AI Is Eating the Career Ladder

Episode transcript

Transcript provided for accessibility and search. Minor transcription errors may remain.

SPEAKER_01 0:01

This is AI Security Multipliers. AI, Cybersecurity, Governance with Steve. This week, AI is eating the career ladder. Artificial intelligence is making organizations more productive, reducing costs, and taking over work traditionally performed by junior employees. That sounds like progress. But what happens when the work we're automating is also the work that develops tomorrow's experts? If we stop bringing new people into our organizations, who replaces the experienced professionals when they retire, leave, or get promoted? To explore the hidden long-term cost of today's AI productivity gains, here's Steve.

SPEAKER_00 0:46

Congratulations! Your AI strategy is working. Productivity is up, costs are down. You've automated the work previously performed by 15 junior analysts. The CFO is delighted. The board is impressed. And somebody in Human Resources has probably put together a rather attractive PowerPoint explaining the transformation. Brilliant. There's just one small problem. Five years from now, three of your senior analysts retire. Two move to another company. One gets promoted. And you suddenly realize something. All those experienced people started as junior analysts. The very positions you stopped hiring for. So who exactly is going to replace them? Because AI may have eliminated the entry-level work, but it hasn't necessarily eliminated the need for experience. I'm Steve, and this is AI Security Multipliers. Today we're talking about a problem I think organizations are in danger of creating without fully appreciating the consequences. We're using AI to improve productivity today, but are we quietly destroying the talent pipeline we'll need tomorrow? Before anyone accuses me of being anti-AI, let's establish something. The productivity benefits are real. If AI can complete routine analysis in minutes rather than hours, that's useful. If it can review thousands of documents, prepare reports, investigate security alerts, or generate working software, that's valuable. And if organizations can deliver the same outcomes with fewer resources, of course they're going to explore that. They should. I'm not suggesting we preserve inefficient processes simply to keep people busy. That would be ridiculous. We don't need somebody manually copying numbers between spreadsheets just because their predecessor spent 20 years doing it. Technology should eliminate unnecessary work. But there's an important distinction between eliminating work and eliminating the opportunity to learn. Those aren't necessarily the same thing, and I think we're beginning to confuse them. Because when organizations calculate the return on an AI investment, they tend to focus on immediate benefits. Hours saved, tasks automated, costs reduced, headcount avoided. All perfectly reasonable measures. But there's another number that rarely appears in the business case. The future expertise we may no longer be developing. And that could turn out to be rather expensive. Think about the most experienced person in your organization. The senior architect, the principal engineer, the incident response specialist, the finance director. The person everyone calls when something complicated goes wrong. They weren't born knowing how to do that. At some point, they were new. They asked questions, they made mistakes, they worked through ordinary problems, they learned from people who knew more than they did, and gradually they developed something that's surprisingly difficult to manufacture. Judgment. Not simply knowledge, judgment. Knowing which details matter, recognizing when something doesn't look right, understanding the consequences of a decision, and knowing when the technically correct answer is the wrong business decision. A great deal of that comes from experience, and experience comes from doing things, including things that, viewed in isolation, don't look particularly valuable. That's the part of the AI productivity conversation I think we're missing. Some entry-level work produces two outputs, the immediate business result and a more experienced employee. We tend to measure the first, we rarely measure the second. Let's use cybersecurity as an example. Traditionally, many security professionals begin by investigating alerts, reviewing logs, documenting incidents, or working through relatively routine technical problems. It's not always glamorous. Sometimes it's positively tedious. Another suspicious login, another phishing email, another endpoint alert, another vulnerability report. But those investigations build experience. You start noticing patterns, you learn what normal activity looks like, you learn what suspicious activity looks like. You discover that a critical alert isn't always a critical incident, and that a low severity alert can occasionally be the first sign of something much worse. You learn which questions to ask, which logs to examine, which assumptions to challenge. And eventually you become the person who can investigate the complicated incidents, the person who can mentor others, the person who can design better detections, the person who can make decisions when the evidence is incomplete. Now introduce AI. It can gather context, correlate events, review telemetry, summarize incidents, and recommend actions. Excellent! But if AI performs most of the routine investigations, where do junior analysts get that experience? We can't simply remove the bottom of the career ladder and assume people will somehow arrive at the top. That's not a workforce strategy, that's a magic trick. And I haven't seen many