HPE Self-Driving Network Fixes 70-80% of Issues [2026]

HPE says its networking software is now closing out the vast majority of trouble tickets without a human ever touching a keyboard. Rami Rahim, Executive Vice President, President & General Manager of Networking at HPE, told Fierce Network that the company’s self-driving network capabilities now detect, diagnose, and resolve between 70% and 80% of all potential network issues entirely on their own, logging the fix and notifying IT staff only after the problem is already gone.

The claim, confirmed across multiple outlets including The Register, Autoblog, Carscoops, Automotive News, and Barron’s, lands at a moment when enterprise IT buyers are being pitched autonomous operations software from nearly every major infrastructure vendor. HPE’s pitch is specific: its Marvis AI engine, running across HPE Mist and HPE Aruba Central, now handles roaming failures, DNS problems, slow connections, and generic performance degradation without opening a ticket for a human to pick up. For software engineers who have spent the last two years watching AI agents creep into code review, CI pipelines, and cloud cost management, this is the same pattern showing up in a different layer of the stack: the network itself.

Google · Preferred Sources

Don't miss new tech stories on Google

Add TrendinTech once in the Google app and our stories appear in your news suggestions.

Add Now

What HPE Actually Announced

The 70-80% figure is not a lab benchmark or a marketing projection pulled from a slide deck. According to Rahim, HPE arrives at the number by logging every autonomous fix the system makes and comparing that tally against the total volume of issues the network detects. “I think we’re now at probably around 70 to 80 percent of all tickets don’t require human intervention,” Rahim said, a figure that, if it holds across HPE’s enterprise customer base, represents a meaningful shift in how IT departments staff and budget for network operations.

The technology underpinning the claim is not brand new. HPE Mist and HPE Aruba Central have offered AI-driven network visibility for several years, and the Marvis AI engine has been HPE’s branded assistant for network troubleshooting since well before the current wave of generative AI hype. What changed, according to HPE’s own announcement, is the shift from passive diagnosis (telling a human what’s wrong) to active, trusted autonomous remediation (fixing it directly, then reporting afterward). That distinction is the entire story here. A dashboard that tells an engineer “DNS resolution is failing on subnet X” is a monitoring tool. A system that silently rewrites the DNS configuration, confirms client connectivity is restored, and only then logs the ticket as closed is something categorically different, and it is the latter that HPE is claiming to run in production today.

HPE’s own announcement frames the shift bluntly: the company describes self-driving networks as no longer aspirational but operational. That framing matters for how the market should read this news. Vendors have talked about “self-healing networks” as a future-state vision for close to a decade. HPE is now arguing that future state has arrived for a defined slice of enterprise network operations, not as a roadmap item but as something customers are running today.

The Issue Types Marvis Handles Without a Human

According to reports, the categories of problems HPE’s self-driving network resolves autonomously are relatively narrow but cover a large share of real-world helpdesk volume. They include poor performance, slow connections, roaming failures (the dropped or degraded connection that happens when a device moves between access points), and DNS problems that prevent clients from connecting in the first place. These four categories, anyone who has worked an enterprise IT helpdesk will recognize immediately, make up a disproportionate share of day-to-day ticket volume. They are also, not coincidentally, the categories of problem that are most amenable to automated detection: they produce clear telemetry signatures, they have a relatively small set of root causes, and the corrective action (reconfigure a channel, flush a cache, adjust a roaming threshold, correct a DNS entry) is usually well-defined once the root cause is identified.

That narrowness is worth sitting with, because it is the detail most likely to get lost in headline coverage of the story. HPE is not claiming its AI can resolve every conceivable network fault autonomously. Novel failure modes, hardware faults, security incidents, and anything requiring a change to physical topology or a capital purchase decision almost certainly still land on a human’s desk. The 70-80% figure describes a subset of total network operations work, not all of it. Still, if that subset previously consumed the majority of a network operations team’s reactive time, automating it changes what those teams spend their day doing, from firefighting to planning.

A Real Customer Number: The UK Ministry of Justice

Rahim’s 70-80% figure is HPE’s own aggregate claim across its installed base, but the company has also published a specific, named customer result that offers a second data point. According to HPE, the UK Ministry of Justice achieved approximately 75% fewer service desk tickets after adopting HPE Self-driving Network capabilities. The Ministry of Justice is quoted in HPE’s own materials describing the system as enabling trusted autonomous actions that help the organization anticipate and resolve issues before users are impacted.

