AI Is Reshaping Services Jobs—Are You Ready?

Let me tell you where my brain went the second I heard this question: "If GPT-6 launched tomorrow and could do your top five tasks at 90% of your best person's quality—who would your client hire? You, or the AI-native competitor with better margins?"

I immediately thought about every entry-level tech job I've seen posted in the last two years. QA tester. Junior data analyst. Content writer. First-rung coding roles. A lot of those postings are just... quieter now. And I cover gadgets—I'm not even in accounting or law. But if that question hit me in the gut, I genuinely don't know how you sit with it if you're 22 and just started at a mid-size insurance brokerage.

That question comes from a recent Bo Sar video by Bogdan, who runs an AI consultancy called Boster Agency. The framing he's pulling from is a Y Combinator talk directed at startup founders—essentially a briefing on which industries are ripe for AI-native disruption. Bogdan's spin: everything YC told those founders about attacking established businesses, established businesses can use to defend themselves first. It's a smart reframe, and the framework underneath it is worth actually thinking through—even with a few important asterisks.

The YC thesis, translated

Y Combinator's argument, as Bogdan lays it out, hinges on a margin collision. Services businesses have historically operated at somewhere around 30% gross margins—though that number deserves a grain of salt, since margins vary wildly across sub-sectors. Staffing firms can run below 20%; boutique consulting shops can clear 50%. The "30%" is a rough composite, not a settled figure. Software businesses, meanwhile, tend to cluster in the 70-80% range (enterprise SaaS typically sits higher; consumer apps lower), but they've historically had much smaller addressable markets than services.

AI scrambles this. If you can now deliver a services-level product—tax prep, mortgage processing, insurance underwriting—with software-level automation underneath, you inherit software margins while keeping the bigger market. That's the prize. And YC is apparently betting that the next generation of foundational companies comes from exactly this collision, naming tax, audit, insurance, mortgages, healthcare, and logistics as the specific battlegrounds.

Bogdan identifies four traits that make an industry a prime target. Low trust at the task level means clients don't care how the work gets done—they want the outcome. Your tax client doesn't care if a human or an AI agent filed their return. They care that it's correct. Low judgment at the task level is trickier: the overall work might require deep expertise, but if you break it into individual tasks, most are routine and rule-based. Filing a motion isn't judgment. Strategizing the case is. AI takes the first layer; senior humans own the second. High intelligence threshold is the counterweight—the work has to be hard enough that a client still needs real expertise, which is what keeps basic ChatGPT wrapper competitors from eating your lunch. And regulation as a moat is the counterintuitive one: licensing requirements and compliance burdens actually slow down the startup attackers, giving incumbents time to adapt.

It's a genuinely useful diagnostic. I found myself mentally running through gadget-adjacent categories while watching it—extended warranty services, device repair networks, tech support operations—and the four-trait test applies surprisingly well. Low client judgment about how the repair gets done ✅. Reasonably complex technical knowledge required ✅. Certification and liability frameworks that a random startup can't skip ✅.

The question that should actually unsettle you

The commoditization test is where Bogdan credits what he describes as a Sam Altman framing—though worth noting, that attribution comes second-hand through the video, so treat it as unverified. The question itself stands regardless of who coined it:

"As the models get better, does your service get stronger—or does the model itself commoditize you?"

If every new Claude or GPT release upgrades your operation for free, you're riding a wave. If every new release brings a competitor closer to doing your job for free, you're riding into a wall.

Here's where I want to be direct with you, because the video sort of breezes past it: this is not just a business-owner question. This is a career question. If you're in your first few years at an accounting firm doing transaction coding, invoice categorization, or routine compliance work—that's the layer Bogdan explicitly says AI takes over. Not in some hypothetical future. In the operational playbook he's describing right now.

The video frames "headcount decouples from revenue" as a win for business owners—same team, 3-5x output, margins expand. And for the owner, sure. For the person being decoupled, that sentence deserves more than a neutral observation. The organizational AI gap between companies that capture this efficiency and those that don't is already widening, and the people caught in the middle aren't usually the ones setting strategy.

I'm not saying don't go into accounting or law or insurance. I'm saying: go in with open eyes about which layer of the work you want to be doing in five years. The judgment layer is durable. The volume layer is not.

The operational playbook (and its honest limits)

The practical moves Bogdan outlines are straightforward: pick one process, automate it properly, then expand. Don't roll out AI company-wide at once and watch it break in five places while your team loses faith. Don't overpromise to clients before the system is actually reliable—inconsistent AI output destroys trust faster than being slightly slower or more expensive.

That second point is underrated. Bogdan calls it the "variance trap," and it maps directly to something I see in consumer tech all the time: products that work great 80% of the time are actually worse than products that work okay 100% of the time. Variance is what breaks the relationship. And in services—where you're talking about someone's taxes, their insurance claim, their mortgage application—a result that swings wildly based on which associate touched it is genuinely damaging.

The shape of the playbook is the same across every industry he mentions: AI absorbs the volume work, humans own the judgment moments. For an accounting firm, that means CPAs doing advisory work rather than transaction coding. For an insurance brokerage, agents handling complex placements and relationships rather than renewal prep. It's not a new idea—it's basically what any good manager should have been doing anyway—but AI makes the economics of it actually achievable at scale.

The part I want you to hold lightly

Here's my honest read: Bogdan is selling AI transformation audits. His agency charges for exactly this kind of operational rebuild. So when he says "the window is open right now and won't stay open forever"—which he says multiple times—understand that a guy who sells urgency-driven consulting always has a window that's closing. That's not a reason to dismiss the framework; the underlying logic is sound. It's a reason to apply it yourself rather than immediately reaching for someone to apply it for you.

The dual-edged reality here is that AI genuinely does create competitive advantage for businesses that implement it thoughtfully—and genuinely does create displacement risk for workers in the volume layers of these industries. Both things can be true at the same time without requiring a consultant to resolve the tension for you.

If I were 22 right now, deciding whether to pursue accounting, paralegal work, healthcare admin, or insurance—here's what I'd actually do with Bogdan's framework: I'd run the commoditization test on the role, not just the industry. Which specific tasks would GPT-6 handle at 90% quality? Which tasks require the kind of contextual judgment, relationship trust, or regulatory accountability that AI genuinely can't replicate yet? I'd build toward the second pile from day one. Not because AI is coming to take everything—it's not—but because the people who understand both layers, who can manage AI systems and do the judgment work, are going to be extremely difficult to replace.

That's not a scary conclusion. It's actually a pretty good map.

Tyler Nakamura is BuzzRAG's Consumer Tech & Gadgets Correspondent.