successful cybersecurity programs built on magic. And this isn't just a cybersecurity problem. Think about software development. Junior developers learn by writing code, debugging, reviewing other people's work, breaking things, fixing them, and discovering why the elegant solution they designed yesterday has brought production to a complete standstill. Now AI can generate increasingly sophisticated code. Wonderful. But if a junior developer primarily asks AI to write software, how do they develop the understanding required to review it, or maintain it, or recognize when the AI has created something dangerous? Think about accounting. Junior accountants learn through reconciliation, analysis, and examining financial records. AI can automate much of that, but the work also teaches people how financial systems behave. Law is similar. Research, document review, drafting, learning how arguments are constructed and where they fail. Engineering, consulting, business analysis, even management. Many professions rely on a progression from relatively simple work to increasingly complex judgment, and AI is particularly attractive for automating some of that foundational work. So the question isn't simply whether AI can perform the task, it's whether removing that task also removes something important from the development of the professional. Sometimes it won't. Sometimes the task really is just repetitive administration, but sometimes it will. And organizations need to know the difference. Here's the scenario that concerns me. An organization decides it can significantly reduce graduate recruitment. AI can handle much of the entry-level workload. The savings look excellent. Year one. Lower costs, higher productivity. Everybody's happy. Year two, more automation, more savings. Another successful transformation presentation. Year three. The organization has a highly productive team of experienced professionals, supported by increasingly capable AI. Still looking good. Then we get to year five. Senior employees begin retiring. Others leave. Some move into leadership positions. And suddenly, the organization discovers that its internal succession pipeline has become rather thin. There aren't enough experienced mid-career professionals ready to move up, because five years earlier the organization stopped developing them. Now what? Hire externally? Possibly. But what happens if every organization in your industry has made the same decision? Experienced talent becomes scarce and expensive. You may find yourself paying a substantial premium for the very capabilities you stopped developing internally, which rather changes the original return on investment calculation. We've saved money today by borrowing against tomorrow's workforce, and eventually the bill arrives. Now, to be fair, the traditional career ladder wasn't perfect. Far from it. Many entry-level roles involve years of repetitive work with very little structured development. Some organizations treat junior employees as inexpensive labor rather than future professionals, and the idea that everyone must spend five years doing tedious tasks before they're allowed to contribute meaningfully is hardly a wonderful model. AI gives us an opportunity to improve that. Perhaps junior employees no longer need to spend hours gathering information. Perhaps they can spend more time analysing it. Instead of manually reviewing every security alert, perhaps they can investigate the interesting ones. Instead of writing routine code from scratch, perhaps they can learn by reviewing, testing, and improving AI-generated code. Instead of preparing endless spreadsheets, perhaps they can focus on understanding the business decisions those spreadsheets support. That could actually accelerate professional development. But there's a condition. We have to design it deliberately. Simply giving a junior employee an AI assistant and expecting expertise to develop automatically isn't enough. Because there's a difference between producing an expert-looking answer and becoming an expert. And this leads to another problem. AI can make relatively inexperienced people look extraordinarily capable. Give someone a complicated cybersecurity question. They ask AI. It produces a detailed technical explanation. Beautifully structured. Correct terminology, recommendations, perhaps even an architecture diagram. Wonderful. But does the person actually understand it? Can they explain why the recommendation is appropriate? Can they recognize when the AI is wrong? Can they defend the decision? Can they adapt it when the situation changes? Or are they simply presenting borrowed expertise? This matters because organizations may begin confusing output quality with professional competence. Someone who can produce a sophisticated analysis isn't necessarily someone who understands the analysis. And that's particularly dangerous in disciplines where mistakes have consequences. Cybersecurity, engineering, finance, healthcare, law. AI can be an extraordinary learning tool, but it can also become a very effective way of avoiding learning. And the difference depends on how we use it. So what do we do? I don't think the answer is to stop automating entry-level work. And I certainly don't think we should force junior employees to perform unnecessary tasks simply because that's how we learned. The answer is to reinvent apprenticeship. We need to be much more deliberate about how people develop expertise. Imagine a junior cybersecurity analyst working alongside an AI investigation assistant. The AI gathers the evidence, but the analyst must interpret it, challenge the conclusions, identify missing information, explain the recommended response, and justify the