One common
platform.
Configured to
your organisation.
The One Arc Platform is the common governed foundation behind every AI Colleague. It works across the HRIS, ERP, ITSM and collaboration systems you already run.
One governed engine.
Every AI Colleague.
AI Colleague Intelligence understands the request and coordinates the work. The Governed Engine controls execution on the One Arc Platform: from the next permitted action to a verified outcome, with human authority at the points that matter.
Understand the request, the person, the channel and the context.
Use approved knowledge and operating rules to choose the next step. Obtain required human approval.
Human decision before actionCarry out permitted actions through governed workflows and permission-aware integrations.
Check that the action happened and the completion conditions are met. Preserve the evidence.
Ask for missing information or retry within defined limits. Escalate when a boundary is reached.
Close only when completion is verified. If blocked, unsafe or needing a decision, pause for a person.
DINA starts with HR Operations. Every future AI Colleague shares the same governed foundation: Control Hubs guide the work, the Trust Layer enforces Privacy, Accuracy and Security, and Ledgers retain the evidence.
Every layer has
a clear responsibility.
From approved knowledge to a verified outcome, the boundaries are part of the product.
Governed execution, from start to finish.
The AI Colleague understands the request and coordinates the work. The Governed Engine controls the sequence and enforces the boundaries.
AI Colleague Intelligence
Context and coordination across the Colleague’s modules.
Governed Engine
Classify → retrieve → draft → validate → gate → execute → trace.
Your operating model becomes the rulebook.
Control Hubs define the work, the knowledge it can use and the conditions it must meet.
Process Hub
SOPs, roles, operating rules and escalation paths.
Workflow Hub
Controlled states, handoffs, approvals and completion conditions.
Knowledge Hub
Approved sources, access scopes and knowledge gaps.
Human authority is part of the architecture.
The Trust Layer implements PAS: Privacy, Accuracy and Security. Consequential actions pause at an authorised person.
Human Decision Gate
AI cannot approve its own consequential actions.
Six Hard Blocks
Fixed classes of action that cannot run autonomously end to end.
Separate ledgers
Case lifecycle, human decisions and audit evidence.
Work across the systems you already run.
Permission-aware adapters connect approved channels and systems of record, with receipts that show what happened.
Approved channels
Console, email, Teams, WhatsApp and secure journeys, subject to configured identity and workflow contracts.
Integration Layer
Scoped actions, explicit evidence levels and verified outcomes.
Knowledge with context. Answers with sources.
Shared knowledge and document capabilities support the AI Colleague without creating a separate knowledge product.
L1 / L2 / L3 knowledge
Global, jurisdiction and customer knowledge, scoped by tenant, role and purpose.
Document Intelligence
Receive, extract, verify and organise documents with recorded provenance.
Learning Loop
Unknowns become knowledge gaps; reusable learning needs human review.
One common platform. Your isolated environment.
One Arc Labs operates the shared product. Your organisation’s variation lives in governed configuration.
Customer Pack
Policies, SOPs, roles, approvals, SLAs, channels and approved knowledge.
Tenant foundation
Your users, legal entities, records, documents and decisions remain tenant-scoped.
Controlled releases
Governed configuration and controlled releases keep your operating rules consistent.
Your organisation.
Your rules.
One shared product.
Your organisation has an isolated tenant and workspace. Policies, SOPs, roles, approvals, service levels and integrations live in a governed Customer Pack.
The core product stays common. Governed configuration releases keep your operating rules consistent. Knowledge is scoped across global, jurisdiction and customer layers.
How we put it to work