The Self-Driving Network: How AI Agents Are Rewiring Global Telecom

• Introduction
• Chapter 1 The World's Largest Machine: Inside the Fragile Giant
• Chapter 2 The Human Bottleneck: Midnight Pagers and the Cost of Downtime
• Chapter 3 The Bell Labs Dream: Early Visions of Autonomous Switching
• Chapter 4 From Copper to Code: Software-Defined Networking and the First Shift
• Chapter 5 The 5G Explosion: When Complexity Outgrew the Human Brain
• Chapter 6 The Cloud Invaders: Hyperscalers Knock on the Telco Door
• Chapter 7 Espoo’s Gambit: Nokia’s Bid to Reinvent Network Intelligence
• Chapter 8 Redmond’s Ambition: Microsoft Azure and the Operator Cloud
• Chapter 9 The Rivalry: Ericsson, Huawei, and the Geopolitics of Connectivity
• Chapter 10 From Scripts to Models: The Evolution of Algorithmic Control
• Chapter 11 The Rise of the Agent: Defining Autonomous Intent in Telecom
• Chapter 12 Digital Twins: Simulating the Global Grid in Silicon
• Chapter 13 The Open RAN Revolution: Disaggregating the Network Stack
• Chapter 14 The Zero-Touch Dream: ETSI, Standards, and the Battle for Protocols
• Chapter 15 Power and Silicon: AI at the Edge of the Radio Network
• Chapter 16 The Hallucination Hazard: When AI Agents Make High-Stakes Mistakes
• Chapter 17 War Games in the Core: Cybersecurity and Self-Defending Infrastructure
• Chapter 18 The Green Algorithm: Slashing Megawatts with Dynamic Power Tuning
• Chapter 19 Field Trials and Friction: Cultural Clashes Inside Legacy Operators
• Chapter 20 Subsea and Orbit: Extending Agent Autonomy Beneath Oceans and Into Space
• Chapter 21 The Great Talent War: Telecom Engineers vs. Silicon Valley
• Chapter 22 Sovereignty, Security, and Regulation: Who Governs Autonomous Signals?
• Chapter 23 The Economic Fallout: Workforce Disruption in the Automated Telco
• Chapter 24 The Unblinking Eye: Real-Time Self-Healing Across Continents
• Chapter 25 The Fully Autonomous Horizon: What Happens When the Network Runs Itself

Every second of every day, humanity breathes life into the largest and most intricate machine ever constructed. It is not the International Space Station, nor the Large Hadron Collider, nor the electrical power grids that span continents. It is the global telecommunications network: a labyrinthine, planet-wrapping nervous system composed of millions of cell towers, tens of thousands of data centers, hundreds of undersea fiber-optic cables, and constellations of low-Earth orbit satellites. Through this colossal web of silicon, glass, and radio waves flows nearly every modern human endeavor—financial markets clearing trillions of dollars in milliseconds, robotic surgical units operating across borders, power grids balancing volatile renewable energy, and billions of people reaching out to connect across oceans. Yet beneath the seamless gloss of our digital lives lies a dirty secret known to every systems engineer from Tokyo to Frankfurt: the world’s largest machine has become too complex for human beings to operate.

For more than a century, the stewardship of this infrastructure relied on human cognition, duct tape, and institutional memory. When a storm severed a fiber trunk in the Midwest, or a configuration error knocked a European national carrier offline, the response was stubbornly analog. Alarms blared in cavernous Network Operations Centers (NOCs), pagers buzzed on nightstands at three in the morning, and bleary-eyed engineers gathered on frantic conference calls to parse arcane logs, debate command-line parameters, and cross their fingers as they pushed updates into the dark. But the rollout of fifth-generation (5G) mobile infrastructure, the disaggregation of hardware into software-defined layers, and the explosive proliferation of connected devices broke that model forever. A modern telecommunications network is no longer a collection of predictable copper switches; it is a hyper-dynamic, hyper-distributed living ecosystem generating billions of state changes per second. The human brain, with its evolutionary limits on processing speed and working memory, has become the ultimate bottleneck.

