HPE says its self-driving network is no longer a lab demo. Rami Rahim, executive vice president, president and general manager of Networking at Hewlett Packard Enterprise, told reporters on October 12, 2026 that the company’s AI-driven network stack now resolves 70 to 80 percent of potential network issues without a single human touching a ticket. The claim, first reported by The Register, puts a hard number on a trend that developer and IT-operations teams have been circling for years: software that fixes itself before anyone opens a dashboard.
For engineers who already lean on AI coding agents to write and ship code, HPE’s announcement is a reminder that the same autonomous-agent logic reshaping developer tooling is now running production infrastructure. The company’s pitch rests on three platforms working together: HPE Mist, HPE Aruba Central, and the Marvis AI engine, which sits at the center of the automated remediation loop. HPE announced autonomous agentic AIOps actions for Mist and Aruba Central in May 2026 and says it is first with fully autonomous networking. Rahim’s 70-80 percent figure, he said, comes from logging every fix the system makes on its own and comparing it against the total ticket volume.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What HPE Actually Announced
According to Rahim, HPE’s self-driving network can detect, diagnose, and resolve a defined set of recurring problems in real time, without a network engineer ever getting paged. The issue types named in the announcement are specific: poor performance, slow connections, roaming failures, and DNS problems that keep client devices from connecting to the network in the first place. These are the bread-and-butter tickets that eat up the bulk of a typical network operations center’s day, and HPE is betting that automating them away is where the immediate return on investment sits.
The mechanics matter here. Rahim described an operational sequence where the Marvis AI engine receives customer service tickets first, before a human ever sees them. From there, Marvis either resolves the issue outright or, if it cannot, it learns from the resolution path a human takes so that the same category of problem can be handled automatically the next time it appears. That learning loop is the part of the story that should interest coding teams: it is functionally the same reinforcement pattern used by autonomous coding agents that improve their suggestions based on accepted or rejected pull requests.
Rahim put the current state of the system plainly. “I think we’re now at probably around 70 to 80 percent of all tickets don’t require human intervention,” he said, according to The Register. He also offered a timeline for where he expects the technology to land next. “Within two to three years, we’ll have no issues that require humans,” Rahim said, a forecast that, if realized, would mark one of the more aggressive automation timelines any major networking vendor has put on record.
Why This Matters Beyond the Network Closet
Network operations and software development have historically run on separate tracks inside most organizations, but the line between them keeps blurring. When a self-driving network autonomously resolves a DNS misconfiguration before it takes down a staging environment, that is effectively the same category of labor-saving automation that a coding agent delivers when it catches a null-pointer bug before a pull request merges. Both depend on an AI system that has been trained on enough historical incident data to recognize a pattern, act on it, and log the outcome for the next iteration.
That overlap is why HPE’s announcement lands squarely in conversations happening across the developer-tools market. Teams evaluating coding agents like Claude Code, Codex, and GitHub Copilot are asking the same fundamental question infrastructure buyers are asking about Marvis: how much of this can I trust to run unsupervised, and what is the failure mode when it gets something wrong? HPE’s 70-80 percent figure is notable precisely because it is a measured number, not a marketing ceiling. It implies that roughly one in four or one in five tickets still needs a human, and that gap is where trust, auditability, and escalation paths still matter most.
The timing also lines up with a broader capital story. Microsoft, Meta, and Google have pushed combined AI infrastructure spending past $725 billion this year, and a meaningful share of that spend goes toward the networking fabric connecting GPU clusters. A network that can self-diagnose and self-heal at scale is not a nice-to-have in that environment; it is close to a requirement, since the ticket volume generated by tens of thousands of interconnected nodes would overwhelm any human-staffed NOC.
HPE’s Self-Driving Network vs. Other Autonomous IT Systems
HPE is not alone in chasing the “no human in the loop” promise. Microsoft has leaned on its own automation research, including the Decision-1 model it says runs dramatically faster than comparable systems, to push autonomous decision-making deeper into enterprise operations. The comparison below lines up what is publicly confirmed about each effort, without extrapolating beyond what each company has stated.
| System | Vendor | Primary function | Confirmed automation claim | Human-in-loop status |
|---|---|---|---|---|
| HPE Self-driving Network (Marvis AI engine) | Hewlett Packard Enterprise | Network issue detection, diagnosis, remediation | 70-80% of tickets resolved without a human, per Rahim | Remaining 20-30% still escalated to engineers |
| HPE Aruba Central / HPE Mist | Hewlett Packard Enterprise | Network management platform hosting the AI engine | Real-time detection and resolution, per HPE | Integrated with Marvis learning loop |
| Decision-1 | Microsoft | AI decision-making model | Claimed 35x speed advantage over comparable models, per Microsoft | Not specific to network operations |
| AI coding agents (general category) | Multiple vendors | Code generation and review | Reported to add roughly 30% more code output, per industry reporting | Pull-request review still typically required |
The comparison underscores a pattern across the industry: the headline automation percentage is always paired with an acknowledgment that some share of work still needs a human reviewer. HPE’s own figure leaves room for that nuance, which is part of why Rahim’s statement was treated as credible reporting rather than dismissed as pure marketing by outlets covering the story.
