Mark Cuban has been saying for a while now that the jobs automation risk conversation is missing the point. Not because the risk isn’t real, but because most people are looking at the wrong level of detail. They’re asking “will AI take jobs?” when the more useful question is “which specific jobs, in which specific ways, and starting when?” His answers to that second question are worth taking seriously.
Cuban, the billionaire entrepreneur and former Shark Tank investor, has been unusually direct about which corners of the white-collar job market look the most exposed right now. He’s not predicting a mass unemployment catastrophe. His view is more unsettling than that, actually: the erosion is happening category by category, inside companies that have already decided to stop growing their headcount in certain roles. The shift is already underway, driven by companies weighing the cost and productivity of AI systems against human labor, and as tools improve and become more cost-effective, he expects businesses to reduce headcount in roles built around repetitive tasks.
What makes his read worth paying attention to isn’t that he’s a tech billionaire talking about AI. Plenty of those exist. It’s that he isn’t particularly panicked about it, which means he’s probably not saying it for the drama. His five categories are specific, grounded, and increasingly backed up by labor market data that’s already moving in the direction he’s describing.
The Entry-Level White-Collar Jobs Automation Risk
The most immediate jobs automation risk sits at the bottom of the office hierarchy, and Cuban has been consistent about this. He believes entry-level white-collar roles are particularly vulnerable as AI removes the redundancies and “binary” tasks, like data-keeping and data entry, that those positions have traditionally handled. AI systems can now run structured, repeatable tasks at a speed no human can match.
People just starting out feel this most acutely. The entry-level job has always served a specific function in a career: you do the repetitive, low-stakes work, you learn the organization from the ground floor, and you move up. That pipeline is narrowing. A 2025 Goldman Sachs economist report, shared first with CNBC, found that unemployment among 20- to 30-year-olds in tech-exposed occupations jumped by 3 percentage points since the start of 2025, a significantly larger increase than for older tech workers or young workers in other fields. The junior roles that once existed as entry points are simply not being replaced when they open up.
Cuban’s framing of this isn’t apocalyptic. Writing on Bluesky, he drew the comparison to what happened when office technology replaced millions of secretarial and dictation roles in the twentieth century: “New companies with new jobs will come from AI and increase TOTAL employment.” Whether that net positive arrives quickly enough to help the people currently entering the workforce is a different, messier question.
Software Development: Not Elimination, But Compression
Cuban’s position on software jobs is more nuanced than a simple warning. He has pushed back against the more alarmist predictions that AI will wipe out developers entirely, pointing to the sheer scale of the technology industry as a buffer. But he does believe something important is shifting at the entry level of the field. In his view, AI will transform how developers work, not eliminate them. Routine coding tasks may be automated, but higher-level thinking, system design, and problem-solving will remain deeply human responsibilities.
The practical consequence of that compression, though, is that the junior developer role, the “write this function, fix this bug” position, is getting harder to justify at the same scale as before. AI coding assistants can now produce working first drafts of code fast enough that a team can run leaner. GitHub reports that 75% of developers now use AI assistants, which is less a warning sign and more a reality check: the workflow has already changed. The question is whether companies hire the same number of people to supervise that workflow, or fewer.
For people mid-career in software, Cuban’s message is relatively reassuring. For people trying to break in with limited experience, it’s worth taking seriously. The ladder has fewer rungs at the bottom than it did three years ago.
Customer Service: Where Automation Is Already Measurable
Customer service is probably the category where the jobs automation risk has moved fastest from theoretical to visible. Call centers and support queues have been testing AI chatbots and voice systems for years, and the results have been good enough that the rollout has accelerated. Cuban has said companies will continue to expand automation in this area, leaving fewer traditional support roles and greater demand for workers who can handle complex or sensitive interactions.
AI bots now handle basic customer queries while human operators, often working alongside an AI assistant, focus on trickier problems. Fewer people are needed to handle the same volume of inquiries, not because the interactions disappeared, but because the routine ones are automated away. What remains are the genuinely difficult conversations: the angry customer, the complicated account issue, the situation that doesn’t fit a script.
Those jobs still exist. They’re just under more pressure, require more skill, and are likely to pay less than the volume of work might suggest, because the headcount around them has been stripped back. Anyone currently in customer service who wants to stay in it needs to be the person who handles what the bot can’t, which means developing skills in judgment, de-escalation, and nuanced communication.
Data Analysis and Research: Speed Without Understanding
Data analysis and research tasks are increasingly automated, with AI tools now able to summarize datasets, generate reports, and identify trends, work traditionally performed by analysts. Cuban expects the focus to shift toward workers who can interpret results and guide AI systems rather than produce analyses from scratch.
