← Back to portfolio
Assessed 2026-07-16

AIDEN

AI Draft Engineering · CSCoE / DDC / Translation & Development
Lead
Monika Broennimann
SMEs
Tiba Razmi
Provider org
CSCoE / DDC / Translation & Development

Overall coverage

Modern
70% 99 / 132 points
Coverage = unweighted mean of the 10 dimension coverages (not points ÷ max)

Maturity distribution

Modern 8Transitional 1Tech Debt 1
Headline finding
The priority for AIden going forward should be to strengthen the overall robustness, reliability, and governance of its LLM capabilities. This should include establishing a comprehensive set of LLM evaluations that goes beyond traditional unit testing and static code analysis, with a focus on model quality, consistency, safety, and expected behaviour across relevant use cases. End-to-end tracing and a complete audit trail should also be implemented to ensure that model inputs, outputs, configurations, and key decisions remain transparent and traceable.
Primary modernization lever: Resilience & Recovery

Coverage vs. portfolio

app vs. avg (dashed)

Score vs. gap by dimension

Prioritized remediation

Establish and maintain a comprehensive suite of LLM evaluations. The scope should extend beyond conventional unit testing and static code analysis to assess model quality, reliability, safety, and expected behaviour.
Deployment Automation Impact High Effort Medium
Implement end-to-end LLM observability, including request tracing, decision-path visibility, and a complete audit trail of model interactions, configurations, inputs, and outputs.
Observability Impact High Effort Medium
Define and document application-level SLAs and SLOs to support the consistent definition and calculation of relevant SLIs. Use the existing dashboard to regularly communicate service performance and reliability trends to stakeholders.
Reliability Impact Medium Effort Low
Document and test lightweight backup, restore, and recovery procedures. Clearly assign roles and responsibilities and validate the procedures through periodic recovery exercises.
Resilience & Recovery Impact Low Effort Low

Per-question assessment

click a dimension to expand · colored answer chips
FAIR & Discoverability 0 gaps 75%
The dataset or system is findable in a searchable catalog or inventory (e.g., ).
Yes
A named data owner or steward is formally assigned and documented.
Yes
Access and usage conditions, including legal and compliance constraints, are documented and available to users.
Yes
What is the percentage of the data uses standardized terminologies such as RTiS.
N/A
Data lineage is captured for key data flows and is maintained over time.
None
IAM, Privacy & Security 0 gaps 100%
The system uses the approved corporate identity provider for user authentication (e.g., Roche SSO via PingFederate, OAuth 2.0, OpenID Connect).
Yes
Data is encrypted in transit.
Yes
Data is encrypted at rest.
Yes
Secrets are managed in an approved central vault (e.g., HashiCorp Vault or AWS Secrets Manager).
Yes
Access rights and permissions are reviewed and formally certified at defined intervals.
Yes
No shared or generic privileged accounts exist in the production environment.
Yes
Platform / Hosting / Infrastructure 0 gaps 100%
Infrastructure is provisioned declaratively using Infrastructure as Code.
Yes
The system uses approved shared internal platforms or Golden Paths, where applicable. (e.g., Minerva Application Blueprint)
Yes
Architecture & Interoperability 2 gaps 75%
Can we swap in and out components without modifying the system.
Partial
All application services are exposed through clear and defined interfaces.
Partial
Data contracts are documented and governed.
Yes
API lifecycle management is actively implemented using Gravitee.
Yes
Reliability 2 gaps 50%
All critical services have defined SLAs and SLOs.
Partial
Major incidents are followed by systematic postmortems, and resulting corrective actions are tracked to closure in the backlog.
Yes
SLIs are defined and calculated for all established SLAs and SLOs, and are visible to stakeholders and system owners.
No
Scalability & Performance 1 gap 75%
The system supports automated vertical and horizontal scaling based on real-time demand, where applicable.
Yes
Load and performance testing are performed routinely and included in major release cycles.
Partial
Deployment Automation 1 gap 80%
Core workloads are supported by a complete CI/CD pipeline.
Yes
Builds, tests, and deployments are executed through fully automated processes.
Yes
Automated quality gates are enforced before production deployment. (e.g., SonarQube)
No
Production rollback or deployment occurs regularly and within the last 6 months.
Yes
Build artifacts and dependencies are managed in a traceable and controlled manner.
Yes
Observability 1 gap 75%
All critical services generate standard telemetry, including logs, metrics, and traces.
Partial
Observability data is available through approved central monitoring and observability tools.
Yes
Incident analysis can be performed across upstream and downstream dependencies.
N/A
Resilience & Recovery 2 gaps 0%
Critical services and data have defined Recovery Time Objective (RTO) and Recovery Point Objective (RPO) targets.
No
Backup, restore, and recovery procedures, including assigned roles and responsibilities, are documented and tested.
No
Cost Transparency & FinOps 2 gaps 75%
Cost visibility is available at component, system, product family, or domain level.
System
Cost is reviewed regularly together with usage and business value by the owning team.
No
Architectural decisions explicitly consider cost as a primary decision variable.
Partial