We might be seeing a hidden operational bottleneck that is quietly strangling enterprise growth. Traditional IT infrastructure remains stubbornly reliant on manual intervention, placing an unsustainable burden on already stretched teams. With global skill shortages reaching crisis levels, organisations are quickly discovering that simply hiring more engineering specialists is no longer a viable way to scale. Instead, the focus has shifted entirely towards autonomous, self-driving networks capable of absorbing the daily grind of diagnostic and remediation work. Yet, as business leaders contemplate handing critical execution over to agentic systems, a deep-seated fear of losing control and visibility continues to hold many back.
Addressing this hesitation requires a fundamental shift in how we approach automation. Rather than treating machine autonomy and human oversight as opposing forces, forward-thinking tech providers are designing networks where both elements work in tandem from the start. embedding artificial intelligence directly into the foundation rather than bolting it on as an afterthought, these next-generation networks can detect misconfigurations and resolve complex issues independently. This critical transition allows IT operators to step back from mundane troubleshooting and reposition themselves as drivers of strategic, intent-based governance.
We explore why Southeast Asia’s next B2B startup opportunity may be hiding on the factory floor
To understand the practical mechanics of this shift and what it means for the future of enterprise infrastructure, we spoke to Mark Ablett, Vice President, Asia Pacific and Japan, HPE Networking. We try to unpack the delicate balance between autonomous execution and human-in-the-loop guardrails, while exploring how a self-running foundation delivers a vital strategic advantage in the real world.
As enterprises consider self-managing infrastructure, what are the most common concerns around losing operational control, and how should they think about balancing autonomy with oversight?
The concern is not about losing control of the network but about losing visibility into how decisions are made and what drives them. IT leaders are increasingly comfortable automating repetitive and predictable tasks. What creates hesitation, however, is the prospect of systems making decisions without a clear, explainable rationale.
The way to approach this is not to view autonomy and oversight as opposing forces existing at opposite ends of a spectrum. True progress comes from designing both to work in tandem from the outset. This year, we announced new self-driving network capabilities, establishing the company as the industry’s first and only provider of fully autonomous, agentic AIOps networking. We further announced major advancements that expand our self-driving networking strategy across AI factories, data centres, and the enterprise edge by introducing new AI data centre networking, routing, Agentic AIOps, and security innovations designed to simplify operations and improve performance across increasingly distributed AI-driven environments.
A self-driving network is not about removing humans from the equation. It is about elevating their role from managing every operational task to guiding the system through intent, policy, and governance. While the network autonomously handles high-volume, repeatable, and well-defined processes, human oversight remains essential for evaluating outcomes, exercising judgement, and intervening where strategic or complex decisions are required.
In practice, this is a staged journey, and the pace of change in this category is moving quickly. Networks move from generating insights to recommending actions, to taking action with a human checking the result, and only then to acting independently in narrow, well-proven domains. Organisations that try to skip straight to full autonomy without going through that trust-building process are the ones that end up nervous about losing control. Organisations that build it in stages tend to find the opposite. They end up with more oversight than they had before, because the system surfaces problems a stretched team would otherwise miss.
With the ongoing shortage of specialised IT talent, how does investing in autonomous technology allow an enterprise to scale its business without needing to constantly hire hard-to-find engineering specialists?
Skilled networking talent is in short supply across markets, and the pressure to bridge that gap is only growing stronger. Over 90 per cent of global enterprises will face critical skills shortages this year, with AI-related gaps alone putting up to $5.5 trillion of economic value at risk through delays, missed revenue, and quality issues. That pressure does not diminish as networks become more complex; it accelerates and compounds over time.
Autonomous networking changes the equation by absorbing the volume of routine diagnostic and remediation work that used to require a person watching a dashboard or working through a ticket queue. Data also shows that AI tools can save IT workers up to 45 per cent of their typical workday as routine tasks are automated. A network that can detect misconfiguration, identify a rogue device, or resolve a roaming issue on its own does not need a specialist on call for every one of those events. That allows the specialists you have to focus on the areas where human judgement delivers the greatest value, instead of spending time on operational tasks that the system can efficiently manage on its own.
We saw this demonstrated at scale during the Milano Cortina Winter Olympics, where the network extended across more than 40 venues spanning three regions. Deploying dedicated engineers at every location was not feasible. What enabled the operation to run effectively was a lean team supported by a system embedded with years of accumulated operational intelligence, allowing even less experienced engineers to benefit from insights and decision patterns the network had already learned over time.
That is the model enterprises can adopt at their own scale, enabling growth and operational complexity to expand without needing headcount to increase at the same pace.
How should business leaders reposition their human IT teams to focus on revenue-generating innovation instead of just treating them as a backend support expense?
The starting point is recognising that every hour an engineer spends manually resolving a known issue is an hour diverted from work that drives meaningful business outcomes. For years, IT teams have been measured primarily on uptime and ticket resolution. While those metrics remain important, they are fundamentally defensive in nature and reveal little about the strategic value technology teams are creating for the business.