The Historical Arc of “Self-Driving” Networking
The term “self-driving network” has been used in networking marketing for roughly a decade, often loosely, to describe anything from basic automated configuration pushes to more sophisticated anomaly detection. What distinguishes HPE’s current claim is the specificity of the figure and the naming of a concrete measurement method. Rahim said the 70-80 percent number comes from logging each autonomous fix, which is a more rigorous standard than the vague “AI-powered” labels that dominated networking announcements through the early 2020s.
Juniper’s 2019 acquisition of Mist, which brought Mist’s AI technology into Juniper switching and routing, laid the technical groundwork for this moment, giving HPE a machine-learning-native management layer to build on rather than retrofitting AI onto legacy network management software. HPE announced Agentic Mesh for Mist in December 2025, which it said supports anomaly detection and root-cause analysis, following the autonomous fixes for misconfigured ports, capacity, and noncompliant hardware that Mist added in August 2025. In June 2026 HPE expanded self-driving networking to CX wired switches in Mist. Aruba Central, meanwhile, has served as the operational dashboard where IT teams historically triaged alerts manually, and it gained wired-port remediation powered by Marvis AI in June 2026. HPE also launched Aruba Central On-Premises 3.0 in December 2025 with proactive remediation and AI alerts, and brought Mist’s Large Experience Model, which uses billions of Zoom and Teams data points, to Aruba Central that same month. The shift described by Rahim is essentially about removing that manual triage step for an increasingly large share of ticket types, starting with performance, connectivity, roaming, and DNS failures.
That history matters for anyone evaluating whether to trust the 70-80 percent figure. It is not a number pulled from a single quarter of data; it reflects years of incremental model training against real customer tickets across HPE’s installed base. Compare that to how Microsoft has framed its own AI trust safeguards, where the emphasis has been on building in an “emergency brake” rather than claiming full autonomy out of the gate. HPE’s framing is more confident, but it is grounded in the same industry-wide tension between automation speed and operational safety.
What This Means for IT and DevOps Teams Right Now
For network administrators managing HPE Aruba or HPE Mist deployments, the practical takeaway is that a growing share of first-line troubleshooting is being handled before a ticket reaches a human inbox. That shift changes staffing math for network operations centers in the same way autonomous coding agents have changed staffing conversations for engineering teams. Fewer tickets reaching Tier 1 support does not necessarily mean fewer jobs; it typically means those staff get reassigned to the harder 20-30 percent of cases that still require judgment, plus capacity planning and architecture work that AI systems are not yet positioned to handle.
There is also a data center capacity angle worth flagging. As AI training clusters grow denser, the network fabric connecting GPUs becomes a more frequent source of performance-degrading issues, and automated remediation becomes less of a convenience and more of an operational necessity. HPE appears to share that view: it introduced new AI-routing and data-center-networking capabilities in June 2026, including AI data-center root-cause remediation, and tied them to self-driving operations. That dynamic echoes concerns raised around FERC’s recent data center registry decision, which highlighted how much operational complexity utilities and network operators are absorbing as AI infrastructure scales. Latency-sensitive deployments, like the kind described in Eletronet’s recent IP transit upgrade, are exactly the environments where a few seconds of automated DNS remediation can prevent a cascading outage.
Security operations teams should also take note. The same self-healing logic that HPE applies to connectivity issues parallels what endpoint detection vendors are building into their platforms. The ongoing competitive jockeying among CrowdStrike, SentinelOne, and Microsoft Defender increasingly centers on how much remediation each platform can execute autonomously before a security analyst gets involved, which is the same metric HPE is now reporting for network operations.
Market Impact and Competitive Pressure
HPE’s networking business sits inside a company that has spent the past several years repositioning itself around AI infrastructure, and a credible, measurable automation claim gives HPE’s sales teams a concrete figure to put in front of enterprise IT buyers who are increasingly skeptical of vague AI marketing. Competing networking vendors, including Cisco and Juniper (now part of HPE following its 2025 acquisition), have pushed their own AI-driven operations tools, but few have attached a specific percentage to autonomous ticket resolution in the way Rahim did this week.