That the Ministry of Justice’s reported reduction (roughly 75%) sits squarely inside Rahim’s broader 70-80% range is a useful sanity check on the headline figure. It suggests the company-wide number is not simply a best-case anecdote dressed up as an average. It also gives enterprise IT buyers a second, independently attributable reference point beyond the vendor’s own aggregate claim, which matters in a market where self-reported automation statistics are common and rarely audited by a third party.

Why This Matters Beyond the Networking Team

Software engineers and platform teams who don’t directly manage switches and access points still have a stake in this story, for three reasons. First, network reliability is the invisible floor underneath every other automation claim in the enterprise stack. A CI/CD pipeline, a Kubernetes cost-monitoring dashboard, or a cloud-based build system is only as reliable as the network connecting engineers to it, and DNS failures or roaming drops are a disproportionately common cause of “why is my deploy stuck” tickets that land on platform engineering teams rather than networking teams. Second, the architecture HPE describes, continuous telemetry collection, an AI reasoning layer, and autonomous closed-loop remediation, is structurally identical to the agentic patterns already appearing in infrastructure-as-code tools like Terraform-driven multi-cloud deployments. Third, this is one more data point in a broader 2026 trend: AI agents are being handed increasing autonomy to execute changes directly in production environments, not just suggest them.

That last point connects directly to a theme running through enterprise AI coverage all year: the move from AI-as-advisor to AI-as-operator. Decision-automation systems built to score and execute choices rather than merely flag them for human review are showing up across the stack, and HPE’s networking pitch is effectively the same philosophy applied to packet loss and DNS resolution instead of business logic.

AIOps in Context: How HPE’s Claim Compares

HPE is far from the only vendor selling autonomous network operations software, and the broader AIOps category (a term coined by Gartner years before generative AI made “agentic” a buzzword) includes competing platforms from Cisco, Juniper (now part of HPE itself, following HPE’s 2025 acquisition), VMware, and a wave of startups focused narrowly on network observability. What distinguishes HPE’s claim from most competitor marketing in this space is specificity: a named executive, a defined percentage range, a stated measurement method (logged autonomous fixes against total detected issues), and a named reference customer. Much of the AIOps marketing circulating in 2026 trades in vaguer language, “AI-powered insights” or “intelligent automation”, without committing to a number a reporter can later check against a customer case study.

The table below places HPE’s reported figures alongside the general categories of autonomous network claims the broader industry has made public, drawing only on the reporting and HPE’s own published statements referenced in this story.

Metric	Figure	Source
Share of network tickets resolved without human intervention	70-80%	Rami Rahim, HPE, via Fierce Network
UK Ministry of Justice service desk ticket reduction	Approximately 75%	HPE customer case study
Issue categories handled autonomously	Poor performance, slow connections, roaming failures, DNS problems	Reports citing HPE
Core products involved	HPE Mist, HPE Aruba Central, Marvis AI engine	HPE
Rahim’s stated measurement method	Logging each autonomous fix against total tickets	Fierce Network interview
Confirmed future target of 100% automation	Not confirmed by HPE on the record in this reporting	N/A

The Historical Arc: From Manual Troubleshooting to Marvis

To understand why this announcement matters, it helps to walk through how enterprise network troubleshooting actually used to work, and still does at most organizations that haven’t adopted an AIOps platform. A user reports a slow or dropped connection. A helpdesk ticket gets filed. A tier-1 support technician triages it, often by asking the user to restart their device or reconnect to Wi-Fi, before escalating to a networking specialist who pulls logs from switches, access points, and DNS servers by hand, correlates timestamps, and manually identifies a root cause. That process, even when well-staffed, routinely takes hours. At under-staffed organizations, especially in the public sector, it can take days.

HPE’s Mist platform, which the company built out through acquisitions over the past several years, replaced much of that manual log correlation with continuous telemetry collection and machine learning-based anomaly detection years before the current conversation about agentic AI began. Marvis, HPE’s branded virtual network assistant, started as a natural-language interface for querying that telemetry, essentially letting an engineer ask “why is this user’s Wi-Fi slow” and get an AI-generated answer instead of manually cross-referencing logs. The shift Rahim is describing now, autonomous remediation rather than just autonomous diagnosis, is the next rung on that ladder, and it mirrors a broader shift happening across enterprise software this year, where AI systems that once only summarized problems are increasingly being trusted to act on them directly.

The table below contrasts the traditional manual troubleshooting workflow with the autonomous process HPE describes for the issue categories Marvis now handles.