This crisis of complexity has triggered the most consequential, high-stakes technological transformation since the dawn of the internet: the race to build the self-driving network. Long treated as a distant, theoretical dream born in the storied halls of Bell Labs, autonomous networking has shifted from academic speculation into an existential survival strategy. The vision is deceptively simple: transform an infrastructure managed by armies of technicians into an autonomous, self-healing, self-optimizing organism driven by artificial intelligence. In this new paradigm, networks are governed not by brittle static scripts, but by cognitive AI agents capable of understanding high-level business and operational intents, diagnosing micro-anomalies before outages manifest, balancing massive electrical loads on the fly, and re-architecting their own routing topologies in fractions of a second—all without a single keystroke from a human hand.

Behind this shift lies a fierce, quietly waged geopolitical and corporate war. On one side stand the historic monarchs of telecom equipment—European industrial heavyweights like Nokia and Ericsson, and China’s embattled titan Huawei—each racing to redefine their identity from hardware builders to software and algorithmic intelligence providers. On the other side are the cloud hyperscalers, spearheaded by Microsoft Azure and Amazon Web Services, who view telecom not merely as a customer base, but as the final, unconquered frontier of the enterprise cloud. By marrying cutting-edge foundation models with carrier-grade radio and core infrastructure, these titans are competing to capture the digital operating system of the physical world. The prize is not merely market share; it is the ultimate gatekeeping authority over the planetary flow of information.

Yet the road to full autonomy is fraught with peril. The stakes of automating global telecommunications are fundamentally different from deploying a customer-service chatbot or generating synthetic images. When an autonomous telecom agent hallucinates, miscalculates an impedance vector, or engages in unintended algorithmic feedback loops, the casualty is not a mangled paragraph—it is 911 emergency services dropping offline, air traffic management losing signal, and banking networks freezing across an entire hemisphere. Operators are forced to navigate profound questions of safety, sovereignty, and trust. How do you govern a self-defending core that can outwit state-sponsored cyber attackers at machine speed, when its own human creators no longer fully grasp the mathematical rationale behind its decisions? What happens to the millions of engineers, line workers, and system administrators whose entire careers were built on manual mastery of the dial tone?

The Self-Driving Network is the inside story of this unprecedented technological inflection point. Drawing on rare access to executive suites, research laboratories, and frontline operations across Espoo, Redmond, Kista, and beyond, this book demystifies the invisible architecture that keeps our civilizations running and uncovers the titanic struggle to automate it. It is a chronicle of audacious technical ambition, brutal cultural friction within legacy monopolies, and the profound economic reallocation of power between telecom stalwarts and Silicon Valley. Above all, it is an exploration of what happens when the most critical machine on Earth wakes up, opens its eyes, and begins to run itself.

On a desolate stretch of coastline near Bude, a small seaside town in North Cornwall, England, the Atlantic Ocean crashes endlessly against jagged cliffs. To a holidaymaker walking the coastal paths, the landscape feels wild, ancient, and entirely untamed. Yet just beneath the surface of the tide, buried under feet of wet sand and protective rock armor, lies a set of thin steel-reinforced tubes. Inside these tubes are strands of glass no thicker than a human hair. Through those strands, pulses of infrared light flash hundreds of billions of times every second, carrying bank transfers, medical imaging, video streams, military communications, and encrypted text messages between Europe and North America.

This spot in Cornwall is not unique. Similar scenes play out in coastal alcoves across the globe: in Halifax, Nova Scotia; in Fort Lauderdale, Florida; in Chiba, Japan; in Djibouti; in Mumbai; and in the suburbs of Marseille. These coastal landing stations are the physical sutures binding the continents together. They are the arterial nodes of an infrastructure so vast, so pervasive, and so continuously active that humanity has largely forgotten it exists until it breaks.

We routinely speak of "the cloud" as if it were a ethereal, weightless medium floating somewhere in the upper atmosphere. In reality, the cloud is heavy, hot, grounded, and intensely physical. It is made of concrete, diesel generators, undersea cables, glass fibers, copper wires, silicon chips, steel towers, and radio frequencies. Together, these elements form the global telecommunications network—a single, interconnected system that spans every continent, reaches into low-Earth orbit, and touches the ocean floor.

By any physical or operational metric, this network is the largest machine ever constructed by human beings. It contains more discrete moving parts, covers more geographical area, consumes more energy, and processes more state changes per millisecond than any other system in existence. The global electrical grid may span vast landmasses, but it operates at a comparatively sluggish frequency of 50 or 60 hertz. Transport networks—highways, railways, shipping lanes—are constrained by the speed of physical engines and the endurance of human operators. The telecommunications network, by contrast, operates at the speed of light, continuously balancing billions of competing data streams across millions of nodes without stopping for a single microsecond.

