Ilan Sade, Division President, GenAI & Data at Amdocs – Interview Series

Ilan Sade, Division President, GenAI & Data at Amdocs brings more than two decades of leadership within the same organization, rising from early technicals as a programmer and project manager to overseeing large-scale global delivery, product strategy, and innovation. Throughout his tenure, he has led critical divisions spanning revenue management, digital and business support systems, and open network initiatives, culminating in his leadership of the T-Mobile division before stepping into his current focused on generative AI and data. His career reflects deep domain expertise in telecommunications, particularly in complex billing systems, customer experience platforms, and large enterprise transformations, positioning him at the forefront of Amdocs’ shift toward AI-driven operations and next-generation data platforms.

Amdocs is a multinational software and services company specializing in solutions for communications, media, and digital service providers, helping them manage everything from billing and customer relationships to network operations and digital transformation. Founded in 1982 and operating in over 90 countries, the company has evolved into a key infrastructure provider for telecom operators, offering cloud-native platforms, AI-driven analytics, and automation tools that enable more efficient service delivery and personalized customer experiences. Its growing focus on generative AI and data platforms reflects a broader industry shift toward intelligent, software-defined networks and fully digitized customer ecosystems.

You’ve spent more than two decades at Amdocs, rising from a developer to leading the GenAI and Data division, and previously overseeing one of the company’s most strategic partnerships with T-Mobile. How has that journey shaped your perspective on what it actually takes to move AI from experimentation into production at telecom scale?

What I have learned over the years is that moving AI into production at telecom scale is not primarily a model problem. It is an operations problem. You need the right data foundations, strong integration into existing systems, clear accountability, and teams that know how to run AI as part of day to day business processes. If any one of those pieces is missing, pilots can look impressive but they do not scale.

My path at Amdocs gave me exposure to all sides of the equation, from engineering to customer delivery to large operator partnerships. That experience shaped my view that success comes from combining technical excellence with execution discipline. In telecom, AI has to work across complex environments, support real service levels, and deliver measurable outcomes. That requires a production mindset from day one.

At the Mobile World Congress (MWC), there was a clear signal that telecom companies are heavily investing in AI pilots but struggling to operationalize them. In your view, what are the biggest blockers preventing operators from moving beyond experimentation today?

I see one of the biggest blockers being fragmentation. Most operators have valuable data and strong use cases, but their environments are split across a wide range of systems, teams, and vendors – making it hard to connect AI outputs to real workflows. This is especially true when those workflows span network, customer care, and business operations. As a result, AI often remains a point solution instead of becoming part of the operating model.

Additionally, another blocker that I’ve witnessed is trust – operators ultimately need reliability, governance, and clear controls before they can integrate AI into critical processes. For instance, if they can’t explain why an AI agent has made a decision or enforce policies around it, that technology will be kept in a pilot lane. Moving forward requires a framework that combines automation with observability, auditability, and human oversight.

Amdocs is positioning aOS as an “agentic operating system.” How do you define agentic AI in the context of telecom, and how is it fundamentally different from earlier AI-driven automation approaches?

Within the telecom space, agentic AI specifically refers to the technology that can understand goals, plan tasks, take actions across multiple systems, and adapt based on results. Instead of simply generating content or predicting outcomes, agents can execute workflows end-to-end. They’re able to reason over context, collaborate with other agents, and operate within governance boundaries to complete real operational tasks.

This is fundamentally different from earlier automation, which was mostly rule-based and static. Traditional automation worked well for repetitive tasks in stable environments, but it struggled with complexity and exceptions. Agentic AI can handle dynamic situations, learn from feedback, and coordinate across domains.

You’ve described a future of AI-native telecom operations. What does that actually look like in practice, and how far away are we from fully autonomous, networks?

AI-native telecom operations looks like AI that is embedded into the core of how a business runs – not just added on top. In practice, that looks like service assurance workflows that detect and resolve issues before customers notice, customer care journeys that are personalized and proactive, and network operations that continuously optimize performance based on real time conditions. The key is that AI is integrated into decisions and execution, not just analytics.

We are not at fully autonomous networks yet, and we should be realistic about that. The next few years will be about progressive autonomy, where operators automate more complex workflows while keeping humans in control of high impact decisions. Full autonomy will require stronger standards, broader interoperability, and continued improvements in reliability and governance.

