By: Bryan Hamman - area vice-president (AVP) for Africa, NETSCOUT

Artificial intelligence (AI) is fast moving from experimentation into everyday operations. Across industries, organisations are looking at how AI can automate processes, speed up decision-making and help teams manage increasingly complex digital environments.

But as AI takes on a greater role in day-to-day operations, another, more difficult to answer question is emerging: how much do businesses actually trust the data those decisions are based on?

Recent findings from NETSCOUT's Cisco Live 2026 survey of nearly 1,000 IT professionals revealed that only 17.2% of respondents did not already have AI projects underway or planned to launch within the year. Adoption, in other words, is not the bottleneck… but trust is. More than half of those surveyed (56.2%) said that their technicians needed hours to resolve AI performance issues, and one in five said that this could take a day or more to do. Only 11% were able to get to the bottom of these issues consistently within minutes. And while the survey was conducted among IT professionals within the US, the pattern it captures – deployment racing ahead of operational readiness – will be familiar in any country.

For African organisations, this is particularly relevant. Companies locally continue to modernise their digital environments, managing a mixture of legacy infrastructure, cloud services, mobile networks, third-party platforms and geographically dispersed operations, all at the same time.

There is understandably considerable attention on how quickly organisations can adopt AI. However, choosing an AI platform and identifying processes that can be automated is the easier part. What is tougher to ascertain, as digital infrastructure becomes more distributed and interconnected, is whether there is enough visibility into the environment within which those AI systems will actually operate. 

Knowing that a system is available is no longer enough. Teams need to understand how applications, networks, infrastructure and users are interacting, and what is happening when something goes wrong. Observability, in that sense, has stopped being an IT monitoring issue and become an operational requirement. 

These implications are particularly significant within Africa, with digital banking, mobile money, e-commerce, healthcare platforms and government services, for instance, all depending on reliable digital experiences. When a mobile money platform goes down for even an hour in a market where it’s the primary way people can move money, it becomes a livelihood problem, affecting revenue, productivity and trust.

‘More data is better’ no longer applies

As AI moves deeper into operations, the quality of the underlying data becomes more and more important. If it is fragmented or lacks the necessary context, AI will still produce a confident-sounding answer, but there is a risk that it will not accurately reflect what is really happening.

This is where the distinction between collecting more data and having better data becomes important, increasingly described in the industry as ‘context engineering’ rather than ‘prompt engineering’. Essentially, adding more telemetry does not necessarily provide a clearer picture or context. 

The NETSCOUT survey draws specific attention to this, with 64% of respondents saying that deep packet inspection (DPI) is important or very important for AI observability, yet only 9.7% actually use it as their primary data source. Most teams are still looking to metrics, logs and traces to infer what happening on the network, rather than looking at DPI data, which remains a vital source of operational evidence. 

Tellingly, the single biggest barrier cited to effective AI operations, named by 27.2%, was a combination of siloed tools, a lack of real-time visibility, incomplete data context, trust, noisy telemetry, high data storage costs and rigid policies.

High-fidelity operational data closes that gap by providing evidence of what is actually happening across the digital environment, giving context through a detailed view on how users, applications and infrastructure are really interacting. NETSCOUT ‘Smart Data’ provides trusted, contextualised operational intelligence that helps organisations to understand how complex digital services actually perform across hybrid and multi-cloud environments.

From automation to operational resilience

The AI conversation also needs to move beyond automation for its own sake. There is real value in using AI to spot anomalies, accelerate troubleshooting and handle repetitive operational tasks. But for businesses running critical digital services, the bigger objective is resilience: can AI help keep services available, help teams to respond faster when problems emerge, and support better decision making under pressure?

Observability earns its keep here too, not just as something used primarily after an incident has occurred, but as the ongoing context needed to catch problems early and understand dependencies. It can help to connect technology performance to business outcomes. If a network problem affects a specific application, customer group or service, understanding that relationship and all the dependencies quickly is often the difference between a contained issue and a much wider business disruption.

Trust and confidence become more important as AI becomes autonomous

Trust matters more as organisations move workstreams towards agentic AI. There is a real difference between an AI system recommending an action and one taking that action itself. When a human stays in the loop, a bad recommendation can be caught before it does damage. As AI systems gain autonomy, however, the quality and reliability of the information feeding those systems becomes far more consequential.

This emphasises the importance of knowing where an AI-driven decision came from: validating the information behind it and tracing the decision back to reliable operational evidence (where necessary); it does not mean removing people from the process. Rather, AI outcomes should enhance human expertise by giving people better information and helping them to act more quickly and confidently.

Africa has an opportunity to build for AI readiness

There is a genuine upside for African organisations here. Many businesses across the continent are still building out their cloud, data and digital capabilities, which means that they can consider observability, operational resilience and governance as foundational elements of AI readiness, rather than trying to bolt them on later as many more mature markets are currently required to do.

The organisations that get this right will not necessarily be those that deploy AI fastest. They will be the ones that understand the environment in which their AI operates and ensure the information available to both their people and their machines is trusted data you can have confidence in when making decisions.

This means asking more difficult questions than ‘do we have enough compute capacity’? It means examining whether the organisation can see what is happening across its digital environment, if that information is sufficiently detailed and contextual, and whether teams can validate the decisions being made from it.

For NETSCOUT, this is the foundation of trusted AI-driven operations: high-fidelity, packet-derived intelligence that can provide an observed view of real digital interactions, not an inferred one. Turning raw network activity into structured, contextual Smart Data is what makes that context usable, alongside existing operational tools, and it’s ultimately what provides a trusted foundation for better decisions. 

AI adoption is accelerating – and NETSCOUT’s own numbers contest to this - but adoption alone is not the measure of success. The real opportunity is to build AI-enabled operations that organisations can understand, validate and trust. For Africa's fast-developing digital economies, closing that gap now, while the infrastructure is still being built, is a more valuable head start.