Product tour
See how governed agentic AI works
Walk through the platform screen by screen — no confidential data, fictional insurance example only. From configuring your agent network to retaining audit evidence, this is what enterprise teams get when they deploy OrqForge.
Built for enterprise operators, not experiments
OrqForge replaces slow, inconsistent human coordination with governed agent networks — while keeping humans accountable for material outcomes. These screens show the production platform using a demo workspace only.
Days
Typical workspace configuration target — not years
11 stages
Visible pipeline on every run
100%
Material outcomes pass human review gates
Any channel
Teams, email, webhooks, schedules
End-to-end flow
From configuration to audit-ready delivery
Insurance claims is the example scenario below — fictional data only. The same eight-stage pattern applies to any enterprise process you configure in OrqForge.
- 1 Configure
Design your agent network in days
Skip multi-year integration projects. Business and IT co-design the operating model together.
The agentic flow builder proposes workspaces, roles, taxonomy, and review paths from your process material. Insurance claims is our walkthrough example — the same builder configures onboarding, approvals, support triage, or any enterprise workflow.
app.orqforge.com/workspaces/demo-insurance/ai-trainerAI Trainer 19 messagesI've reviewed your process docs. For this insurance example I'm proposing 6 roles across intake, evaluation, adjustment, and publishing. Enable orchestrator + FNOL + damage eval for the demo?Yes — enable those three plus adjuster. Keep fraud detection disabled for now.Done. Taxonomy updated: claim_intake, damage_assessment, adjustment_report. Review gates set to draft mode.Attach documents or describe changes…Send - 2 Organise
Define hierarchy and accountability
Chain of command, not a single prompt. Every agent has scoped tools, owners, and review paths.
Orchestrators delegate to specialists. Stream owners assign human reviewers. Data boundaries stay inside the workspace — no unbounded agent access.
app.orqforge.com/workspaces/demo-insurance/org/organisationLiveAgent hierarchy
Insurance claims (demo) · 6 roles · 3 tiers
1Process Orchestrator
Stream owner · Ops Lead (human)
Draft execution
-
Intake Agent
Teams · Email · Webhook
-
Classification Agent
Taxonomy routing
2Delivery Orchestrator
Parallel specialist coordination
-
Damage Evaluation
Scoped tools · Evidence pack
-
Adjuster Agent
Structured output · Review path
-
FNOL Agent
First notice · Complete
3Review Coordinator
Human accountability gate
-
Stream reviewers
2 assigned · Draft mode
-
- 3 Trigger
Intake from every channel your teams use
One platform for Teams, email, webhooks, and schedules — consistent governance regardless of entry point.
Rule-based routing classifies imperfect input: subject lines, senders, attachments, or chat context. Work enters the same eleven-stage pipeline every time.
app.orqforge.com/workspaces/demo-insurance/triggers/triggersMulti-channel intake
Meet teams where they work — same governed pipeline for every trigger
TMicrosoft Teams
ActiveBot installed · 3 workspaces
@Email listener
Activeclaims-inbox@demo.example · Rule-based routing
WWebhook
ActiveAzure DevOps · Jira · Custom
⏱Scheduled jobs
ConfiguredDaily heartbeat · Weekly reports
Enterprise pattern
Imperfect input from any channel → classified → orchestrated → reviewed → published to your systems.
- 4 Execute
Track every run through the pipeline
Operations leaders see stage progress, elapsed time, and bottlenecks — not black-box AI.
Eleven visible stages from lobby to learn. Active runs show which agents are working, which stage they are in, and how long work has been in flight.
app.orqforge.com/workflows/ai-workflowsExecuteMotor claim intake — CLM-2026-0847
claim_intake · Demo Insurance POC · Active · 12m 34s
LobbyIngestTriagePlanHierarchyExecutePost-ExecReviewPublishLearnDoneDamage assessment report
Execute · 8m 12s · 24.1k tokens
Claims orchestrator
Execute · Delegating · 12m 34s
- 5 Collaborate
Operators work alongside agents
Transparency builds trust. Teams see reasoning, delegation, and status without switching tools.
Chat with role agents, attach documents, and trigger work from the interface your operators already use. Delegation events appear inline with full lineage.
app.orqforge.com/chat/agent-chatCOClaims Orchestrator
● Working
New motor claim — rear-end collision, 2 photos attached. Policy holder reports vehicle not drivable. Please process.Received. Classifying as claim_intake. I'll delegate damage evaluation and adjuster review. Estimated completion ~15 minutes.Delegated
→ Damage Evaluation Agent (working)
→ Adjuster Agent (queued)
Message Claims Orchestrator… - 6 Delegate
Specialists execute in parallel
Scale throughput without scaling headcount — coordinated specialists, not one overloaded assistant.
When work exceeds one agent’s scope, orchestrators delegate across the network. Reasoning streams stay visible per agent so supervisors can intervene early.
app.orqforge.com/runs/clm-2026-0847/mission-treeAgent network
- Claims Orchestrator Monitoring
- Damage Evaluation Working
- Adjuster Agent Working
- FNOL Agent Complete
- Fraud Detection Disabled
Reasoning — Damage Evaluation
Analysing attached images for bumper and panel damage…
Estimating repair bracket: £2,400 – £3,800 based on visible impact zone.
Missing: garage estimate, policy excess confirmation.
→ Passing partial assessment to Adjuster Agent.
- 7 Govern
Humans approve material outcomes
Defensible to audit and procurement. Draft, assist, and auto modes configured per workflow type.