decision. A senior analyst reviews their reasoning, not simply whether they closed the ticket. Now the AI is supporting learning rather than replacing it. We can use simulations, realistic incidents, controlled exercises, rotations across security disciplines, supervised investigations, progressively increasing responsibility, and structured mentoring. The same principle applies to software development, accounting, engineering, and other professions. AI can remove some of the repetitive work, but we must preserve the opportunity to develop judgment, perhaps even improve it. Because if we're deliberate, AI could help people become capable professionals faster, but if we're careless, it could create a generation of people who can produce impressive work without understanding it. Those are two very different futures. There's another side to this. Experienced professionals need to change how they develop others. Traditionally, mentoring often happened informally. A junior employee asks a question, a senior employee explains. They work through a problem together. Over time, experience transfers. But if AI answers every question before the junior employee asks a colleague, those interactions may disappear. And that would be a loss. Because some of the most valuable lessons I've learned in technology and cybersecurity didn't come from documentation. They came from people. People who explained not just what to do, but why. People who challenged my assumptions, people who let me make mistakes, and occasionally people who stop me from making particularly expensive ones. AI can explain technology, but experienced professionals provide organizational context, practical judgment, and the lessons that rarely appear in a training manual. So mentoring may need to become a deliberate part of the operating model, not something we hope happens when everyone has a spare 15 minutes, because otherwise, the expertise we've spent decades developing may leave the organization when the people who hold it do. I also think leadership teams need to change how they evaluate AI investments. When we automate a process, we naturally ask about efficiency. How much time will we save? How much will it cost? How quickly will we see a return? But we should also ask what happens to the people who traditionally performed that work? Not simply whether their positions disappear, what happens to the talent pipeline? What skills are we no longer developing? Which roles depend on experience gained in those positions? How will we replace people who leave? What new development pathways do we need? And who is accountable for maintaining the organization's future capabilities? This isn't just an HR problem. It's an operating model problem, a business resilience problem. And, in cybersecurity, potentially a security risk. Because if your most experienced incident responder leaves and nobody is ready to replace them, that's a capability gap. If your senior security architect retires and all the architectural knowledge leaves with them, that's a risk. If your future security leaders never developed practical operational experience, that's a risk too. AI governance shouldn't stop at what the technology is allowed to do. It should also consider what the organization needs to remain capable of doing. So here's this episode's multiplier. Pick one role in your organization where AI is reducing entry-level work. Cybersecurity analyst, software developer, business analyst, finance associate. Whatever makes sense. Now ask five questions. First, what work are we automating? Be specific. Not just the job title, the actual tasks. Second, what did people learn by doing that work? Was it simply repetitive administration, or was it helping them develop judgment, technical understanding, and practical experience? Third, where do our future senior professionals come from? Look at your current experts. How did they develop? What opportunities did they have that new employees may no longer receive? Fourth, what's our replacement plan? When experienced employees retire, leave, or get promoted, who is ready to step into those roles? And finally, are we measuring the talent we're creating or just the work we're eliminating? Because productivity is important, cost reduction is important, but so is having people capable of running the organization five years from now. The goal shouldn't be preserving entry-level work forever, it should be preserving the development of expertise. And if AI changes how that happens, then we need to redesign the career ladder, not simply remove the bottom of it. That's the multiplier. AI is going to change work. Some entry-level tasks will disappear, some jobs will change dramatically, and new opportunities will emerge. I think much of that can be positive. We can eliminate tedious work, improve productivity, give people access to extraordinary capabilities, and potentially accelerate professional development. But we need to think beyond the next quarterly financial results. Because organizations don't simply need work completed today, they need capable people tomorrow. People who understand the technology, people who can challenge it, people who can make difficult decisions, people who can lead, and people who can develop the generation that comes after them. Every senior professional was once a junior professional, every expert was once a beginner, and every experienced leader learned from someone. So if AI eliminates the beginning of the career ladder, we'd better have a very good plan for how people reach the top. Otherwise, we may discover that our wonderfully productive AI-enabled organization has a rather inconvenient problem. Nobody left who knows how to run it. I'm Steve.