Yet despite its staggering scale and apparent ubiquity, this planetary machine is extraordinarily fragile. It exists in a state of perpetual, controlled degradation, held together by a brittle assembly of physical hardware, physical constraints, and delicate mathematical balances.

To understand why the world's leading technology companies and telecom operators are engaged in a multi-billion-dollar race to hand control of this system over to artificial intelligence, one must first step inside the machine itself and understand how it was built, how it works, and why its sheer physical reality makes it an engineering nightmare.

The physical anatomy of the global network can be understood as a series of concentric layers, starting at the deep-sea bed and expanding outward into the air around us.

At the deepest foundation lies the subsea cable network. Approximately 1.4 million kilometers of subsea fiber-optic cable currently carpet the ocean floors. If uncoiled, these cables would stretch to the moon and back nearly four times. These are not massive, cavernous pipelines; most subsea cables are roughly the thickness of a garden hose. At the very center of the cable sit the optical fibers—pure, synthetic silica glass engineered to such precise optical clarity that if you were to look through a block of this glass three miles thick, it would appear as clear as a clean living-room window.

Surrounding these delicate strands of glass are layers of protective casing: petroleum jelly to block water penetration, a copper tube to carry high-voltage direct current power, stranded steel wires for tensile strength, a aluminum water barrier, and a thick outer sheath of high-density polyethylene. In shallow coastal waters where commercial fishing trawlers and ship anchors pose a constant threat, the cables are further encased in heavy steel armor and buried deep into the seabed by specialized underwater plows. In the abyssal plains of the deep ocean, where human activity is nonexistent, the cables drop down to the size of a marker pen, resting quietly on the dark ocean floor under immense pressure.

Light traveling through glass does not move forever without losing energy. As photons travel along thousands of kilometers of subsea cable, they scatter and attenuate. To prevent the signal from vanishing into background noise, marine engineers install optical amplifiers—known as repeaters—every 50 to 100 kilometers along the seabed. These repeaters are massive, torpedo-shaped cylinders of beryllium copper weighing several hundred pounds each. Inside each repeater, laser diodes pump energy into short lengths of erbium-doped optical fiber. When the weakened signal light passes through the excited erbium ions, it triggers a chain reaction of stimulated emission: new photons are released at exactly the same phase and frequency as the incoming light. The signal is amplified directly in the optical domain, without ever being converted back into electrical signals.

To power these submerged amplifiers, thousands of volts of direct current are pumped down the copper layer of the cable from power stations located on the coast. A single break in that copper line, or a failure in one internal diode thousands of meters beneath the surface, can sever an entire transoceanic link, instantly starving an entire country or region of bandwidth.

Once the subsea cables make landfall, they enter landing stations—unassuming, heavily fortified concrete buildings designed to withstand natural disasters and sabotage. From there, the data flows into the terrestrial backbone network.

If the subsea cables are the transoceanic arteries of the global machine, the terrestrial backbone is its internal skeleton. This network consists of millions of miles of dark fiber laid along railway rights-of-way, buried alongside interstate highways, threaded through gas pipelines, and strung across utility poles.

To maximize the amount of information that can be pushed through a single strand of terrestrial glass, network operators rely on a technology called Dense Wavelength Division Multiplexing (DWDM). Rather than sending a single pulse of light down a fiber, DWDM uses precision lasers to transmit dozens of distinct colors, or wavelengths, of light simultaneously through the exact same glass core. Each wavelength acts as an independent channel, carrying hundreds of gigabits of data per second. Modern optical equipment can squeeze more than one hundred distinct wavelengths onto a single fiber pair, allowing a cable no thicker than a pencil to transport tens of terabits per second across continents.

This spectral complexity, however, introduces immense operational volatility. Light of different wavelengths travels at slightly different speeds through glass, a phenomenon known as chromatic dispersion. As signals travel across hundreds of miles, the sharp optical pulses spread out and begin to overlap, turning crisp digital ones and zeros into an illegible smear. Furthermore, intense laser light packed into microscopic fiber cores triggers non-linear optical effects: wavelengths interact with one another, generating interference and noise. To keep the signal readable, terrestrial networks require constant, precise manipulation using dynamic reconfigurable optical add-drop multiplexers (ROADMs)—complex optical switches that use tiny micro-electromechanical mirrors to route, boost, and flatten light frequencies on the fly.