Review gates map to role labels and stream owners. Approvers see recommended actions, contributors, confidence, and attachments before anything publishes.
app.orqforge.com/runs/clm-2026-0847/review/reviewPending approval
Damage assessment report
CLM-2026-0847 · Draft mode · Awaiting stream owner
Recommended action
Approve structured assessment and route to claims management system.
Confidence: High · 3 evidence attachments · 2 agent contributors
Approve & publishRequest changesReview context
-
Execution mode
Draft — human gate required
-
Reviewer
Ops Lead (assigned)
-
Contributors
Orchestrator, Damage Eval, Adjuster
-
Stage history
11 stages · 12m 34s elapsed
Material outcomes never publish without explicit human approval — configurable per workflow type.
-
- 8 Assure
Evidence retained on every material run
Answer “what happened, who approved it, and why” — without reconstructing from chat logs.
Stage history, input snapshots, delegation lineage, and review decisions export for audit. Learning proposals stay governed — nothing silently changes agent behaviour.
app.orqforge.com/runs/clm-2026-0847/evidence/audit-trailPDF · JSONRetained evidence — CLM-2026-0847
Immutable stage log · Exportable for audit & procurement
09:14:02Ingest Intake AgentEmail + 2 attachments indexed
09:14:18Triage ClassificationRouted to claim_intake · Priority: standard
09:15:41Hierarchy OrchestratorDelegated to Damage Eval + Adjuster
09:22:07Execute Damage EvaluationAssessment draft · £2,400–£3,800 bracket
09:26:33Review SystemQueued for Ops Lead approval
Input snapshots
3 retained
Delegation lineage
2 hops
Review decisions
Pending
Product
All platform screens
Every view below uses fictional demo data. The same interfaces power claims, onboarding, support, approvals, and any workflow you define.
Configure workspaces, roles, taxonomy, and review paths in a guided session — days, not years.
Agent hierarchy
Insurance claims (demo) · 6 roles · 3 tiers
Process Orchestrator
Stream owner · Ops Lead (human)
Draft execution
-
Intake Agent
Teams · Email · Webhook
-
Classification Agent
Taxonomy routing
Delivery Orchestrator
Parallel specialist coordination
-
Damage Evaluation
Scoped tools · Evidence pack
-
Adjuster Agent
Structured output · Review path
-
FNOL Agent
First notice · Complete
Review Coordinator
Human accountability gate
-
Stream reviewers
2 assigned · Draft mode
Mirror your operating model: orchestrators, specialists, scoped tools, and human accountability gates.
Multi-channel intake
Meet teams where they work — same governed pipeline for every trigger
Microsoft Teams
ActiveBot installed · 3 workspaces
Email listener
Activeclaims-inbox@demo.example · Rule-based routing
Webhook
ActiveAzure DevOps · Jira · Custom
Scheduled jobs
ConfiguredDaily heartbeat · Weekly reports
Enterprise pattern
Imperfect input from any channel → classified → orchestrated → reviewed → published to your systems.
Teams, email, webhooks, and schedules — every intake path feeds the same governed pipeline.
Motor claim intake — CLM-2026-0847
claim_intake · Demo Insurance POC · Active · 12m 34s
Damage assessment report
Execute · 8m 12s · 24.1k tokens
Claims orchestrator
Execute · Delegating · 12m 34s
Live pipeline stages, elapsed time, token usage, and status on every run — inspectable end to end.
Claims Orchestrator
● Working
Delegated
→ Damage Evaluation Agent (working)
→ Adjuster Agent (queued)
Talk to role agents directly. Attach documents, trigger work, and follow reasoning in real time.
Agent network
- Claims Orchestrator Monitoring
- Damage Evaluation Working
- Adjuster Agent Working
- FNOL Agent Complete
- Fraud Detection Disabled
Reasoning — Damage Evaluation
Analysing attached images for bumper and panel damage…
Estimating repair bracket: £2,400 – £3,800 based on visible impact zone.
Missing: garage estimate, policy excess confirmation.
→ Passing partial assessment to Adjuster Agent.
Orchestrators coordinate specialists in parallel — with full visibility into who is doing what.
Pending approval
Damage assessment report
CLM-2026-0847 · Draft mode · Awaiting stream owner
Recommended action
Approve structured assessment and route to claims management system.
Confidence: High · 3 evidence attachments · 2 agent contributors
Review context
-
Execution mode
Draft — human gate required
-
Reviewer
Ops Lead (assigned)
-
Contributors
Orchestrator, Damage Eval, Adjuster
-
Stage history
11 stages · 12m 34s elapsed
Material outcomes never publish without explicit human approval — configurable per workflow type.
Material outcomes wait for explicit human approval. Execution mode, reviewers, and context on one screen.
Retained evidence — CLM-2026-0847
Immutable stage log · Exportable for audit & procurement
Email + 2 attachments indexed
Routed to claim_intake · Priority: standard
Delegated to Damage Eval + Adjuster
Assessment draft · £2,400–£3,800 bracket
Queued for Ops Lead approval
Input snapshots
3 retained
Delegation lineage
2 hops
Review decisions
Pending
Stage history, delegation lineage, and input snapshots — retained for audit, procurement, and assurance.
Ready for your processes?
We will walk you through a live session configured around your roles, review paths, and integration requirements — not a generic slide deck.
- Map your operating model to workspaces and agent hierarchy
- Define execution modes and review gates per workflow type
- Connect Teams, email, or webhooks to your first agentic flow
- Discuss procurement, audit, and deployment with your stakeholders
Book a tailored walkthrough
Tell us your use case — we'll show OrqForge configured for your enterprise, not ours.