Workflow Step	Traditional Manual Process	HPE Self-Driving Network
Detection	User reports issue; ticket filed with helpdesk	Continuous telemetry monitoring flags anomaly automatically
Triage	Tier-1 support technician manually reviews report	Marvis AI engine correlates telemetry patterns in real time
Root-cause diagnosis	Specialist pulls logs from switches, access points, DNS servers by hand	AI reasoning layer identifies root cause from aggregated telemetry
Remediation	Engineer manually applies configuration change	System executes the fix autonomously (for the four covered issue types)
Notification	Ticket updated and closed after human confirms fix	IT staff notified only after the fix is already applied
Human involvement required	Every step	Only for the roughly 20-30% of cases outside the automated categories

Rahim has described the underlying philosophy in plain terms: “A self-driving network is a network that configures, optimizes and heals itself,” he said, adding that “it is run not by humans, but by AI agents and software that oversee the day-to-day operations of the entire network and every device connected to it, collecting telemetry and proactively making changes to ensure the user experience is always amazing.” That framing, an AI layer that acts as the operator rather than an advisor sitting beside one, is the throughline connecting this networking announcement to the rest of 2026’s enterprise automation push.

What HPE Has Not Confirmed

It is worth being precise about the boundaries of HPE’s claim, because autonomous-AI announcements tend to get rounded up in secondary coverage. HPE has not confirmed, on the record in the reporting reviewed for this story, that its system fixes all network issues, that it has reached or will imminently reach 100% automation, or a specific committed date by which every remaining human-required issue will be eliminated. The 70-80% figure is a current-state operational claim, not a forecast with a committed deadline. Any reporting suggesting a firm, dated path to full automation should be treated as speculative unless HPE issues a separate, specific statement confirming it.

It is worth noting that Rahim himself has gone further in describing his own expectations for where the technology is headed, telling The Register, “Within two to three years, we’ll have no issues that require humans.” That is Rahim’s personal projection about the trajectory of the technology, not a confirmed HPE product roadmap or a guaranteed delivery date, and the distinction between an executive’s forecast and a company’s committed milestone matters here.

That distinction matters for IT buyers evaluating the technology, and it matters for how seriously the rest of the industry should take the 70-80% figure itself. A vendor claiming to have already automated the large majority of routine ticket volume, while being explicit that a meaningful remainder still requires human judgment, is a more credible and more useful claim than a vendor promising imminent full automation. It also lines up with how AI deployment has generally gone across other categories of enterprise software this year: strong gains on well-defined, high-volume, pattern-matchable problems, with the long tail of novel or ambiguous cases still requiring a person.

Market Impact for HPE and Rival Vendors

HPE’s networking division has been under pressure to prove the value of its 2025 Juniper acquisition, and autonomous-operations claims like this one are precisely the kind of differentiator the combined company needs to justify premium pricing against Cisco in enterprise bake-offs. A verifiable reduction in service desk ticket volume is a number that resonates directly with CIOs and IT budget owners in a way that abstract “AI-powered” marketing language does not: fewer tickets means fewer support staff hours, faster mean-time-to-resolution, and a more defensible ROI case for the upfront cost of deploying Mist and Aruba Central hardware and licensing across an enterprise campus or branch network.

For competitors, the pressure is now to match or beat a specific, attributable number rather than a vague promise. Cisco, which has invested heavily in its own AI-driven networking assistants, and smaller AIOps-focused vendors will likely face questions from analysts and prospective customers about whether they can produce a comparable figure, backed by a comparable measurement method and a named reference customer. That shifts the competitive conversation in enterprise networking away from feature checklists and toward outcome metrics, a shift that mirrors what has already happened in adjacent enterprise software categories like cloud infrastructure pricing, where buyers increasingly demand hard numbers rather than marketing language.

What Enterprise IT Teams Should Actually Do With This News

For IT leaders evaluating whether to invest in autonomous network operations software, the HPE announcement offers three practical takeaways rather than a single “adopt this now” directive. First, ask any vendor making an autonomous-remediation claim for their specific measurement methodology, the same way Rahim described logging each autonomous fix against total ticket volume. A vendor that cannot explain how its percentage was calculated is not offering a comparable number. Second, look for named customer case studies with specific percentage reductions, not just logo walls. The UK Ministry of Justice figure is useful precisely because it is attributable and specific. Third, recognize that the categories of issue being automated (DNS, roaming, performance degradation, slow connections) are a well-understood subset of network operations, and evaluate any vendor’s claim against whether it covers the issue types that actually dominate your organization’s own ticket volume.

Organizations already running infrastructure-as-code pipelines through tools like Terraform, or managing container costs through platforms discussed in coverage of ECS versus EKS cost tradeoffs, should see this as part of a continuum rather than an isolated networking story. The operational playbook, instrument everything, let an AI layer detect and diagnose first, then gradually extend trust to autonomous remediation, is the same playbook showing up across cloud cost management, CI/CD, and now core network operations.