That specificity is likely to pressure competitors to publish comparable figures of their own, if they have them, or risk looking vague by comparison. It also puts HPE in a stronger position when bidding for large enterprise and data center networking contracts, where procurement teams are increasingly asking vendors to quantify operational savings rather than accept generic AI-enhanced messaging.
| Issue type named by HPE | Typical cause | Autonomous handling per HPE |
|---|---|---|
| Poor performance | Congestion, interference, misconfiguration | Detected and diagnosed by Marvis AI engine |
| Slow connections | Bandwidth contention, roaming delays | Flagged and resolved without a ticket escalation |
| Roaming failures | Handoff issues between access points | Automated detection and correction, per HPE |
| DNS problems | Resolution failures preventing client connections | Logged and remediated as part of the 70-80% figure |
What Remains Unconfirmed
It is worth being precise about what HPE has and has not confirmed. The 70-80 percent figure is explicitly a range, not a fixed number, and Rahim did not claim the system resolves exactly 70 percent or exactly 80 percent of issues. HPE also has not stated that this capability applies uniformly across every customer deployment; the figure reflects an aggregate measurement across HPE’s logged ticket data, and individual network environments will vary based on complexity, device density, and how long a given deployment has been running on Mist and Aruba Central. Readers should treat the capability as representative of HPE’s current reported average, not a guaranteed outcome for any specific network.
Similarly, while Rahim offered a two-to-three-year horizon for reaching a state where “no issues” require human intervention, that is a forward-looking projection from a company executive, not a confirmed roadmap with committed features or dates. Forecasts like this one are common in the networking and AI industries and should be read as directional rather than contractual.
How This Fits Into HPE’s Broader AI Strategy
HPE’s networking division has spent the past two years folding Juniper’s engineering talent and product lines into its own AI-operations roadmap, and the self-driving network announcement is the clearest public signal yet of what that combined team is shipping. Rather than treating AI as a bolt-on feature inside Aruba Central’s dashboard, HPE is positioning Marvis as the primary decision-maker for a defined category of incidents, with human engineers acting as the exception-handling layer rather than the default responder.
That strategic bet mirrors what HPE said in its own announcement describing the shift from vision to reality for autonomous networking capabilities across HPE Mist and HPE Aruba Central, first detailed on HPE’s newsroom earlier this year. The company has been explicit that this is a staged rollout tied to specific issue categories, not a blanket claim that every conceivable network fault is now automated. That staged approach is consistent with how other large infrastructure vendors have approached autonomous operations: start with high-frequency, well-understood failure modes, then expand the model’s authority as confidence in its decisions grows.
Where HPE’s Self-Driving Network Is Being Deployed
HPE has been attaching the Self-Driving Network name to named customer deployments. On September 29, 2026, Saudi German Health UAE selected HPE’s Self-Driving Network portfolio for wired, wireless, and network-access-control modernization, managed through HPE Aruba Central. On October 5, 2026, HPE announced that it will deploy a Self-Driving Network and Connected Intelligence Center for the 2027 Ryder Cup. The planned network will use Wi‑Fi 7 and continuously monitor, adapt, and optimize user experience.
HPE says the Ryder Cup deployment is intended to prevent issues before they occur, but the announcement does not state a 70–80% issue-resolution rate. Rahim’s range should therefore be read as a separate claim from these customer announcements. On October 19, 2026, HPE and the Miami HEAT announced a self-driving-network deployment at Kaseya Center using Aruba wireless and Juniper switching. HPE’s own pages also say the world’s largest almond manufacturers use HPE Aruba Networking Central intelligence to resolve Wi‑Fi issues 50% faster, which is a customer result and not the same measurement as the ticket range.
HPE is converging technology beneath the platforms rather than renaming Mist as Aruba Central, so buyers should expect both names to keep appearing in product material. HPE is also positioned furthest in “Completeness of Vision” and highest for “Ability to Execute” in the 2026 Magic Quadrant™ for Enterprise Wired and Wireless LAN Infrastructure by Gartner® for the fifth consecutive time. For an IT team weighing an expansion, these deployments and analyst placements are useful context, but the most practical step is still to ask for ticket-resolution data from a deployment that resembles your own network.
Industry Reaction and What Comes Next for IT Buyers
Reaction to Rahim’s comments has centered less on skepticism of the technology and more on questions about measurement transparency. Because HPE is calculating its 70-80 percent figure from internally logged autonomous fixes, independent verification of the exact methodology is not yet publicly available outside of what Rahim described to reporters. That has not stopped the figure from circulating widely across network-operations and IT-management circles, in part because it is specific enough to be testable by any enterprise customer running Mist or Aruba Central in production.
