Telecom operators are sitting on massive volumes of network, customer, and operational data, but “having the data” no longer wins. What wins is turning that data into actions that reliably improve outcomes: SLA adherence, customer experience, energy efficiency, fraud prevention, and network performance.
In a recent AppDevANGLE episode, I spoke with Molham Aref, CEO and Founder at RelationalAI, Joji Philip, Head of Industry GTM, Global Communications Vertical at Snowflake, and Sreedar Rao, Global Telecom CTO at Snowflake to unpack why decision intelligence (DI) is emerging as the missing layer between BI dashboards, generative AI, and autonomous telecom operations.
“If running your business was like driving a car, business intelligence is the dashboard,” Aref said. “Decision intelligence is the navigation system. It figures out the best way to get to your desired destination.”
From BI dashboards to Outcome-Driven Decisions
Most telecom organizations have mature reporting and analytics capabilities. BI is good at showing what happened and what’s happening now (e.g., congestion hotspots, churn trends, capacity utilization, and operational KPIs).
DI shifts the focus from visibility to decisioning: predictive and prescriptive reasoning that recommends, and increasingly executes, the best action given a set of constraints.
Molham Aref outlined multiple telecom-specific DI domains:
- Network optimization and management: predictive maintenance, dynamic traffic routing, deployment planning, digital twins
- Customer experience and retention: churn prediction and prevention, hyper-personalization, proactive support
- Revenue protection and operations: fraud detection, revenue assurance, strategic marketing
- Strategic planning: tower/radio placement, spectrum allocation, supply chain optimization, ESG/compliance tracking
The throughline: DI becomes the “how do we get from here to the outcome” engine that BI alone can’t provide.
Programmable networks need auditable decisioning
Joji Philip framed DI as the point where programmable connectivity stops being hype and becomes a product reality.
Operators are pushing network capabilities up the stack through APIs, and those APIs carry implicit promises of stable latency, prioritized throughput, verified device status, and trusted identity signals. But as soon as enterprises can request outcomes, the network needs to prove it can deliver them.
“Once you expose network capabilities via APIs, you’re effectively selling decision connectivity,” Philip said.
He described three “decision families” that show up immediately in real networks:
- Service assurance decisions: not just identifying congestion, but deciding how to steer traffic, prioritize users, and place workloads (including at the edge) to meet QoE and SLA intent
- Energy and capacity tradeoffs: shifting energy modes, balancing power versus throughput without breaking experience targets
- Intent validation: proving that the network delivered the requested outcome, and if not, producing an evidence trail and corrective action
That last point matters: DI isn’t only “automating actions.” It’s making actions defensible.
Closed-Loop Automation vs. Decision Intelligence
Sreedar Rao drew a sharp line between traditional closed-loop automation and DI.
Closed-loop automation typically encodes tradeoffs ahead of time: fixed rules and predetermined outcomes. That works until the environment becomes dynamic enough that rules don’t generalize and “unknown unknowns” start showing up.
“The distinction… is that there’s an inherent trade-off that happens,” Rao explained. “With traditional closed-loop automation, those trade-offs are baked in. In a dynamically changing environment, a predetermined set of rules aren’t sufficient.”
DI platforms are built to reason through tradeoffs in real time, and critically, to do it in an auditable way: what decision was made, why it was made, and what alternatives were discarded. Rao tied that directly to operator ambitions for L4/L5 autonomous network operations, where explainability and governance are non-negotiable.
Why Governance and Architecture Matter Now
Aref pointed out a practical reason DI has historically been hard: it often required introducing new stacks into the footprint. That meant copying data, re-securing it, re-governing it, and integrating it back into operational workflows. The architectural bet RelationalAI and Snowflake are making is to bring DI into the platform where telecom data already lives and is governed.
“No data leaves,” Aref said. “It all runs inside the security perimeter. It’s governed the same way.”
This is more than convenience. It changes what teams can safely do with agents and GenAI-driven workflows because it reduces the “silo tax:” inconsistent semantics, duplicated pipelines, and fractured controls.
Rao extended that architecture view into an execution model telecom teams can act on:
- Data accelerators: shorten the path from data generation to actionable intelligence
- A knowledge plane: abstract raw data into telco-specific semantics and constructs
- Decision as a service: expose DI as a shared service for many apps/agents (rather than embedding decision engines inside monoliths)
- A marketplace mindset: data + AI + agents consuming a common decision layer
The goal is intelligent infrastructure that’s deterministic, logical, and auditable, not “autonomous” in a way that feels random.
GenAI Needs DI to Unlock Enterprise Value
Aref closed with a hard truth we’re seeing across industries: frontier models are powerful, but enterprise value creation isn’t scaling at the same rate.
There’s a gap between what GenAI can do and what organizations can operationalize. DI helps close that gap by giving agents tools for predictive and prescriptive reasoning that go beyond BI:
- Predictive reasoning for what’s likely to happen
- Prescriptive reasoning for what action best meets objectives
- Graph and rule-based reasoning to enforce constraints and policies
He also noted an important feedback loop: GenAI can reduce the labor required to implement DI (semantic modeling, prescriptive model development, eval creation), making DI more achievable at scale, and DI makes GenAI more outcome-driven.
Why This Matters for Developers and Platform Teams
Telecom is becoming an application platform story: programmable APIs, edge workload placement, intent-driven SLAs, sustainability targets, and autonomous operations. DI is emerging as the control plane that ties those goals to governed data and auditable decisions.
If you’re building in or around telecom platforms, the takeaway is simple: insights don’t create outcomes; decisions do. And the organizations that treat DI as a first-class architecture layer will be the ones that turn AI from experimentation into production-grade automation with proof.
Watch the AppDevANGLE episode with Molham Aref (RelationalAI), Joji Philip (Snowflake), and Sreedar Rao (Snowflake) to hear how decision intelligence is being positioned as the outcome engine for telecom’s next wave of AI modernization.
SUSE SUSECON 2026: VMware Migration Strategy and AI Platform Play
Synopsys and TSMC Deepen AI Silicon Partnership at Advanced Nodes
Contrast and Google Cloud Bring Runtime Context to AI-Driven SOCs
Eclipse Foundation Brings Enterprise-Grade Operations to Open VSX
TinyMCE AI Embeds Writing Intelligence Directly Into Dev Workflows
