Login
Register
As AI adoption accelerates across telecoms, fragmented model access, rising inference costs and ungoverned agent sprawl are creating new operational risk. This Catalyst shows how a unified Model-as-a-Service control plane can give CSPs one governed route to any AI model, improving security, cost control and developer velocity.
Helping telcos scale AI safely without slowing innovation
Communications service providers are moving rapidly from AI pilots to enterprise-wide adoption, but the operating model for consuming AI models has not always kept pace. Different teams often connect directly to different model providers, build their own credentials and retry logic, and manage costs, security and compliance in disconnected ways. As agentic AI becomes embedded in telecom operations, this fragmentation can lead to agent sprawl, inconsistent governance and limited visibility into how models are being used across the business.
The Catalyst project, MODaaS: Accelerating agent velocity, showcased at DTW Ignite in June 2026, addresses this challenge by creating a Model-as-a-Service control plane for telecom AI. It gives operators a single, governed way to access models from any provider, while improving observability, controlling cost and reducing the integration burden that slows AI development.
AI adoption inside telecom operators is accelerating, but many organizations still consume models in an unstructured way. Teams may choose different frontier models or specialist providers for different tasks, often without centralized routing, monitoring or compliance controls. This creates several risks at once: duplicated integration work, limited real-time cost visibility, unnoticed model quality drift and inconsistent handling of sensitive data.
The problem is particularly acute in telecoms because general-purpose frontier models can perform poorly on domain-specific telecom tasks unless they are selected, governed and monitored carefully. Operators need the flexibility to use the right model for the right use case, with assurance that every request is secure, auditable and commercially sustainable.
The Catalyst proposes a unified MODaaS control plane that combines the Databricks Model Pool with the Tecnotree Model Decision Layer, enabling the platform to select the right model for each agent based on its specific needs and Agent KPIs. it introduces a Token Economics Framework built on six levers: capability fit, accuracy trade-off, latency, throughput, context window, and verbosity enabling operators to engineer the precise price-performance point for every workload. By unifying discovery, routing, security, and FinOps into one enterprise capability, MODaaS shifts CSPs from "highest intelligence by default" to "sufficient intelligence at optimal cost," turning AI spend into a governed business lever
In fact , these capabilities provide a standards-aligned operating layer for model access, governance, observability, and lifecycle management. Applications and AI agents can access models through a single API front door, while the platform handles routing, fallback, security, auditing, and cost controls behind the scenes.
This creates a more flexible and resilient model consumption pattern. Operators can swap model providers without changing application code, route requests intelligently across providers and use automatic fallback when a model or provider is unavailable. Real-time guardrails can enforce security policies, redact sensitive data and provide traceability from request to response.
For AI developers and product teams, the value is speed. Instead of spending time on integration, routing and governance plumbing, teams can focus on building and testing use cases. For operations and risk teams, the value is control. Model consumption becomes visible, auditable and enforceable through a common layer rather than hidden across multiple applications, teams and suppliers. As Justin Michaels, Director Field Engineering at Databricks, explains, the Catalyst is creating “a unified AI model governance control plane” that enables telecom operators “to scale AI safely without vendor lock-in.”
MODaaS is designed to support end-to-end governance across the entire AI and model lifecycle, as well as every AI request. The Tecnotree MODaaS platform and Databricks Mosaic AI Gateway both provide governance and traceability across models, experiments, and deployments, while the AI Gateway gives operators a controlled point through which applications and agents can access model services.
This enables operators to apply consistent security controls, gain visibility into usage patterns, monitor model performance and quality, and detect drift before it impacts customer experience or operational outcomes.
The commercial value of MODaaS lies in reducing the friction, risk and cost that can otherwise slow AI adoption at scale. A single, governed model access layer can shorten developer onboarding, reduce duplicated integration work and make it easier for teams to build AI-enabled services quickly. At the same time, real-time cost visibility and guardrails help operators prevent spending surprises as model usage grows across the organization.
Operationally, the solution improves resilience by introducing intelligent routing and automatic fallback across providers. It strengthens governance by applying unified security, privacy and audit controls to every AI request. For the wider industry, the Catalyst provides a practical blueprint for moving from fragmented AI experimentation to trusted, managed and monetizable AI at enterprise scale.
At DTW Ignite 2026, MODaaS: Accelerating agent velocity was named a finalist in the Catalyst awards, in the categories Best Moonshot Catalyst – Trustworthy AI and Data challenge, and Best Moonshot Catalyst – Showcase and storyline excellence. The recognition reflects both the relevance of its approach to trusted AI and the clarity of its demonstration. As operators industrialize AI and agentic systems, MODaaS shows how a governed Model-as-a-Service layer can help them accelerate innovation without losing control of security, cost, quality or supplier choice.
