Every few weeks another company announces cuts and says the word AI in the same sentence. If you strip away the noise, a consistent pattern shows up across very different companies: the people building and directing AI systems keep their jobs, and the people doing repeatable, rules-based work do not. This article breaks down that pattern using real, sourced examples, then gives you a straightforward way to check where you stand and what to do about it.

The pattern shows up the same way at company after company

Look at Salesforce. In its June 2026 round, the company cut 86 jobs concentrated in sales, general administration, and technology and product functions tied to Agentforce, MuleSoft, and Marketing Cloud, while the core Agentforce engineering team was not directly affected, according to Metaintro (metaintro.com). CEO Marc Benioff has said engineering headcount has stayed mostly flat because the company is using AI to make its engineers more efficient, per the same report. This was Salesforce's third reduction round in roughly nine months, and Metaintro notes the throughline: roles tied to repeatable, rules-based tasks face the most pressure, while positions that design and direct automation remain in demand. Now look at IBM. Bloomberg reporting cited by Yahoo Finance found that IBM eliminated roughly 200 HR jobs by replacing them with AI agents, part of a broader push of 7,000 to 9,000 U.S. position cuts between Q4 2025 and April 2026, even as the company redirects its remaining U.S. workforce toward AI and hybrid-cloud priorities (finance.yahoo.com). Two very different companies, the same shape of decision.

Support, admin, and rules-based roles are absorbing the pressure

The functions showing up on cut lists share a common trait: predictable inputs and predictable outputs. At Salesforce, the roles reduced were in sales operations, general administration, and customer support, and the company's support division had already shrunk from roughly 9,000 workers to about 5,000, a reduction near 4,000 jobs, according to Metaintro (metaintro.com). At IBM, the approximately 200 HR roles replaced by AI agents fit the same profile, per Yahoo Finance (finance.yahoo.com). This is not unique to these two companies, but they illustrate the mechanism clearly: when a workflow can be described as a set of steps a system can learn to follow, that workflow becomes a target for automation, and the headcount behind it becomes a target for reduction. This is not about whether a department is important. Support, admin, and HR are essential functions. It is about how repeatable the day-to-day tasks inside that function are, which is a very different question.

The roles that survive are the ones directing the automation

Metaintro's reporting on Salesforce puts it plainly: roles closest to building the product customers pay for survive, and people who can design, deploy, and supervise AI agents are the people companies still compete hard for (metaintro.com). Metaintro CEO Lacey Kaelani frames it as restructuring rather than pure elimination: AI is not completely eliminating roles, but instead restructuring roles and slowing hiring for others. IBM's own framing supports this too. Rather than simply cutting headcount across the board, Bloomberg reported IBM is concentrating its remaining U.S. workforce on AI and hybrid-cloud priorities, according to Yahoo Finance (finance.yahoo.com). The dividing line is not seniority or tenure. It is whether your work sits upstream of the automation, shaping and steering it, or downstream, performing tasks the automation is being built to absorb. This is why two people with the same job title at two different companies can have very different levels of exposure: what matters is whether the daily work is about directing systems or executing within a script the systems are learning to run themselves.

A calm framework for assessing your own exposure

You do not need to panic to take this seriously. Start by listing your core weekly tasks and asking, honestly, how many of them follow a repeatable script: same inputs, same steps, same output format. If most of your role fits that description, you sit closer to the exposed side of the pattern seen at Salesforce and IBM. That does not mean your job disappears tomorrow, but it means the trend line is not in your favor. Next, look at whether your work involves judgment calls that require context an AI system does not have: relationship management, ambiguous prioritization, cross-team negotiation, or building the systems themselves. Those are the roles Metaintro describes as directing AI rather than being streamlined by it (metaintro.com). If you are exposed, the practical move is to shift your resume and your actual work toward the supervising, building, or judgment-heavy parts of your function before a restructuring forces the question. Standout can help you tailor your resume and applications toward those higher-leverage roles without spending hours reworking every application by hand.

Want fewer, better-targeted applications instead of guessing? Standout uses application pattern recognition and real resume tailoring to help you apply less and get more interviews.

Frequently asked questions

Does working in a protected function like engineering guarantee safety?

No. According to Metaintro (metaintro.com), even within engineering at Salesforce the distinction was about proximity to building and directing AI systems, not the department label itself. Engineering roles doing repeatable work can still be exposed.

Is this pattern limited to tech companies?

The examples with the clearest public documentation are tech companies like Salesforce and IBM, according to Metaintro (metaintro.com) and Yahoo Finance (finance.yahoo.com), but the underlying logic of automating repeatable workflows is not exclusive to any one industry.

What should I do if my current role looks exposed?

Start documenting judgment-based and cross-functional work you already do, since that is the kind of contribution Metaintro describes as harder to automate (metaintro.com). Then begin applying to roles that emphasize directing or building AI systems rather than roles built around repeatable task execution.