From the high-capacity optical backbone, data flows into the nerve centers of the global network: Internet Exchange Points (IXPs) and carrier-neutral data centers.

Locations like Equinix FR2 in Frankfurt, Telehouse East in London, or Equinix Ashburn in Northern Virginia are the modern cathedrals of telecommunications. Inside these massive, windowless, climate-controlled facilities, hundreds of independent telecommunications carriers, cloud providers, content delivery networks, and enterprise networks physically interconnect their infrastructure.

Here, the network transitions from the transport of raw light to the complex logical routing of packetized data. Thousands of enterprise-grade routers—monolithic chassis the size of industrial refrigerators, consuming tens of kilowatts of power—scan every incoming packet of data, read its destination address, and make a decision about where to send it next.

These routing decisions are governed by the Border Gateway Protocol (BGP), a distributed dynamic routing protocol that effectively acts as the map-making engine of the internet. BGP allows tens of thousands of independently managed networks, known as Autonomous Systems (AS), to constantly announce which IP address ranges they can reach and at what cost.

BGP is an extraordinarily powerful protocol, but it was designed decades ago for a smaller, far more trustworthy internet. It relies heavily on implicit trust between network operators. If a single router in a small network in South America or East Asia misconfigures its BGP tables and accidentally announces that it offers the shortest path to Google's or Apple's servers, BGP will faithfully take that router at its word. Within minutes, global traffic bound for those tech giants will reroute itself, surging toward the small, unprepared network like water flooding a narrow pipe, causing massive global service disruptions. These events, known as BGP route leaks or hijackings, occur with alarming frequency, highlighting the razor-thin line between global stability and widespread failure.

The final mile of this planetary machine—the part that human beings actually interact with every day—is the Radio Access Network (RAN). This is where the fixed, glass-and-copper infrastructure gives way to the ethereal world of electromagnetic radiation.

Across the globe, millions of cellular sites sit perched on specialized steel towers, building rooftops, water tanks, utility poles, and inside athletic stadiums. Each cell site consists of a baseband unit (the computational engine that processes digital signals), remote radio heads (which convert digital data into radio frequency energy), and heavy antenna arrays that beam these radio waves through the air to mobile phones, connected vehicles, industrial sensors, and smart home appliances.

The physics of wireless communication are inherently chaotic. Unlike light contained inside a clean glass tube, radio waves propagate through an open, uncontrolled environment. They bounce off concrete buildings, get absorbed by foliage and heavy rainfall, suffer from atmospheric refraction, and collide with interference generated by other electronic devices.

As a subscriber moves down a city street in a car or train, their smartphone is constantly measuring the signal strength and quality of nearby cell sites. The network must perform a high-speed ballet, orchestrating a seamless "handoff" of the subscriber's active session from one cell tower to another in a few milliseconds without dropping a single packet of a voice call or video stream. Multiply this single interaction by billions of active mobile devices moving simultaneously through complex urban environments, and the scale of the mathematical challenge becomes staggering.

To manage this, modern mobile networks rely on cellular frequency reuse, dynamic power control, and precise phase synchronization. Base stations must synchronize their operational clocks down to the nanosecond level. To achieve this synchronization, network operators rely on atomic clocks installed at core sites, supplemented by Global Navigation Satellite Systems (GNSS) like GPS and Galileo. If a cell tower loses its timing signal, its radio transmissions will slowly drift out of phase, causing severe signal interference with neighboring towers and rendering high-speed mobile data impossible across entire metropolitan sectors.

What makes this global machine uniquely difficult to operate is not merely its physical extent, but its radical heterogeneity. The global network is not a single, unified entity designed by a single architect and built with uniform parts. It is a palimpsest—a sprawling, Frankensteinian structure built in layers over generations, where cutting-edge technology is grafted directly onto legacy infrastructure.

In any given national telecommunications operator, you will find modern virtualized cloud software running alongside physical hardware installed in the 1990s, which in turn links to copper cabling laid in the 1970s. A single customer request—such as placing a mobile call or loading a webpage—triggers a chain reaction that moves through dozens of different technological epochs and vendor platforms.

