Case Studies
An 18-week Connect engagement deploying EA GraphLink across multiple repository databases, building a TM Forum-aligned Power BI semantic model, and rolling out Kernaro AI Hub to a non-technical stakeholder population: transforming how 200+ business stakeholders interact with architecture intelligence.
Global telecoms operator with commercial operations across 8 markets. The architecture function comprised 400+ architects and analysts, all working in Sparx EA. The repository was large, actively maintained, and well-governed: this was not a client with an EA capability problem.
The problem was a different one.
This engagement is anonymised. Metrics are as reported by the client.
This client had done the hard work. Years of investment in Sparx EA had produced a repository with strong MDG governance, consistent TM Forum ODA-aligned element tagging, and an active architecture community across all eight markets. The repository contained thousands of modelled elements covering business capabilities, application components, technology services, and their relationships: a significant intellectual asset.
The problem was access. The repository was useful to the 400+ architects and analysts who worked inside the EA client every day. It was invisible to everyone else.
Business stakeholders: product managers, program managers, CTO staff, commercial directors: had no direct access to architecture data. Their window into the architecture landscape was the monthly PowerPoint report produced by the architecture team: a 40-slide deck assembled from EA diagrams, manually curated, formatted for an executive audience, and delivered four weeks after the data it represented was current.
By the time the PowerPoint reached the people who needed it, it was stale. And the architecture team had spent 40 hours producing it.
The business had stopped trusting the reports. They knew the data was historical. They had started making technology decisions without consulting the architecture function because the architecture function’s outputs weren’t timely enough to be useful in fast-moving commercial decisions.
The situation was a common one, but the scale and complexity of the technical challenge was not.
Scale of the EA GraphLink deployment. The operator ran multiple Sparx EA repository databases: one per market, with a group-level repository for cross-market architecture. EA GraphLink needed to be deployed across all repository instances, with a unified GraphQL endpoint that allowed Power BI to query across the full architecture landscape. Multi-repository GraphQL deployment at this scale required careful schema design and testing.
Complexity of the data model. The repository was structured around TM Forum ODA: the Open Digital Architecture framework used widely in telecoms. The GraphQL schema needed to reflect this structure, exposing TM Forum domains and components in a way that was meaningful to telecoms business stakeholders, not just to architects familiar with the underlying modeling conventions. Generic schema design would have produced dashboards that were architecturally accurate but operationally meaningless.
Rollout to a non-technical population. Kernaro AI Hub was being rolled out to business stakeholders who had never used an AI tool for architecture queries: and who, in many cases, had never directly engaged with architecture data in any form. Configuration, onboarding, and adoption required attention. A tool deployment without a stakeholder adoption plan produces license usage statistics, not business value.
EA GraphLink was deployed across all repository instances: market repositories and the group-level repository. This required:
The deployment surface was larger than a typical single-repository Connect engagement. Week five revealed a data quality issue in two of the eight market repositories: MDG tag inconsistencies that would have produced incorrect query results in the Power BI model. These were resolved in collaboration with the client’s market architecture teams before the schema was finalized.
The GraphQL schema was designed to expose the TM Forum ODA structure in a form optimized for Power BI consumption and Kernaro AI Hub querying.
This was not a generic schema. TM Forum ODA organizes the telecoms digital enterprise across six domain groupings: Operations, Business Management, Intelligence, Partner, Employee, and Customer Experience: with a defined set of ODA components within each. The schema exposed this structure explicitly, with TM Forum domain and component identifiers as first-class query parameters.
Schema design required three iteration rounds: an initial version reviewed against the data model; a revised version tested against representative dashboard queries; and a final version optimized for Power BI DirectQuery performance at scale. Each iteration was tested against real repository data: synthetic testing at this stage would have missed the domain-specific query patterns that drove the final optimization decisions.
Power BI dashboards were built over the EA GraphLink GraphQL endpoint using DirectQuery: ensuring that every view reflects the current state of the repository at query time.
Five dashboards were delivered:
Group Architecture Overview: cross-market view of application and technology landscape by TM Forum ODA domain, highlighting coverage and dependency density.
Market Architecture Comparison: side-by-side view of each market’s architecture profile against the group reference architecture, showing standardisation status and local variation.
Application Portfolio by Domain: application inventory organized by TM Forum ODA component, with lifecycle status and technology risk indicators.
Integration Landscape: integration interface inventory showing the connections between ODA components, with interface type, protocol, and operational status.
Program Architecture Impact: active transformation programs mapped against the architecture domains they affect, supporting dependency and impact assessment at program level.
Dashboard design went through two stakeholder review rounds. The first round identified that several initially planned visualizations, while architecturally interesting, were not useful to the business stakeholder population: they were oriented towards architects, not decision-makers. The second revision shifted the emphasis to operational questions: what is running, what is at risk, what is changing, and where are the dependencies that matter for commercial decisions.
Kernaro AI Hub was configured and rolled out to a defined stakeholder population of approximately 200 business users across the group.
Configuration included:
Rollout was staged: group CTO staff and head-office program managers in the first cohort, market architecture leads in the second, broader stakeholder population in the third. Staged rollout allowed adoption feedback to inform configuration adjustments before the full population was onboarded.
Live architecture dashboards available to 200+ business stakeholders: this cohort previously had no direct access to architecture data in any form.
Architecture report production time: from 40 hours per month to near-zero: the monthly PowerPoint report was retired. Stakeholders access current architecture data directly, on demand, rather than waiting for a periodic report.
Kernaro AI Hub handling 50+ architecture queries per week from business stakeholders: without architect involvement. Common queries: “which applications support the [market] customer experience domain?”, “what integrations are affected by the [program] changes?”, “show me the technology risk profile for the partner management domain.”
Data freshness: from monthly to daily: Power BI DirectQuery against EA GraphLink ensures that dashboard data reflects the current repository state. Architecture team updates are visible to business stakeholders the same day.
> “Our architects stopped being report factories and started being architects again.” > >: Head of Architecture Practice
Starting from a well-governed repository. EA GraphLink and Power BI deliver clean, trustworthy dashboards when the underlying repository data is consistent and governed. This client had done the MDG governance work. The Connect engagement could build on that foundation rather than having to fix it first. Organizations with ungoverned repositories typically need an MDG remediation phase before a Connect engagement can deliver the results shown here.
Domain-specific schema design. A generic GraphQL schema over a Sparx EA repository exposes elements and relationships in the modeling layer’s own terms. A TM Forum-aligned schema exposes the architecture in the business vocabulary the client’s stakeholders actually use. The difference in dashboard adoption between the two approaches is significant: stakeholders engage with data they recognize.
Staged rollout with feedback loops. Rolling out Kernaro AI Hub to 200 business stakeholders in a single wave would have produced a mixed adoption result: high engagement from the cohort whose questions the initial configuration served well, low engagement from everyone else. The staged approach with configuration refinement between cohorts produced more consistent adoption across the full user population.
If your architecture team is producing reports rather than doing architecture: or if your repository is well-governed but invisible to the business stakeholders who should be using it: a Connect engagement is the appropriate next step.
Explore the Connect Service: EA GraphLink deployment, Power BI integration, Kernaro AI Hub. $50K–$185K+ depending on scope and number of repository instances.
Contact Sparx Services: describe your access challenge and we will assess the appropriate engagement scope.
Talk to a Sparx Services architect about where your organization is on the journey and what the next stage looks like.