Telecom systems have historically been fragmented across Operations Support Systems (OSS) and Business Support Systems (BSS) layers, which has made end-to-end automation difficult. How does an agentic architecture help unify these domains and enable cross-functional workflows?

Agentic architecture helps by introducing a coordination layer that can work across OSS and BSS without forcing a complete system replacement. Agents can connect to existing platforms through APIs, understand the context of a business objective, and then orchestrate the right sequence of actions across network, service, and customer systems. This allows operators to automate workflows that previously broke at domain boundaries.

For example, if there is a network issue affecting a high value enterprise customer, an agentic system can correlate the fault, assess impact, trigger remediation steps, and update the customer communication flow in parallel. That kind of cross functional execution is difficult with traditional automation because each domain operates in isolation. Agentic workflows help close that gap.

One of the interesting aspects of agentic systems is the collaboration between AI agents and human operators. Where do you see the balance landing between automation and human oversight in telecom environments?

The balance between AI agents and human operators will always depend on the specific use case, but will largely be human-led and AI accelerated for the foreseeable future. AI agents are excellent at speed, scale, and pattern recognition, while human operators bring judgment, accountability, and context. The goal is not to remove people from the loop. It is to let people focus on decisions that require expertise while AI handles the heavy operational workload.

In practice, that means setting clear thresholds for autonomous actions and escalation paths for exceptions. Low risk, repetitive tasks can be automated with minimal oversight, while high impact decisions should always include human approval. This approach builds trust and helps operators scale AI safely across mission critical environments.

There’s a lot of hype around generative AI, but telecom operators are ultimately focused on ROI. What are the most important metrics CSPs should be tracking to determine whether AI deployments are actually delivering value?

Operators should track metrics that tieectly to business outcomes, not just technical performance. On the customer side, that includes first contact resolution, average handling time, churn reduction, and customer satisfaction. On the network side, it includes mean time to detect and mean time to resolve incidents, service availability, and operational efficiency gains.

It is also important to measure adoption and reliability. If agents are deployed but teams do not trust them, the value will not materialize. CSPs should track how often AI recommendations are accepted, how often workflows complete successfully, and how often human intervention is required. ROI comes from sustained operational impact, not isolated pilot results.

aOS emphasizes multi-agent workflows that can execute complex, end-to-end processes across telecom environments. How do you ensure coordination, reliability, and governance when multiple AI agents are operating simultaneously across critical systems?

Coordination starts with a clear orchestration model. In a multi agent environment, each agent should have a defined, access boundaries, and success criteria. A central orchestration layer manages task sequencing, conflict resolution, and state tracking so that agents do not work at cross purposes. This keeps workflows predictable even when they span many systems.

Reliability and governance require strong controls by design. That includes policy enforcement, audit trails, explainability, and real time monitoring of agent behavior. It also means having fallback mechanisms so workflows can pause, escalate, or roll back safely if something unexpected happens. In critical telecom systems, governance is more than an add-on – it’s a core requirement.

In a recent aOS announcement, Amdocs positions generative AI as evolving from a “sidecar” capability to a core operational layer embedded across customer, network, and business processes. What changed in the last 12 to 24 months that makes this shift possible today?

Several things matured at the same time. Foundation models improved significantly in reasoning quality and tool use, which made them more capable in enterprise workflows. At the same time, the ecosystem around them improved, including orchestration frameworks, observability tools, and governance controls. That made it practical to move from isolated use cases to coordinated operational workflows.

The other major change is organizational readiness. Operators now have clearer priorities around AI and stronger pressure to deliver measurable results. They are no longer experimenting just to learn. They are looking for platforms that can scale AI across functions with security and control. That shift in maturity on both the technology and business sides is what makes this moment different.

If aOS represents a turning point toward AI-native telecom operations, what does the next phase look like? Are we heading toward fully autonomous telecom networks, and what challenges still need to be solved before that becomes reality?

The next phase is about scaling from isolated automation to coordinated autonomy across the enterprise. We will likely see more multi agent workflows that connect customer care, service operations, and network teams in real time. Operators may move from reactive operations to predictive and proactive models, where AI can identify risks early and execute remediation before issues escalate.

Fully autonomous networks are a long-termection, but there are still important challenges to solve. We need stronger interoperability across vendor ecosystems, more robust governance standards, and continued progress in reliability and explainability. Most importantly, the industry needs confidence that autonomous systems can perform safely under real world conditions. The path forward will be incremental, with clear controls at every step.

Thank you for the great interview, readers who wish to learn more should visit Amdocs.