The Knowledge Gap AI Is Creating Is the New Security Gap. Here’s How to Close It.

Tech companies are making a familiar bet: that AI can do the work faster and cheaper than people. While in many cases this may be true, it’s also being leveraged as an easy-button for layoffs. 

In a PwC report shared earlier this year, 56% of CEOs said they’d seen no significant financial benefit from their AI investments thus far, and only 12% reported seeing cost savings and revenue growth from AI. Additionally, a February study by the National Bureau of Economic Research found that nearly 90% of firms said AI has had no impact on employment or productivity over the past three years. This all underscores a revealing truth: many organizations have yet to see the revenue and productivity returns initially promised by AI, and to cover their losses, they’re continuing to cut thousands from the workforce.  

But here’s what the earnings calls aren’t saying: when you eliminate the account manager who carries ten years of relationship context in their head, or the QA engineer whose institutional knowledge is the only thing catching defects before they reach customers, you are not cutting fat. You are cutting load-bearing walls. The rework costs, the customer churn, the quality failures that follow rarely show up in the same slide deck that celebrated the headcount reduction. They get quietly absorbed into future quarters, attributed to market conditions, and forgotten. The gap between what AI-driven restructuring promises and what it actually delivers is real, it’s growing, and for many organizations, it remains entirely unacknowledged. 

The real costs of this cycle do not land in one place. They spread across every team left standing after the cuts. Developers absorb the workload of engineers who are gone, while simultaneously being handed AI tooling that is still too immature to compensate for the lost headcount. Product teams lose the context that kept roadmaps grounded in customer reality. Support organizations are stretched thin and response quality degrades as customers immediately recognize “slop” responses to their tickets. The burden is not concentrated, it is distributed, and that diffusion is part of what makes it so easy to ignore in a quarterly report. But within that broader spread, IT teams face a distinct and compounding version of the problem. They are not just doing more with less. They are being asked to manage infrastructure that was built fast, without the institutional knowledge that gave it coherence, while simultaneously executing the offboarding of departed colleagues and the onboarding of their replacements as companies quietly reverse course on cuts they publicly celebrated. 

As we continue through the year, and organizations look to make good on AI’s promises, it’s about time we realize that there’s a lot more than revenue that gets lost when you cut human expertise in the name of AI-enabled efficiency. The good news? You can have the best of both worlds. But that requires taking a more practical, pragmatic approach to AI. One that keeps humans at the center of your strategy.  

What gets lost when knowledge walks out the door 

Critical operational knowledge lives in people and is applied through human judgement, not systems. When those people leave, that knowledge leaves with them. No matter how sophisticated the model, AI can’t fill that gap today.  

Even as AI capabilities expand, automation without nuanced context creates dangerous blind spots in critical areas:  

• Judgment and strategy: Intuition, institutional knowledge, and market awareness don’t live neatly in data sets, even unstructured ones. 
• Context and situational awareness: Humans recognize nuance, intent, and can navigate the grey areas where rigid rules may break down. 
• Empathy and trust: High-impact or emotionally charged moments (especially those involving customers or employees) demand a human-centered response. 
• Oversight and accountability: Someone has to be charged with identifying failures, bias, or getting ahead of misinformation. Simply saying, “You’re right, let me try that again,” after providing entirely incorrect information doesn’t cut it. 
• Institutional memory and system context: Understanding why teams built infrastructure a certain way (or didn’t), what they’ve tried before, and what edge cases exist that no one documented.  
• Leadership and governance: Defining policy, risk tolerance, and sticking to ethical boundaries remains a fundamental human responsibility. 

AI can surface insights, recommend actions, and accelerate execution, but it cannot replace institutional memory that prevents disasters, or offer the accountability that catches them before they spiral.  

What smart organizations are doing differently  

Organizations that succeed with AI aren’t removing humans from the equation. They’re using AI to streamline the rote, repetitive processes that add friction to everyday operations. Freeing humans to dedicate time and other critical resources elsewhere to find new creative ways to drive greater business benefits. They’re leveraging AI as a “race-with” tool in order to force-amplify. Consider the world of IT operations. These teams are already stretched thin, managing sprawling infrastructure with limited means. AI, or perhaps more specifically, machine learning (ML), can deliver real value here. AI/ML can automate formerly time consuming, rote tasks that bury valuable, subject-matter experts in operational grunt work: patch testing and research, ticket classification, workflow generation, discovery, and risk mapping.  

These tasks form the foundation of IT operations, but they consume massive amounts of time (especially as organization’s digital estates widen, the vulnerability landscape shifts, and employees increasingly expect to have positive digital experiences with the devices they use to work). When teams fail to do the basics well (managing or monitoring these capabilities and processes) because they’re underwater, the consequences hit hard: downtime, missed patches, data leaks, security breaches.  

On the other hand, when you use AI well to streamline work in these areas, you give IT staff time to do what AI can’t: clear technical debt, improve existing systems and processes, rearchitect for resilience, and preserve the institutional knowledge that keeps critical systems running. That’s more than just efficiency. It’s risk mitigation. And that sets a firmer foundation for long term growth. 

Getting AI adoption right 

Organizations that start with thoughtful AI application and adoption, while also prioritizing secure, effective enablement, are seeing (and will continue to see) greater wins from their AI strategies. Optimization doesn’t come from cutting people. It stems from investing in them, redesigning workflows, and educating teams on how to best use AI.  

This really isn’t new. Technology has been advancing through generational cycles for hundreds of years. Each time we tend to repeat the same mistakes. We over-rotate and try to make a specific advancement a panacea for whatever ails us. There’s no shortage of snake-oil salesmen out there who’ll take advantage of this desperation, over-promising and disappearing before the outcomes can be measured. The reality is that AI/ML can do what computers have always done; perform certain tasks at higher speed and accuracy than a human. Those who find those opportunities and apply AI in a pragmatic, race-with manner, will see very rapid returns on their investment. Practically, that means: 

• Leveraging AI to eliminate low-value tasks, not high-value judgment 
• Designing human-in-the-loop workflows for de-risking high-impact decisions 
• Defining clear escalation paths when AI systems fail or produce uncertain outcomes 
• Assigning explicit ownership for AI-driven decisions  
• Training teams to scrutinize AI output, not blindly accept it 

The real cost of getting this wrong 

The tech industry’s current cycle of mass layoffs followed by quiet rehiring isn’t a talent shift. It’s a transfer of risk. Organizations are trading decades of institutional memory for a few months of automated speed, and IT leaders are the ones left holding the bag when it comes to managing the fallout.  

AI offers real gains, but those mean the most when you use AI to preserve and amplify expertise, not replace it. The companies that will continue to thrive in the AI era aren’t the ones cutting the deepest or deploying the most AI tools. They’re the ones thoughtfully using AI to automate monotony, and augment human creativity, collaboration, and expertise.  

Because when the next system fails, or the next outage or cybersecurity incident occurs, you don’t want a robot or a machine at the forefront of system remediation or customer communication. That’s where you’ll want (and need) the person with deep historical context and knowledge still in the building.  

If we want to unlock AI’s full potential, it’s crucial to concentrate on leveraging AI for the tasks mired in tedium, and stop using it as an excuse for poor business performance.