The analyst who spent three days pulling together a market report can now get a first draft in twenty minutes. That’s genuinely useful. It’s also genuinely threatening to anyone whose value proposition was that they could pull together a market report faster than anyone else. The differentiator is no longer the speed of production. It’s the quality of the judgment about what the output means, what question to ask next, and whether the AI’s confident-sounding answer is actually right.
Goldman Sachs research paints a broad picture: globally, around 300 million jobs are exposed to AI automation, and in the US, AI can potentially automate tasks that account for 25% of all work hours. Analyst and research roles sit squarely in the category of knowledge work that makes up a large chunk of that exposure. The roles won’t vanish, but the number of people needed to fill them is likely to fall.
Finance and Legal Support: The Routine Work Is Going First
Finance and legal support round out Cuban’s list, and both categories share the same underlying pattern: the most time-consuming work in those fields is also often the most mechanical. Document review in law, compliance checks in finance, basic bookkeeping, contract summarization – these are tasks that require attention and accuracy but not creativity or judgment. Cuban has argued online that repetitive and routine tasks like review, inventory, bookkeeping, and compliance are likely to be handled by AI, resulting in more demand for workers who enter with relevant experience in those fields.
Research has found that up to 46% of tasks in administrative work and 44% in legal professions could be automated. For a junior paralegal whose primary job is document review, or a bookkeeper whose primary job is reconciling accounts, that’s a material threat. For a senior attorney or a CFO whose job involves strategy, client relationships, and judgment calls under uncertainty, it’s less so. The gap between those two positions, in terms of what AI can and can’t replace, is growing wider.
Lawyers who can work alongside an AI model fine-tuned for legal research can outcompete those who don’t, making it plausible that small numbers of attorneys, aided by AI tools, could produce the same output as larger teams. That’s a reduction in how many people are needed to do the work, not a revolution in who does it.
Cuban’s Actual Advice: Don’t Wait
Cuban hasn’t issued these warnings without a practical suggestion attached. His position, consistently, is that the people who do worst in this transition are the ones who treat AI as something that is happening around them rather than something they can use. “There’s only two types of companies in this world,” he has said, “those who are great at AI and everybody else,” and that workers “gonna have to understand how it impacts your job, or how you can use it to be better at your job.”
He has also pushed back on the assumption that AI is economically straightforward to deploy at scale. After investors on the All-In podcast revealed that some AI agents are costing more than $300 per day, adding up to over $100,000 annually, Cuban engaged with the question of whether replacing employees with AI tools is actually economically sustainable for many businesses. His skepticism about runaway automation isn’t wishful thinking. It’s based on real cost structures that make wholesale replacement of human workers less straightforward than the headlines suggest.
“Humans have a far greater capacity to know the outcomes of their actions,” Cuban has said, adding that AI agents “never do,” and that AI systems still lack real-world judgment in ways that make replacing workers risky. The technology is good at doing what it’s told, in the way a very diligent intern who has read every relevant document might be good at following instructions. It falls short when the situation doesn’t match the document.
His advice to workers in exposed roles has centered on two things: get fluent with AI tools in your field, and consider smaller companies where that fluency is visible and valued. In a small organization, someone who can set up an AI-assisted workflow for a task that used to take the team three days is immediately useful in a way that’s hard to miss. That visibility matters when decisions about headcount are being made.
What to Do With This
Cuban’s warnings are not a prediction that your specific job will disappear by next year. He has suggested that businesses may find automation still depends heavily on human oversight, particularly when decisions involve risk, judgment, or unpredictable outcomes, and he has questioned whether large-scale AI deployment even makes financial sense for many companies. The cost of AI systems, the oversight they require, and the edge cases they get wrong all eat into the economic case for full replacement. The math on wholesale automation often doesn’t add up, and Cuban has been vocal about saying so.
What the warnings do signal, clearly, is a direction. The five categories Cuban identified, entry-level white-collar work, routine coding, basic customer service, data compilation, and repetitive finance and legal tasks, all share the same characteristic: they involve doing the same kind of thing, over and over, with information that can be fed to an AI. That’s precisely what current AI systems are built for. The practical move, then, isn’t to panic and it isn’t to dismiss it. Get specific about which parts of your job fall into the “binary task” category Cuban described, and which parts require judgment, relationship, or contextual awareness that AI still genuinely can’t replicate. Put more of your energy into the second column. It won’t make your job immune to change. But it’s the difference between being the person who gets replaced by an AI tool and being the person who knows how to use one.
Disclaimer: This article was created with AI assistance and edited by a human for accuracy and clarity.