For IT buyers evaluating whether to expand their HPE networking footprint, the practical next step is asking account teams for deployment-specific ticket-resolution data rather than relying on the company-wide average. A network with heavy legacy hardware or unusual topology may see a lower autonomous-resolution rate than HPE’s aggregate figure, while a greenfield deployment built natively on Mist and Aruba Central from day one may see a higher one. That variance is normal for any AI system trained on operational data, and it is the same caveat that applies when evaluating autonomous coding tools against a vendor’s headline benchmark scores.
Predictions: Where Self-Driving Networks Go From Here
- Expect competing networking vendors to publish their own autonomous-resolution percentages within the next two to three quarters, following HPE’s lead in quantifying the metric publicly.
- The gap between the current 70-80 percent figure and Rahim’s “no issues require humans” target will likely close unevenly, with simple connectivity and DNS issues automated first and complex multi-site roaming or security-adjacent incidents remaining human-reviewed the longest.
- Enterprise buyers will increasingly request autonomous-resolution benchmarks as a standard line item in networking RFPs, mirroring how AI coding-assistant adoption rates have become a procurement criterion for developer tooling.
- Integration between network-operations AI (like Marvis) and security-operations AI (like the tools used by CrowdStrike, SentinelOne, and Microsoft Defender) will deepen, since both categories are converging on the same autonomous-remediation playbook.
- As AI data center buildouts accelerate under the kind of capital commitments seen from Microsoft, Meta, and Google, network automation will shift from a cost-saving feature to a scaling necessity, since human-staffed NOCs cannot keep pace with ticket volume generated by GPU-dense clusters.
The Bigger Picture for Developer and IT Automation
HPE’s announcement is a data point in a much larger story about where autonomous software is actually delivering measurable results today, as opposed to where it remains aspirational. Rahim’s statement that “the self-driving network is no longer aspirational; it’s operational,” made during HPE’s broader announcement of autonomous networking capabilities, captures the tone the company wants enterprise buyers to take away: this is infrastructure already running, not a future promise.
That framing matters for how IT leaders think about budget allocation heading into 2027 planning cycles. If a self-driving network genuinely resolves 70-80 percent of tickets without human intervention, the return on investment calculation for adopting or expanding HPE Mist and Aruba Central deployments becomes considerably easier to model than it would be for a less quantified automation claim. Rahim also noted that the shift “fundamentally changes the role of networking from a system that informs to one that takes action on behalf of the business, freeing customer networking teams to focus on innovation instead of operations,” a line that echoes almost exactly the pitch coding-agent vendors have made to engineering leaders over the past two years.
Whether HPE hits its two-to-three-year target for eliminating human-required tickets entirely remains to be seen, and that gap between the current 70-80 percent and a theoretical 100 percent is likely to be the hardest part of the curve. The easy automation wins, the recurring, pattern-matched issues like DNS failures and roaming handoffs, tend to get solved first. What is left over skews toward novel, ambiguous, or high-stakes incidents that are exactly the kind of problems AI systems still struggle to resolve without oversight, whether in network operations or in code review.
Frequently Asked Questions
What is HPE’s self-driving network?
It is HPE’s branding for an AI-managed networking stack built on HPE Mist and HPE Aruba Central, powered by the Marvis AI engine, which detects, diagnoses, and resolves network issues automatically.
How much of network troubleshooting does HPE’s AI actually handle?
Rami Rahim, HPE’s networking chief, said the system resolves 70 to 80 percent of potential issues without requiring a human to step in, based on logged data of autonomous fixes.
What kinds of problems does the self-driving network fix on its own?
HPE named four specific categories: poor performance, slow connections, roaming failures, and DNS problems that prevent devices from connecting.
What is the Marvis AI engine?
Marvis is the AI engine at the center of HPE’s self-driving network. It receives customer service tickets first, resolves what it can automatically, and learns from human-resolved tickets to improve future automated fixes.
When does HPE expect to fully eliminate human-required network tickets?
Rahim said he expects the system to reach a point within two to three years where no issues require human intervention, though this is a forward-looking projection rather than a confirmed roadmap.
Is HPE’s self-driving network the same as HPE Aruba Central?
No. HPE Aruba Central and HPE Mist are the management platforms where the AI capabilities run; the self-driving network and Marvis AI engine are the automation and remediation layer built on top of them.
Does this automation apply to every HPE network deployment?
HPE has not confirmed that the 70-80 percent figure applies uniformly across all deployments. It reflects an aggregate measurement across HPE’s logged customer ticket data, and results can vary by network complexity and deployment maturity.
How does this compare to AI automation in software development?
The underlying pattern is similar: both network automation and AI coding agents rely on models trained on historical resolution data to act autonomously on routine issues while escalating harder cases to humans. Coding tools like Claude Code, Codex, and GitHub Copilot follow a comparable human-in-the-loop structure.
