As artificial intelligence reshapes enterprise communications, forward-thinking companies are moving beyond simple AI adoption to comprehensive transformation strategies. Matija Kapic, Head of Solution Engineering for Asia at Infobip, offers a compelling roadmap for CIOs navigating this complex landscape. From architecting multi-model AI ecosystems to deploying agentic intelligence at scale, Kapic shares insights on building adaptable infrastructures that can evolve with rapidly changing foundation models. With Infobip’s recent launch of its Conversational Experience Orchestration Platform (CXOP), the company demonstrates how AI-first approaches are revolutionizing customer engagement across messaging channels while maintaining security and human oversight.
Head of Solution Engineering – Asia
Infobip
CIO&Leader: How should CIOs architect their AI strategy to adapt to rapidly evolving foundation models and emerging AI technologies?
Matija Kapic: At Infobip, AI is no longer an add‑on; it’s becoming the backbone of the enterprise stack. The CIO’s task, therefore, is to build a future‑proof architecture rather than “adopt AI.” With foundation models evolving every month, we need to design an API-centric infrastructure that allows the team to switch between open-source, proprietary, or fine-tuned models without complete re-engineering.
We also reject one‑size‑fits‑all thinking. Each use case — conversational commerce, fraud detection, internal productivity — must get the best‑fit engine, whether a general LLM, a domain model, or a custom agent. A multi‑model posture plus partnerships across cloud providers, open‑source communities, and nimble AI startups keep innovation flowing.
Ultimately, CIOs must act as system architects, product strategists, and change leaders. Success won’t come from chasing each new model, but from cultivating an adaptable, ethical, and outcome‑driven AI ecosystem.
CIO&Leader: What technical capabilities do IT teams need to develop to support and manage agentic AI deployments?
Matija Kapic: With a rapidly changing technology landscape and new developments emerging every day, IT teams need to stay ahead of the curve. They must build capabilities across both foundational software skills and specialized AI frameworks to support and manage agentic AI deployments. Key focus areas include data, orchestration, and governance. This means developing unified data platforms, enabling real-time adaptability, and adopting an agentic AI mesh architecture to support modular and scalable deployment.
Teams must implement MLOps (Machine Learning Operations) practices for continuous model lifecycle management and establish robust security frameworks incorporating access control, auditability, and explainability. As workflows shift from manual to agent-first, new oversight roles will be required to supervise autonomous agents and ensure alignment with business objectives.
A vendor-neutral, composable architecture is essential for long-term flexibility. Equally important is governed autonomy, enabling proactive, self-acting systems that can adapt and collaborate across enterprise environments while managing risk and maintaining trust. Together, these capabilities are crucial for realizing the full potential of agentic AI.
CIO&Leader: What security frameworks are essential when agentic AI handles sensitive customer data across multiple touchpoints?
Matija Kapic: Security can’t be an afterthought while deploying Agentic AI; it needs to be a part of the process. To ensure that sensitive customer data across multiple touchpoints is handled securely, ensure the agentic framework supports encryption for data both at rest and in transit. Implement strict access controls and mechanisms to redact or remove sensitive information where necessary. Since these agents source and act on large volumes of customer data, safeguards around data collection, storage, and usage must be clearly defined and enforced. It’s also essential to work with partners who understand communication compliance and can help navigate regional data privacy regulations. Ultimately, privacy should be embedded into the architecture from day one.
CIO&Leader: What are the core infrastructure requirements for deploying agentic AI at enterprise scale, and how does CXOP integrate with existing tech stacks?
Matija Kapic: For large-scale enterprise-level deployment of Agentic AI, companies must have an infrastructure that can support real-time processing, seamless orchestration, and integration across multiple systems and channels. Primarily, one needs a platform that can move beyond static, rules-based workflows to dynamic, goal-oriented conversations across functions like marketing, sales, and support.
Our Conversational Experience Orchestration Platform (CXOP) addresses this by natively infusing agentic AI across Infobip’s product stack by unifying channels, data, and automation into a single intelligent platform. Built on Microsoft Azure OpenAI in Foundry Models, CXOP uses intelligent AI agents to orchestrate customer journeys across channels such as WhatsApp, RCS, and web chat. These agents understand context, act autonomously, and collaborate with human teams when needed. This infrastructure setup allows for scalable deployment while reducing time to resolution, improving customer loyalty, and managing costs. All this is done without requiring major overhauls to the existing tech stack. CXOP’s integration ensures enterprises can adopt agentic AI while maintaining consistency across systems and touchpoints.
CIO&Leader: How is Infobip positioning itself in the competitive communications platform landscape with its AI-first approach?
Matija Kapic: As the shift toward conversational engagement accelerates, Infobip is responding with an AI-first approach designed to meet evolving customer expectations. Consumers increasingly want to interact with brands through messaging channels, and Infobip’s AI-powered Experiences helps businesses deliver precisely what they need. While resolving pain points and advancing digital transformation, the company is working towards deepening customer loyalty. The launch of the CXOP builds on this by unifying Infobip’s SaaS and CPaaS platforms and infusing them with advanced AI capabilities.
At the center of CXOP are agentic experience agents that personalize and orchestrate conversations across all user touchpoints. This layered, intelligent approach, combined with value-added features like generative AI, QR codes, and integrated payment options, allows Infobip to add depth to every channel. Recognition from Gartner as a CPaaS Leader and from Juniper Research for embedding generative AI and value-added services reflects Infobip’s positioning as a continuously evolving, AI-driven leader in the communications platform landscape.
CIO&Leader: How do you maintain oversight and control when AI agents make autonomous decisions in customer interactions?
Matija Kapic: Maintaining oversight when AI agents make autonomous decisions starts with building transparency and trust into every interaction. Ensure agents are trained to communicate clearly, so customers know what’s happening and feel confident their issue is being handled appropriately. Always offer the option to escalate to a human agent to maintain trust. While autonomy is a strength, agentic systems can still make mistakes. That’s why it’s essential to keep a human in the loop, especially for monitoring decisions in sensitive or complex scenarios. Oversight should be proactive, not reactive, with controls embedded across the customer journey.