A typical data packet might begin its journey as an encrypted digital payload generated by a modern arm-based processor in a smartphone. It travels over a 5G radio interface using advanced beamforming technologies, hits an antenna manufactured by Ericsson or Huawei, travels down a fiber cable to a baseband processor running on a customized Linux kernel, and passes onto a backhaul network powered by Cisco routers. From there, it enters an optical backbone built on Ciena or Nokia DWDM equipment, dives into an undersea cable manufactured by SubCom, lands at an international gateway, passes through a legacy legacy circuit-switched firewall, and finally reaches a cloud data center owned by Microsoft Azure or Amazon Web Services, where it is processed by a virtual server running inside a hypervisor.

Every single vendor along this chain utilizes its own proprietary management interfaces, logging formats, operating systems, and diagnostic tools. There is no central dashboard for the planet's network. There is no master control room where a single engineer can flip a switch to see the end-to-end health of the system. Instead, the global machine operates through a fragile patchwork of standardized protocols, customized translation scripts, and manual operational procedures developed over decades.

This layered complexity creates a system that is profoundly opaque. When something goes wrong in the middle of this stack, tracing the root cause is notoriously difficult. A degraded video call in Chicago might be caused by an oversubscribed radio sector, a misconfigured router queue in Denver, an optical signal degradation on an undersea cable near Halifax, a minor software bug in a firewall vendor's firmware update, or a sudden spike in solar radiation affecting a synchronization satellite.

In a system of this size, tiny micro-events occur continuously. Transistors overheat; physical fibers stretch and crack as ambient ground temperatures change; electrical power grids fluctuate; cosmic rays flip bits in unbuffered memory modules; shipping vessels drag their anchors across deep-sea trenches; road crews dig up municipal streets and sever fiber trunks; rainstorms attenuate millimeter-wave radio signals.

The machine is so large and exposed to the elements that it is impossible to maintain in a state of static perfection. At any given moment, thousands of individual hardware failures, signal degradations, and software exceptions are occurring simultaneously across the global footprint. The network only functions because it was engineered with immense redundancy and resilience—traffic is continuously re-routed around dead links, error-correcting codes fix corrupted bits, and backup generators kick in when main power fails.

However, this systemic redundancy creates its own subtle hazards. Because the machine is built to hide its own flaws and absorb localized damage, minor, hidden failures can accumulate silently across the system over time without triggering immediate outages. The network continues to function, but its structural margin of safety steadily shrinks.

Then, a seemingly minor triggering event occurs—a routine software update, a small spike in traffic during an sporting event, or a localized hardware fault. The remaining healthy nodes, operating at or near their maximum physical capacity, attempt to absorb the sudden additional load. They quickly overheat or run out of memory, shedding their own loads onto adjacent systems.

A localized fault transforms in seconds into a cascading structural collapse. Routers across multiple continents begin dropping offline in a domino pattern, BGP routes flap wildly back and forth as systems attempt to find non-existent paths, radio networks lose synchronization, and vast swathes of the global economy abruptly go silent.

The fragility of this giant is exacerbated by a physical reality that soft tech companies rarely have to consider: the stubborn physics of distance and energy.

Software engineers building web platforms can easily add server capacity by clicking a button in a cloud console, instantly spinning up thousands of virtual machines in a matter of seconds. Telecom engineers do not have this luxury. Their domain is bounded by the speed of light in optical glass—roughly 200,000 kilometers per second, or about two-thirds of its speed in a vacuum—and the physical laws of thermodynamics.

You cannot patch a physical ocean cable with a remote software update. If a subsea cable is cut by an anchor off the coast of West Africa, repairing it requires deploying specialized cable-repair ships—heavy, specialized vessels equipped with deep-sea ROVs (Remotely Operated Vehicles), grapples, and specialized optical splicing teams. It can take days for a cable ship to reach the site of the break, weeks to hook the severed ends from the seabed thousands of meters down, bring them up to a climate-controlled room on the deck of the vessel, manually splice glass fibers with microscopic precision, and drop the repaired cable back down to the ocean floor.

On land, the physical energy constraints are equally unyielding. The global telecommunications network is one of the largest consumers of electrical power on Earth. Base stations, central offices, and data centers consume vast quantities of gigawatt-hours to power radio amplifiers, routing silicon, and the massive industrial cooling systems required to keep those chips from melting.