As autonomous systems take on more of the firefighting, the conversation about IT’s role has to change with it. The useful question is not how much uptime the team delivered, but what they would build if routine maintenance stopped eating their week. For some organisations, that means faster product launches because the underlying infrastructure can be trusted to support them. For others, it is using the operational data that the network now surfaces to inform decisions elsewhere in the business.
This requires a genuine shift in how IT is resourced and measured. As AI scales across the enterprise, IT teams will increasingly be responsible for governing thousands of agents operating as part of the workforce itself, a fundamentally different job description from managing helpdesk tickets. That only becomes realistic if the network can absorb the operational weight of routine management at scale, freeing a leaner, more specialised team to keep pace with AI-driven demand rather than being buried by it.
How can tech providers design ‘human-in-the-loop’ guardrails that give corporate executives peace of mind without accidentally bottlenecking the speed and efficiency advantages of AI?
It helps to start with what can actually go wrong when AI is applied to a network without the right guardrails. Autonomous systems that act on incomplete context or low-quality signals do not just fail quietly, but they can generate exactly the kind of confident, plausible-looking but wrong output that is now widely referred to as AI slop. On a network, that means misconfigurations or remediations that look reasonable but add more complexity than they remove. That risk is precisely why human-in-the-loop design matters so much, and why getting it right is not optional.
The mistake to avoid is treating human-in-the-loop as a single checkpoint that every action has to pass through. That just rebuilds the bottleneck you were trying to remove. The better design is tiered, based on risk and reversibility rather than a blanket rule. Low-risk, easily reversible actions, like adjusting a radio parameter or fixing a VLAN misconfiguration, can run fully autonomously because the cost of being wrong is low and the system has demonstrated it gets this right consistently. Higher-stakes or harder-to-reverse actions keep a human in the approval path until the system has earned enough of a track record to expand its authority.
We believe the real differentiator is not just automation, but explainable automation. IT teams need visibility into why the system made a recommendation, what telemetry informed it, and what the expected operational impact will be. That transparency builds trust and allows enterprises to expand autonomous operations without losing governance or control.
The goal is not to remove humans from network operations entirely. It is to eliminate repetitive troubleshooting and manual intervention, so IT teams can focus on resilience, performance, security, and business outcomes. The most effective AI-driven networks will be the ones that combine autonomy with clear policy guardrails, observability, and human accountability.
Why does it matter whether security is natively built into the infrastructure versus bolted on afterwards, and what does that distinction mean in practice for enterprises evaluating their options?
The difference becomes clear the moment a technical team moves beyond a demo and begins evaluating the platform in a real-world environment. When AI is layered onto an existing architecture as an afterthought, it often feels additive rather than intrinsic, with intelligence operating at the surface instead of being embedded into the core of the platform. In contrast, a platform designed with AI at its foundation behaves differently from the outset, with automation, observability, and decision-making integrated natively into the operational experience rather than bolted on later.
Structurally, this means every layer, from access control to threat detection, was designed to work together from day one, rather than being added later and kept in sync after the fact. The practical payoff is real. When a network failure spans multiple sites at once, a natively built system can pinpoint the exact timestamp and scope of the failure, moving a team straight to the root cause and a same-day fix instead of spending days chasing the source. In regulated environments, especially, less disruption means less business risk, and that is exactly where this distinction stops being theoretical.
For enterprises evaluating their options, the real test is not a vendor’s demo. It is whether the platform was designed around AI from the start, because that is what determines whether an issue gets caught in real time or only after someone goes looking for it.
For a consumer or retail enterprise looking to aggressively scale its digital presence over the next few years, why is a self-running foundation a vital strategic advantage rather than just an optional IT upgrade?
In retail, the impact of technology disruptions is often immediate. A failed transaction can lead to a poor customer experience and, in some cases, affect customer trust. As retailers continue to expand their digital operations, having infrastructure that can operate reliably with minimal manual intervention is becoming increasingly important. When networks underpin everything from payments to in-store experiences, manual operations can become a constraint on scalability and efficiency rather than simply an IT concern.
What changes with a self-running foundation is what happens as the business scales. Retailers can expand into new stores and add services like digital signage, smart checkout, or edge analytics without reworking the network each time. The alternative, manually re-provisioning and troubleshooting for every new location or device, is exactly the operational drag that slows an expansion plan down.
Recently, we announced the expansion of our retail-ready portfolio to help customers improve connectivity, security, insights, and performance across their operations. The solutions enable retailers to confidently manage transactions, data, and shopping experiences across their entire retail environment, from warehouses and back offices to storefronts and curbside pickup. At NRF 2026, HPE demonstrated how these capabilities come together by combining the self-driving networks of HPE Aruba Networking CX switching and Mist AIOps at the branch edge with the latest HPE Nonstop solutions at the core, demonstrating how secure, resilient connectivity underpins retail modernisation.
The strategic case is simple. A retail business that treats network autonomy as core infrastructure, not an optional upgrade, is refusing to let IT become the constraint on how fast it can grow. The businesses that get this right are not the ones with the flashiest AI strategy on paper. They are the ones whose foundation can absorb new stores, devices, and services without being rebuilt every time ambition increases.