Menu
The AI Speed Gap: A New Reality
By Gavin Roberts| 2nd December 2025
Consider this: while a human cybersecurity analyst carefully examines a potential threat, weighing context and consulting protocols over minutes or hours, an AI system can process thousands of similar threats in milliseconds, making autonomous decisions and executing responses before a human even completes their initial assessment. To put this into perspective, humans can fire a few hundred neurons a second, while AI operates trillions of times faster. This isn’t a hypothetical future – it’s happening now, and the implications extend far beyond cybersecurity operations.
As a CTO in the cybersecurity industry, I occupy a unique vantage point. I attend industry events where the trajectory of these changes becomes clear, but more importantly, I see the evolution firsthand in our daily operations. The threat landscape isn’t changing gradually – it’s transforming at a pace that would have seemed implausible just two years ago.
Nowhere is this more evident than in email security, where we’re witnessing what may be the most significant tactical shift in years: agentic, socially engineered attacks that leverage AI to craft highly personalized, contextually appropriate phishing campaigns at scale. These are sophisticated communications that reference real projects, mimic writing styles of actual colleagues, and exploit current organizational contexts. The AI doesn’t just automate the attack – it reasons about the target, adapts its approach, and continuously refines its tactics.
When AI can generate thousands of variations, test them, and optimize for success faster than we can update our training materials, we’re fundamentally outmatched by speed alone. It’s this lived experience that informs my perspective on the broader implications across industries.
We’re witnessing three converging forces: the transformation of cybersecurity into an AI-driven arms race, the rise of fully automated manufacturing through dark factories and agentic tools that are scooping up large portions of what would have been human tasks, and the fundamental disruption of how young people must approach career planning.
These forces are already reshaping organizational structures in ways that are difficult to reverse. Large businesses are deploying agentic AI systems that are fundamentally altering team compositions and operational processes. Teams that once required ten people now operate with five humans and several AI agents. Processes designed around human handoffs are being rebuilt around AI decision-making with human oversight at key junctures rather than at every step.
What connects these trends is speed – specifically, the growing chasm between the pace at which AI systems operate and the rate at which humans can think, adapt, and make decisions. The speed differential is creating entirely new paradigms for work, threat landscapes that evolve in real-time, and career paths that can become obsolete within years rather than decades.
The fundamental shift we’re experiencing isn’t just about automation replacing repetitive tasks – it’s about agentic AI systems that can plan, decide, and execute complex multi-step operations autonomously. Today’s AI agents can process payments, detect fraud, coordinate shipping logistics, and complete what would traditionally be four days of human work without any supervision.
Traditional automation followed scripts and rules defined by humans. Agentic AI, by contrast, makes contextual decisions, adapts to new situations, and pursues goals with minimal human oversight. When an AI system encounters an obstacle, it doesn’t stop and wait – it evaluates alternatives and proceeds.
The implications for workplace dynamics are profound. Nearly seven in ten experts now acknowledge that managing agentic AI requires entirely new management frameworks because these systems operate at a scale and speed that traditional human oversight cannot match. We’re not simply adding tools to the workplace; we’re introducing autonomous colleagues that operate on fundamentally different timescales. The question becomes: how do humans remain relevant partners in systems that can outpace our cognitive processing by orders of magnitude?
I particularly struggle with this question, especially when faced with questions from my own children about the decisions they are facing in their future. Who can answer those questions with any confidence?
Cyberattack breakout times have compressed to under an hour in many cases. Attackers can penetrate systems, move laterally, and achieve their objectives faster than most organizations can even convene their incident response teams. In September 2025, security researchers identified the first AI-orchestrated cyber espionage campaign, where attackers leveraged AI’s agentic capabilities to execute sophisticated attacks with minimal human direction. The AI adapted tactics in real-time, probed for vulnerabilities, and modified its approach based on defensive responses.
Nearly three-quarters of cybersecurity professionals report that AI-enabled threats are already having significant impact, with nine in ten anticipating such threats will dominate the landscape within the next two years. We’ve entered an AI arms race, where both attackers and defenders deploy increasingly sophisticated AI systems to outmanoeuvre each other at machine speed.
The problem is defending is difficult when we haven’t fully witnessed the threat. Over the last two years I have seen the pressure in all industries to have AI act as the defender, and I have also seen the cyber world answer the call, saturating the market with ‘AI this’ and ‘AI that’, at a time when it was not possible to know yet how the bad actors would use the new technology. We have reached a turning point, and I’m a strong believer that proven tools, combined with new tech we are starting to develop now, will ultimately be pivotal in the cause.
