Skip to content
Last updated: 23. apr. 2026, 06.10

Project: Agentic Orchestration Studio

Considerations

Open questions and areas that need clarification before moving from research to implementation.


Target Audience

Who is the platform for?

AudienceWhat they needPriority
IT developersAPIs, code-level orchestration, agent developmentPhase 1
Process ownersVisual process mapping, monitoring dashboardsPhase 1–2
Business unit managersROI dashboards, process status overviewPhase 2
CaseworkersTask inbox (human-in-the-loop), process statusPhase 1
Municipal leadershipStrategic overview, aggregate business valuePhase 3

Open question

The platform's UI complexity depends entirely on this answer. If only IT uses it, we can be more technical. If process owners need to model processes themselves, we need a polished visual modeler.


Roles and Permissions

Who can do what?

RoleMap processesDeploy automationsHandle tasksView dashboardsAdmin
AdministratorYesYesYesYesYes
Process designerYesNoNoYesNo
DeveloperYesYesNoYesNo
CaseworkerNoNoYesLimitedNo
ManagerNoNoNoYesNo

Open questions

  • How does this map to existing municipal AD groups?
  • Should permissions be per-process or global?
  • Who approves deploying a new automation to production?
  • Do we need an approval workflow for process changes themselves?

Existing Systems Integration

Day-one integrations to plan for:

SystemTypeIntegration approach
KMD (various)Case management, financen8n connector or REST API
SBSYSDocument/case managementn8n connector or REST API
OS2FormsForm submissionWebhook trigger
Active Directory / Entra IDIdentity, groupsKeycloak federation (or direct OIDC)
Email (Exchange)Notifications, task routingn8n email connector
Existing RPA botsCurrent automationsRobot Framework migration or API bridge

Open question

What other systems are critical? We need an inventory of systems currently involved in automated processes.


Security and Data Sovereignty

Data classification

  • What data flows through the platform? Personal data (CPR numbers)? Sensitive case data?
  • If AI agents process municipal data, where does inference happen? On-premise only?
  • Audit trail requirements — what must be logged and for how long?

Network architecture

  • Should the platform run in the municipality's existing infrastructure?
  • DMZ placement for components that receive external input (e.g., form submissions)?
  • Network segmentation between process engine, agent layer, and external integrations?

Compliance

  • GDPR data processing agreements for any cloud components
  • Danish public sector security requirements (ISO 27001, ISAE 3402?)
  • Data retention and right-to-deletion in process logs

Open question

A security review must happen before any prototype handles real data. The PoC should use synthetic data only.


Scalability

User scaling

ScenarioUsersImplications
IT team only10–30Minimal infrastructure, simple auth
IT + process owners50–200Need visual tools, role-based access
Org-wide (caseworkers)1,000–5,000Task inbox at scale, SSO required
Cross-municipal (OS2)10,000+Multi-tenant, shared infrastructure

A key advantage of open-source: no per-user licensing. The platform scales with infrastructure, not license cost.

Technical scaling

The minimal stack (Flowable + n8n + Grafana + PostgreSQL) handles hundreds of concurrent processes on modest hardware. For thousands, add:

  • PostgreSQL read replicas
  • Flowable async executors
  • n8n worker scaling
  • Redis for caching

OS2 Collaboration

The opportunity

If we build this as an OS2 product, we:

  • Share development costs across municipalities
  • Get contributions from other IT departments
  • Establish a standard for process orchestration in Danish public sector
  • Align with Denmark's Digitalisation Strategy (DKK 740M for AI, 2026)

Questions

  • [ ] Is there existing interest from other municipalities?
  • [ ] Would this fit under an existing OS2 product group, or need a new one?
  • [ ] What governance model? OS2 standard operating procedures?
  • [ ] Should we present at an OS2 event to gauge interest?

Funding and Business Case

How to sell it internally

The key argument is not "we need a cool platform" — it's:

  1. Fragmentation has a cost — maintaining N separate automations in N different systems costs more than one shared platform
  2. Visibility has value — leadership can't invest in what they can't see or measure
  3. Scaling is blocked — without shared tooling, every new automation starts from scratch
  4. Vendor risk — commercial platforms create dependency; open-source creates sovereignty

What to quantify

  • [ ] How many automations exist today? In how many different systems?
  • [ ] What is the annual maintenance cost of the current fragmented approach?
  • [ ] How many FTEs are spent on manual process steps that could be automated?
  • [ ] What is the estimated time-to-value for a new automation today vs. with a shared platform?

Funding models

  • Internal IT budget (Phase 1: PoC)
  • Digitalisation strategy funds (Phase 2: pilot)
  • OS2 co-funding if other municipalities join (Phase 3: production)

Next Steps to Resolve

QuestionOwnerDeadline
Define target audience and rolesProject teamTBD
Inventory current automationsIT departmentTBD
Security classification of process dataIT securityTBD
OS2 interest from other municipalitiesProject leadTBD
Quantify cost of current fragmentationBusiness analysisTBD