Five Predictions for Autonomous Networking Through 2027

Expect at least one major rival (Cisco, Juniper’s remaining independent product lines, or a VMware-descended networking vendor) to publish a comparable autonomous-resolution percentage within the next two quarters, under competitive pressure from HPE’s public number.
The 70-80% figure will likely climb modestly, but reaching a genuine 90%+ range will require automating categories beyond DNS, roaming, and performance issues, such as security-adjacent misconfigurations, which carry higher risk tolerance requirements for autonomous action.
Public sector adoption, following the UK Ministry of Justice’s lead, will likely accelerate faster than private enterprise adoption in some regions, since government IT departments face more acute staffing shortages relative to ticket volume.
Scrutiny of self-reported automation percentages will increase, with analyst firms and procurement teams pushing vendors toward third-party-audited or at least methodologically transparent metrics rather than vendor-calculated aggregates.
The line between “network operations automation” and general enterprise AIOps will continue to blur, as the same agentic patterns, telemetry plus reasoning plus autonomous action, get applied across networking, cloud cost management, and application performance monitoring under a shared platform narrative.

The Bigger Picture: Autonomous IT Operations in 2026

HPE’s self-driving network claim arrives alongside a broader industry pattern where enterprise software vendors across categories are racing to demonstrate that their AI layers can act, not just advise. The same quarter that produced this networking announcement also saw decision-scoring AI models designed to execute choices rather than merely flag them, and continued momentum behind infrastructure-as-code tooling that lets a handful of engineers manage deployments across multiple cloud providers simultaneously. Networking has historically been one of the more conservative corners of enterprise IT when it comes to handing control to automated systems, given the direct and immediate impact a bad autonomous change can have on every other system riding on top of it. HPE’s willingness to put a specific percentage on the record, backed by a named customer result, signals a level of institutional confidence in these systems that would have been unusual even two years ago.

Whether that confidence is fully earned will depend on how the 70-80% figure holds up as more customers deploy the technology at scale, and on whether HPE or any competitor can responsibly extend autonomous action into higher-risk categories like security incident response without a corresponding increase in costly mistakes. For now, the figure stands as the most specific, attributable claim any major networking vendor has made about autonomous operations in 2026, and it is likely to become the benchmark the rest of the industry gets measured against.

Frequently Asked Questions

What is HPE’s self-driving network?

HPE’s self-driving network is the company’s branded set of autonomous networking capabilities, built on HPE Mist, HPE Aruba Central, and the Marvis AI engine, designed to detect, diagnose, and resolve network issues in real time without requiring a human operator to intervene.

How much of network troubleshooting does HPE’s system automate?

According to Rami Rahim, Executive Vice President, President & General Manager of Networking at HPE, the system now resolves between 70% and 80% of all potential network tickets without human intervention, based on logging each autonomous fix against the total volume of detected issues.

What types of network issues can it fix on its own?

Reports citing HPE indicate the autonomous system handles poor performance, slow connections, roaming failures, and DNS problems that prevent clients from connecting.

Has any real customer confirmed a result from this technology?

Yes. According to HPE, the UK Ministry of Justice achieved approximately 75% fewer service desk tickets after adopting HPE Self-driving Network capabilities, a figure that aligns closely with Rahim’s broader 70-80% claim.

Did HPE say the network will eventually reach 100% automation?

HPE has not confirmed on the record, in the reporting reviewed for this story, that the system will reach full automation or eliminate all human-required issues within a specific stated timeframe. Readers should treat any claim of guaranteed 100% automation as unconfirmed.

Is HPE’s self-driving network the same as AIOps?

HPE’s self-driving network capabilities fall under the broader AIOps category, a term used across the industry to describe AI-driven monitoring, diagnosis, and remediation of IT operations issues, applied specifically to enterprise networking infrastructure.

Why does this matter to software engineers who don’t manage networks directly?

Network reliability underlies every other layer of enterprise software operations, from CI/CD pipelines to cloud deployments, and the same agentic automation pattern HPE describes (telemetry, reasoning, autonomous action) is already showing up in infrastructure-as-code and cloud cost management tooling that engineers use directly.

Which products power HPE’s autonomous networking claims?

The capabilities run across HPE Mist and HPE Aruba Central, powered by HPE’s Marvis AI engine, according to HPE’s own announcements.

HPE’s Self-Driving Network Fixes 70-80% of Issues [2026]