OSS/BSS Data Strategies for AI Success — Data-Driven Telco Summit 2026, Panel Discussion 2 video

The Data-Driven Telco Summit 2026, Panel Discussion 2
— OSS/BSS Data Strategies for AI Success

• Moderator: Matt Leary, TelcoTitans.
• Panelists: John Abraham (Appledore Research), Bob Dietrich (Wavelo), Koenraad Peeters (Ciminko).

Panel overview

Decades of best-of-breed, one very expensive legacy

John Abraham of Appledore Research opened by tracing telecom’s data problem back decades, to an era when systemic transformation was “highly compartmentalised” and best-of-breed procurement won out over integration [00:00:43]. Old systems were rarely retired — just scoped down as new ones layered on top — and M&A compounded it further. His firm’s estimate: roughly two-thirds of annual legacy IT spend goes to systems that “don’t have a future” [00:02:00]. The example that lands hardest: a US tier-one operator that at one point ran “over 300 billing systems” [00:02:35].

Why event-driven architecture, and why now

Abraham described EDA’s pub/sub model plainly — publishers register events, subscribers pull what they need, without the point-to-point integration mesh that made 300 billing systems unmanageable [00:04:38]. The historic objection — that EDA is “fire and forget”, with no state or context — is, in his view, resolved: modern tooling gives full observability into “who publishes, who subscribes, what is the state of the information” [00:06:00].

Inside TM Forum’s async API Catalysts

Bob Dietrich detailed Wavelo’s role reviving TM Forum’s async API working group, aiming to bridge the huge embedded base of existing REST APIs into event streams without forcing a rip-and-replace [00:07:16]. Koenraad Peeters, who led the underlying TM Forum Catalyst work, explained its evolution: an initial “request-reply” pattern that technically used event infrastructure but “is not very event driven” [00:11:19], followed by a second Catalyst explicitly built for “transitional brownfield” telcos — because, as he put it, “no telco is greenfield” [00:12:12]. The goal: let telcos keep their REST integrations while gaining real access to their own data for AI.

Ontology as the missing layer

Abraham introduced Appledore’s three-tier model — data at the base, ontology/semantics in the middle, AI applications on top [00:15:57] — and broke ontology into three components: semantic (what a “customer” actually is, however each system names it), decision-layer reasoning, and the resulting action [00:17:58]. Dietrich described how Wavelo implements this as a dynamic, contextual model rather than a static schema, with built-in validation for non-TMF-compliant source systems [00:19:19].

Is OSS/BSS actually dead?

Peeters’ answer: no — the event bus itself becomes “your storage of data… your source of truth”, with roughly 100 TM Forum-defined business entities providing a common model that legacy and new applications alike can plug into [00:21:48]. Dietrich backed this with a concrete result: Wavelo migrated a legacy billing system in under six months by streaming data incrementally rather than moving it in bulk [00:25:22]. Peeters added a second pattern — running five CRMs in parallel against the same event bus during a migration, so cutover happens once the data is proven, not before [00:26:03].

All three panellists ended firmly bullish, with Abraham noting that the old determinism and observability concerns are, in his words, now handled by “a carrier-grade solution” [00:24:39].