By Bernd Greifeneder, CTO and Founder, Dynatrace

As artificial intelligence moves beyond experimentation and into the core of digital operations, organisations are entering a new phase defined by autonomy, scale and complexity. Agentic AI systems, distributed cloud architectures and rising user expectations are fundamentally changing how digital services are built, operated and governed. Observability is no longer a supporting capability—it is becoming a prerequisite for control, trust and resilience.

Six key trends illustrate how AI and observability will evolve together in the coming phase.

1. Agentic AI Drives a New Level of System Complexity

Agentic AI systems are significantly more powerful than traditional models, but they are also far harder to manage. When multiple AI agents coordinate tasks, exchange context and trigger follow-on actions, even well-structured digital environments can exhibit unpredictable behaviour. Without comprehensive observability and clear governance, such systems quickly become opaque.

Each agent operates autonomously based on instructions and inputs not only from humans, but also from other first- and third-party agents. A single customer interaction can initiate hundreds of background processes in which agents independently make decisions, shift roles and delegate tasks.

Use cases such as connected vehicles or digital travel assistants illustrate this dynamic. One detected issue can activate numerous agents that analyse data, evaluate options, coordinate actions and adapt plans in real time. The volume and speed of these interactions grow exponentially. While humans remain ultimately accountable, the sheer scale of agent-to-agent communication is no longer manageable without end-to-end observability.

Organisations deploying agentic AI without a unified context risk higher costs, unpredictable outcomes and increased operational exposure. The challenge is no longer the optimisation of individual models, but the real-time control of autonomous interaction networks. Observability becomes the foundation for scalable, secure and governable agentic ecosystems.

2. Autonomy Requires Proven Operational Maturity

The shift towards autonomous operations is inevitable—but it cannot be rushed. Autonomy depends on maturity, not ambition. AI can only act independently when systems, processes and automation are stable, transparent and well understood.

Most organisations follow a gradual path. Initial stages focus on prevention, where AI detects and resolves issues before they impact users. This is followed by guided automation, with AI recommending actions and humans retaining full oversight. True autonomy emerges only once outcomes are reliable, verifiable and bounded by clear rules.

Early investments in preventive workflows and recommendation-based automation are essential. They expose gaps in data quality, system performance and contextual signals—elements on which AI critically depends. Only when these foundations are in place can autonomous operation be introduced safely and responsibly.

3. Resilience Becomes the Benchmark for Operational Excellence

Resilience is emerging as the defining metric of digital performance. In highly distributed and interconnected environments, minor disruptions can cascade rapidly across applications, cloud regions, payment systems and external services.

Reliability, availability, security and observability are no longer treated as separate disciplines. Instead, they converge into a single objective: the ability of a system to absorb disruption, respond effectively and maintain a consistent user experience under pressure.

Research commissioned by Dynatrace highlights how fragile digital ecosystems have become. In the UK alone, payment failures put an estimated £1.6 billion in annual revenue at risk; in France, the figure is approximately €1.9 billion. Customers feel failures almost immediately, and tolerance is low—many abandon transactions within minutes, while the average incident lasts far longer.

Resilience is measured not by how systems perform under ideal conditions, but by how they behave during failure. Achieving it requires shared visibility into service behaviour, fault propagation and recovery impact across the entire customer journey.

4. Reliability Becomes the Foundation of AI Progress

The next phase of AI depends as much on deterministic, fact-based signals as on generative capabilities. Creativity alone is insufficient. Trustworthy AI requires structured inputs, validated context and mechanisms that ensure confidence in outcomes.

Agentic systems amplify both progress and error. When agents coordinate and trigger downstream actions, even small inaccuracies can propagate rapidly across the system. This creates “system-level hallucinations”—not from a single model, but from reinforcing inaccuracies across many interactions.

End-to-end observability and deterministic guardrails are essential to prevent this. They ensure that all agents operate on shared facts, remain aligned with defined rules and stay under human oversight. Reliable AI is not accidental; it is engineered through transparency, consistency and control.

5. Humans and Machines Evolve into Shared Growth Engines

Agentic AI is redefining the human role in digital operations. Humans increasingly set objectives, define constraints and oversee outcomes, while AI executes well-defined tasks at speed and scale.

In practice, agentic AI behaves like a highly capable, fast-learning intern: productive with clear goals, tools and context, but still requiring guidance. AI analyses relationships, identifies risks and initiates actions. Humans interpret implications, resolve ambiguity and remain accountable.

This division of labour allows organisations to manage complexity more effectively. AI handles repetitive and time-critical execution, while humans focus on strategic decisions and system-level understanding. Sustainable growth emerges where human judgement and AI execution interact transparently and in direct alignment with business goals.

6. AI and Cloud Teams Continue to Converge

AI is no longer a standalone discipline. It is becoming a standard component of cloud-native software development, integrated as naturally as databases or APIs. As a result, AI engineering, cloud engineering, site reliability engineering (SRE) and security teams are converging into a shared operating model.

AI-driven features affect cost, latency, behaviour and compliance across the entire stack. They cannot be monitored or governed in isolation. To be production-ready, AI must follow the same delivery pipelines, standards and operational controls as other cloud-native components.

End-to-end observability is critical in this context. It must connect agent behaviour, application logic, infrastructure performance and cost signals into a single, coherent view that reflects real user experience. Organisations that adopt this model treat AI as standard software—governed centrally, implemented locally and operated with the same discipline and predictability as the rest of the digital platform.

Conclusion

The next phase of AI is not defined by smarter models alone, but by the ability to operate autonomous systems reliably, transparently and at scale. Observability is the common thread that enables control, resilience and trust across increasingly complex digital ecosystems. Organisations that invest in these foundations today will be best positioned to harness agentic AI safely—and turn complexity into competitive advantage.