OSS/BSS Data Strategies for AI Success
AI dominated the conversation at this year’s industry events. Agentic AI. Generative AI. Autonomous operations.
But behind the excitement, a more technical — and more urgent — discussion is emerging:
If AI runs on data, what happens when your data architecture wasn’t built for it?
In this panel discussion, industry leaders unpack the foundational challenge facing telecom: legacy OSS/BSS environments that were never designed to support real-time, AI-driven use cases at scale.
The Legacy Reality: Silos by Design
Telecom didn’t get here overnight.
Over decades, operators adopted “best-of-breed” systems across billing, CRM, ordering, network management, and more. Each transformation was often department-led. Each new system solved a specific problem. Integration was deferred.
The result?
- Hundreds of siloed systems at large operators
- Massive point-to-point integrations
- Two-thirds of billing spend going toward legacy maintenance
- Data trapped inside applications
For years, the industry learned to live with it. Work around it. Add another layer. Build another integration.
AI changes the equation.
Because AI doesn’t tolerate fragmented, inaccessible, or context-less data.
Why APIs Alone Aren’t Enough
REST APIs helped standardize integrations. TM Forum APIs, in particular, saw widespread adoption across global operators.
But APIs — especially synchronous, point-to-point APIs — were not designed for:
- One-to-many data distribution
- Real-time streaming
- Persistent telemetry and historical replay
- Feeding AI agents with contextual, cross-domain data
Even when data is extracted into lakes, it often becomes static and detached from operational context.
AI needs something different.
It needs live, trustworthy, semantically aligned streams of data.
Event-Driven Architecture: From Integration Mesh to Data Backbone
The panel explored how Event-Driven Architecture (EDA) changes the model.
Instead of point-to-point integrations, systems publish events to a central event broker. Any authorized system — or AI agent — can subscribe.
This shift delivers:
- One-to-many communication at scale
- Real-time data streaming
- Built-in telemetry and observability
- Reduced integration complexity
- Incremental modernization without shutdowns
In this model, the event bus becomes more than plumbing.
It becomes the source of truth.
Every state change. Every business entity. Every interaction — captured as a stream.
And critically: legacy systems don’t have to disappear overnight. They can coexist, publishing into the same backbone while modernization happens incrementally.
The Missing Layer: Ontology and Context
Data alone isn’t enough.
Telecom has its own language — acronyms, entities, and domain-specific definitions that vary across systems. A “customer” in one system may not match a “customer” in another.
That’s where ontology enters the conversation.
Think of it as the semantic backbone between raw data and AI applications:
- Semantic layer: What is a customer? A product? An order?
- Decision layer: How are business rules applied?
- Action layer: What should happen next?
Without this shared semantic structure, AI pilots remain siloed experiments.
With it, AI can operate across domains — with context.
Bridging Brownfield to AI-Ready
One of the most practical themes in the discussion was coexistence.
No telco is greenfield.
The real challenge is transitional:
- How do you keep REST APIs working?
- How do you translate synchronous calls into event-driven streams?
- How do you enable legacy and modern systems to operate in parallel?
The answer isn’t a five-year rip-and-replace.
It’s layering event infrastructure over existing environments — allowing:
- Parallel system operation
- Gradual migration
- Real-time hydration of new systems
- Reduced transformation risk
In practice, this approach has enabled major migrations in months — not years — without disrupting live operations.
Is OSS/BSS Dead?
Far from it.
In fact, OSS/BSS becomes more valuable in an event-driven model.
When every application publishes state changes into a shared event backbone:
- Legacy systems continue operating
- AI agents access the same source of truth
- Business entities align to common models
- Optimization becomes continuous
OSS/BSS doesn’t disappear.
It evolves — from siloed systems of record to participants in a unified, streaming architecture.
The Real AI Readiness Test
The panel closed on a note of optimism.
Event-driven architecture isn’t experimental. It’s mature. It powers global-scale digital platforms today. Observability concerns have been addressed. Scalability has been proven.
The remaining question is no longer technical feasibility. It’s organizational momentum.
AI success in telecom won’t be determined by model sophistication alone. It will be determined by whether operators modernize the data foundations underneath their OSS/BSS environments.
Because AI isn’t blocked by imagination.
It’s blocked by architecture.
If AI readiness, legacy modernization, or event-driven architecture is on your roadmap, let’s continue the conversation.