But here’s the critical insight: the solution isn’t simply to deploy defensive AI and step back. Human cybersecurity professionals remain essential, but their role is evolving from frontline responders to strategic partners with AI systems. The most effective security operations now involve AI handling high-speed detection and initial response while humans provide contextual judgment, strategic oversight, and ethical guardrails. This partnership model represents our best path forward in cybersecurity.
The concept of the “dark factory” – a fully automated manufacturing facility that can operate in complete darkness because no humans need to see – has moved from theoretical possibility to practical reality with stunning speed. Industry analysts project that by 2025, sixty percent of manufacturers will have implemented some form of lights-out manufacturing. China has emerged as the global leader, with robot density reaching 392 units per 10,000 workers in 2023, nearly three times the global average.
Foxconn replaced 60,000 workers with robots at a single plant in Kunshan and has committed to automating thirty percent of its operations by 2025. These are mainstream manufacturing operations producing the devices we use daily.
The displacement isn’t limited to assembly line workers. Dark factories require different skillsets entirely – robotics engineers, AI system managers, and maintenance specialists. The traditional pathway from factory floor worker to line supervisor to plant manager doesn’t exist in a facility with no floor workers. This represents not just job displacement but the elimination of entire career ladders that have provided economic mobility for generations.
Yet the partnership model offers a path forward. The most sophisticated manufacturing operations are hybrid facilities where AI and robotics handle precision and repetitive tasks while humans manage exception handling, quality oversight, and continuous improvement.
Career Decision-Making in an Accelerated World
For young people entering the workforce, the traditional career planning framework has become dangerously obsolete. When AI systems can master most tasks within a job category, research shows that employment in that role drops by approximately fourteen percent. This isn’t a one-time disruption; it’s a continuous process as AI capabilities expand.
Around a quarter of early career workers now say they may transition away from tech-dependent industries entirely to minimize AI’s impact on their careers. Gen Z is increasingly pursuing blue-collar trades and entrepreneurial ventures rather than traditional white-collar paths – a rational response to genuine uncertainty about which careers will remain viable.
The fundamental problem is that career decisions require long-term commitments – years of education and skill development – but the landscape now shifts faster than those investments can mature. A career path that seems secure today may be substantially automated within the timeframe it takes to develop expertise.
This is where the partnership framework becomes essential. Rather than viewing AI as a competitor, young people need to identify where human capabilities complement AI strengths. The careers most likely to remain robust are those where human-AI collaboration creates value that neither could achieve independently.
In cybersecurity, pure pattern recognition increasingly belongs to AI, but strategic thinking and ethical judgment remain distinctly human. In manufacturing, operating robots may be automated, but designing production systems and handling complex exceptions still require human insight. The key is developing skills in judgment, creativity, and contextual understanding while becoming fluent in working alongside AI systems as partners.
Instead of choosing a single path and developing deep, narrow expertise, young people need to cultivate adaptability, continuous learning capabilities, and “AI fluency” – the ability to effectively collaborate with and direct AI systems. Educational institutions and employers need training programs that emphasize partnership with AI rather than competition against it.
Conclusion: Navigating the Speed Barrier
We have entered an era where the speed of technological change has outpaced our traditional frameworks for adaptation. The gap between machine-speed operations and human-speed decision-making isn’t going to close – if anything, it will widen.
But speed differentials don’t have to mean human obsolescence. The partnership model – where humans contribute judgment, context, creativity, and ethical oversight while AI handles speed, scale, and pattern recognition – offers a viable path forward. This isn’t about humans trying to match AI capabilities or simply accepting displacement. It’s about deliberately architecting systems, careers, and institutions around productive human-AI collaboration.
Without thoughtful intervention, the default path leads toward widening displacement and economic disruption. We need to be intentional about creating partnership frameworks, redesigning education for adaptability, and building workplace structures that value human-AI collaboration rather than simple automation.
For young people making career decisions today, the message is both sobering and hopeful: traditional career stability is gone, but opportunities exist for those who can position themselves as effective partners with AI systems. The key is accepting the speed differential as reality while identifying where human capabilities remain essential to the partnership. That requires courage, adaptability, and a willingness to continuously learn – but it also offers a path forward in an AI-accelerated world.
Gavin is CTO at Topsec, bringing over 25 years of expertise in Financial IT, Secure Payments, and Email Security. Gavin dedicated his focus to cybersecurity and email protection, where he continues to drive innovation and secure solutions.
Sign up to get regular updates about email security