As data traffic continues to explode exponentially—driven by high-definition video, cloud computing, real-time gaming, and the proliferation of connected devices—telecom operators face a daunting physical wall. Simply scaling up the traditional infrastructure by adding more powered equipment, more radio towers, and higher-wattage amplifiers is economically and thermodynamically unsustainable. The physical materials required—copper, lithium, specialized synthetic silica—and the sheer electrical energy required to run them would strain global resources and push operational costs beyond profitability.

The machine, in other words, cannot simply grow bigger along its traditional architectural trajectories. It must become radically more efficient. It must learn how to dynamically turn off unused radio sectors during low-traffic periods without dropping unexpected emergency calls; it must adjust laser power output based on real-time atmospheric conditions; it must anticipate route congestion minutes before it happens and re-engineer light paths across continents without human intervention.

For decades, the engineering community managed this fragile giant through an elaborate hierarchy of human labor and manual abstractions. They constructed Network Operations Centers (NOCs)—cavernous, dim rooms equipped with massive wall-sized monitors that tracked the health of global networks using color-coded nodes. Green meant healthy; yellow meant warning; red meant critical failure.

When a node turned red, human operators sat down at command-line terminals, typed out diagnostic commands, analyzed log files, referenced physical paper manuals, and executed carefully scripted procedures to restore service. If a physical component was broken, they dispatched a field technician in a van with a toolkit and a ladder to physically replace a line card, clear a branch from a line, or reset a circuit breaker.

This human-centric operational model worked acceptably well when networks were static, centralized, and slow-moving. In the era of fixed-line copper telephony, a network topology remained essentially unchanged for years at a time. Call volumes followed predictable human routines: peaks during business hours, lulls at night, and sharp spikes on holidays like Mother's Day. System changes were planned months in advance, subjected to rigorous change-control committees, and executed during carefully scheduled maintenance windows in the dead of night.

That world no longer exists.

The modern telecommunications network is no longer a static collection of physical wires and dedicated hardware switches. It has evolved into a hyper-dynamic, highly distributed software environment where network functions are continually instantiated, torn down, moved, and reconfigured on demand across thousands of edge compute sites.

A modern network site can change its internal configuration thousands of times in a single minute. It alters radio beam directions to follow moving crowds, creates temporary virtual network slices with guaranteed low latency for remote surgical procedures or automated industrial vehicles, and dynamically reallocates bandwidth between fixed home broadband and mobile users.

The sheer volume of operational state data generated by this modern architecture is mind-boggling. A major national tier-one telecommunications operator generates tens of billions of performance telemetry events, log lines, alarm signals, and metric data points every single day.

No team of human operators, regardless of how well-trained, experienced, or well-rested they might be, can process this overwhelming deluge of real-time information. A human operator looking at a dashboard in a Network Operations Center sees only a heavily summarized, delayed snapshot of what the network was doing minutes ago. By the time a human being notices that a performance metric has entered a critical threshold, reads the log output, and decides which command to type into a command-line interface, the underlying conditions inside the network have already shifted thousands of times over.

The machine has scaled far beyond the limits of human cognition. Its physical reach is too wide, its internal dynamics are too fast, and its architecture is too mathematically complex for the human mind to actively manage.

The global telecommunications system has become, in the most literal sense, an alien territory to its own creators. Human engineers designed every individual component—the lasers, the routing protocols, the radio antennas, the software hypervisors—yet no single human being, nor any collective group of humans sitting in a operations center, can fully comprehend or control the real-time operational behavior of the system as a whole.

This fundamental mismatch between the physical complexity of the network and the cognitive capacity of human operators is the core structural crisis of modern telecommunications. The global machine has reached its physical limit under manual human stewardship. It has become a fragile giant, continually balancing on the edge of failure, held together by legacy scripts and human brute force that are rapidly reaching their breaking point.

To keep the signals flowing, to prevent the global digital economy from collapsing under the weight of its own complexity, the architects of this system have realized that they must take a radical step. They must remove human hands from the operational loop entirely. They must give the machine its own brain—a cognitive control system capable of perceiving physical reality, reasoning through hyper-complex topological trade-offs, and acting at machine speed across millions of nodes simultaneously.

The era of manual network management is coming to an end. The race to build the self-driving network has begun.