Almost 9 in 10 businesses use AI somewhere in their operations. That number comes from McKinsey's 2025 State of AI research, up from about 8 in 10 the year before. Hiring hasn't slowed down to match it. PwC's 2026 Global AI Jobs Barometer found that headcount is growing faster at companies that use AI heavily than at companies that barely use it, 52% against 36% since 2018.
McKinsey's data offers a reason why. Businesses seeing real value from AI mostly reworked how the job gets done. They didn't just add a chatbot to the old process.
PwC found something similar on a bigger scale: roles where AI takes the routine work and a person keeps the harder calls, what PwC calls "professionalized" roles, are growing twice as fast and paying more than roles where AI just makes a task easy enough for anyone to do.
McKinsey also found that only about a quarter of businesses using generative AI check everything it produces before it goes out the door. Close to half have already had a real problem because AI got something wrong.
AI use is spreading faster than anyone's checking it, and as a result, we see a role's value has also changed.
What AI is genuinely good at
AI drafts a first version of almost anything: an email, a caption. It turns a messy call into notes someone can use in five minutes. It pulls numbers out of a document faster than a person could, and it sorts and labels data without complaint. None of that takes judgment. It takes speed, and that's where AI performs best.
How different Roles have changed
Administrative
Calendar work and routine emails ate up a big chunk of a virtual assistant's week. AI handles most of that now. What's left is making the harder call. For example, when three requests land at once, which one matters most to the founder today? That would not be a scheduling problem. It's a priority problem, and it still needs someone who knows the business.
Customer Support
A chat agent answering the same five questions all day is largely gone. AI took that over. But an upset customer asking for an exception, or one who's about to cancel, was never a "questions" problem in the first place. It's now a people problem, and it stays with a person.
Marketing & Creative
Writing the first draft used to fill most of a content writer's or a social media manager's day. Now that's the fast part. The slower, harder part is deciding what's worth publishing under the brand's name and catching the one line that is not aligned with the business before a client ever sees it.
IT & Software Development
A developer still writes code, but AI writes a lot of the first draft of it. What takes real skill now is spotting the shortcut that works fine today and breaks something in six months. A code generator has no way to know that. A person does.
Sales & Lead Generation
Writing outreach emails by hand used to eat a big part of an SDR's day. That's mostly automated. What isn't automated: noticing a prospect go quiet during a call, and knowing whether to push harder or back off. No tool does that yet.
Finance & Accounting
Sorting transactions by hand used to be most of a bookkeeper's week. That runs on its own now. The priority becomes reviewing the exceptions AI flags and catching the one misclassified expense before it turns into a compliance problem.
HR & Recruitment
Screening a stack of resumes used to take up most of a recruiter's schedule. AI quickly reduces that pile fast now. Whether a candidate is genuinely the right fit only shows up in a day-to-day conversation, and that's still on the person doing the hiring.
What Your Next Hire Looks like
A business doesn't need someone spending all day writing first drafts anymore. AI does that part in minutes. What a business needs is someone who takes what AI produces and makes it accurate, on-brand, and ready to send out under the company's name.
That's a different job than what most businesses were hiring for two years ago. That also appears across every role above, from a bookkeeper reviewing flagged exceptions to a recruiter making a call a resume can't make for them.
An offshore hire in any of these areas isn't competing with AI for the first draft. They own what happens after it, usually for a fraction of what a local hire doing the same job would cost.
The numbers back this up. An offshore bookkeeper typically starts around $1,675 a month. A sales development representative starts around $2,600. A software developer starts around $2,500.
None of those rates assume the person spends their day producing first drafts by hand. They assume the person spends their time on the checking and judgment work AI still can't do, and they cost a small part of what the same role would run locally.
This is also why the role matters more than the job title. Two businesses hiring the same title, say a marketing coordinator, can end up needing very different people.
One might mostly need someone reviewing AI-drafted content before it goes to a client. The other might need someone building the whole content calendar from scratch because no workflow exists yet. Figuring out which one applies before the search starts saves weeks of back and forth later.
FAQs
Won't this mean needing fewer people eventually?
PwC's data says the opposite so far. Hiring has grown faster at the companies using AI the most, especially the ones using it to take on new work instead of just cutting costs.
Couldn't an existing team just handle this with some AI training?
Sometimes, if there's spare time to take on a new layer of checking and ownership. Many founders don't have that spare time. That's usually the real gap a new hire fills.
Isn't hiring more expensive than a subscription?
Only if a founder's own time is free. It isn't. An offshore hire who owns this full-time usually costs far less than the hours a founder spends checking AI's work themselves, week after week.
A Simple Test for Any Role
PwC draws a line between roles AI is "professionalising" and roles it's "democratising." That's a useful question to ask about any specific hire: does AI mostly make the person doing this role more capable, or does it mostly make the task so easy that almost anyone, including the founder for now, can do it well enough?
If it's the first, the role is worth hiring for properly. If it's the second, it might not be yet, and that's worth knowing before the job goes up, not after.
One more thing worth separating out: if the hesitation is about offshore hiring itself, that's a different question with a different answer. Vetting someone well and handling compliance is a meaningful job. It's usually where an established offshore partner earns its place. Our own process for getting that right is covered here.
