When everyone
has AI, what are
businesses actually
paying for?

A client wanted a social media strategist to build a Claude-based video-editing workflow. A bot, a skill, an automation, whatever the eventual setup would be.

The expectation was that someone hired for social media strategy should also be able to figure out how AI could do the editing.

I thought that was an unfair expectation. But it also captured something I have been seeing more often: once a business knows AI can help with a task, that task starts getting added to someone's job description.

I founded GHL Scale Up, an agency that helps businesses implement GoHighLevel, migrate their CRM systems and build marketing and customer follow-up automations. That work has given me a close view of how businesses actually interact with AI, including the gap between what they think it can do and what it takes to make it useful.

Many clients actively appreciate heavy AI use when outsourcing work. They want fast turnaround, good quality and someone checking the result. They are quite happy for AI to do much of the production, provided the finished work meets their expectations.

That last part matters. The client usually has little interest in supervising our prompts, checking every output or diagnosing why an automation failed. If they have to do all of that, they have taken back much of the work they outsourced.

This is the shift I keep coming back to: people are increasingly expected to own a business function, even at a junior level, as AI takes over the repetitive tasks within it.

For a social media hire, that might mean taking responsibility for the journey from an idea to a published video and then understanding how it performed. Strategy, production and reporting start becoming one person's problem.

There is an uncomfortable hiring mismatch here. In my experience, demand for juniors who can handle this broader brief feels ahead of the supply of people ready to do it. Yet the budget can still be a junior budget. Whoever gets hired is expected to deliver more because AI is available.

Some of this is a real expansion of what people can do. Some of it is employers expanding the brief faster than they reconsider pay, training or support.

It is a weird time to build a business. Nobody really knows where this settles, and the shifts are happening in several directions at once. Looking beyond my own experience, the research makes that picture more interesting.

One of the more troubling possibilities is that AI may be removing the work through which people used to become experts.

Stanford's August 2026 update to its Canaries in the Coal Mine research found that employment among 22–25-year-olds in AI-exposed occupations stood 19% below where it would have been if it had kept pace with less-exposed peers. The gap mainly reflected reduced hiring. The researchers found no evidence of widespread, economy-wide displacement and explicitly describe these findings as descriptive, rather than proof that AI caused the gap. ¹

That distinction changes the story. The disruption may arrive through a company deciding it no longer needs its next junior hire.

Meanwhile, PwC's 2026 analysis found that US entry-level roles most exposed to AI were seven times more likely to ask for traditionally senior skills, including leadership and judgment. ²

Put those findings beside each other and a difficult question appears: how do people develop the judgment employers want if fewer employers hire them to do the work that builds it?

Writing a first draft, fixing a small bug or preparing a basic report can be repetitive. It can also teach someone what a good brief looks like, where mistakes happen and how to recognise a weak answer.

AI could help people learn those things faster. But employers will have to make room for learning deliberately. A subscription alone does not give a junior employee the experience needed to check everything it produces.

The social media request makes more sense when viewed alongside what is happening in product and engineering.

In a 2026 survey by Notion, Amplify Partners and Vercel, 44% of respondents reported significant blurring of traditional product roles, with another 37% reporting some blurring. The survey covered 1,053 respondents and leaned toward people already building with AI, so it is a view of an active adopter group rather than every workplace. ³

McKinsey's research into organisations seeing strong gains from AI describes engineers moving into requirements and solution design, while product and design responsibilities converge. ⁴

These changes make communication more consequential. Someone covering more of the journey needs to understand the customer, explain trade-offs and know when to bring in deeper expertise. Generating code or a mockup does not resolve a disagreement about what the customer needs.

My social media example feels like another version of this. A strategist is expected to understand production because AI makes parts of production more accessible. An engineer may face a broader expectation around the interface and the customer experience for the same reason.

AI's productivity benefit can show up as a wider job description rather than a lighter workload.

This is where I see the opportunity for capable generalists. Someone who understands a function, can work productively with AI and can carry a problem through to completion becomes useful across several parts of a business.

Founder’s office roles are one expression of that kind of ownership. But the same expectation can exist inside marketing, operations or customer support. The important quality is being able to decide what needs doing, get it done and recognise when the result is inadequate.

The evidence supports broader responsibilities. It does not yet give us a clean measure of an AI-driven generalist hiring boom. Still, this is the direction I would watch closely as a founder and employer.

There is another opportunity in the gap between installing AI and getting it to work reliably.

BetterUp and the Stanford Social Media Lab surveyed US desk workers about low-quality AI-generated work. Respondents reported spending an average of one hour and 51 minutes dealing with each incident. It is a self-reported estimate, but it captures a problem that a simple measure of output speed misses. ⁵

One person's AI productivity gain can become another person's cleanup bill.

A report produced quickly still needs its claims checked. A chatbot connected to a CRM still needs to handle exceptions. A workflow that performs well in a demonstration can fail when real customer information is incomplete or inconsistent.

This creates a plausible service category around AI rescue: auditing a setup, finding where it breaks, repairing the integrations and making it usable for the team that has to operate it.

There is evidence of spending growth in the surrounding work. Upwork reported that freelancer earnings from AI integration grew 178% in 2025 compared with 2024, based on contracted work with US-origin demand. That measures integration work broadly, including new implementations, rather than rescue projects specifically. ⁶

To me, this is part of the selling-shovels opportunity in the AI gold rush. Businesses are buying the promise of automation. They also need people who can help them realise it, including when the first attempt goes wrong.

That opportunity may change as the tools improve. I would be cautious about building a business whose entire value depends on AI continuing to make today's mistakes. Understanding the customer's process gives the service a stronger foundation.

Repairing broken systems is one kind of work. Reviewing and improving the output of a functioning system is another.

Fiverr's Spring 2025 Business Trends Index reported a 641% increase in searches for services to “humanize AI content” over six months. Searches are not completed purchases, but they show people looking for help after the initial output already exists. ⁷

That is an interesting change in what someone is paying for. A first draft may already be available. The buyer needs someone to decide whether it says the right thing, sounds appropriate and is worth putting in front of a customer.

For some services, I can see a smaller team handling substantial delivery this way. AI does much of the production, while people contribute business context, editorial judgment, exception handling and quality control.

But calling something “human reviewed” cannot become an excuse for a quick glance. The reviewer needs enough expertise, time and authority to reject bad work. Otherwise, the human judgment layer becomes another claim the client has to verify.

Several expert views help frame where these changes might lead. They are arguments and forecasts, not settled outcomes.

I believe work will continue to exist, although the way people earn from it could change substantially.

There is a basic economic tension in the idea that businesses can remove employment everywhere and keep growing indefinitely. Employment gives people income. That income becomes spending, and that spending supports other businesses. Workers are also customers. A company can reduce its own wage bill, but widespread loss of income would weaken the customer base businesses depend on.

That does not guarantee a smooth transition or enough good jobs for everyone. It does mean the future of work has to include a conversation about who earns from AI's productivity gains.

The question I find most interesting is how work evolves, and what people become responsible for as it does.

I expect more roles to move toward ownership of a business function. Someone might begin with a narrow set of tasks and gradually become responsible for deciding what needs doing, building the workflow, using AI to execute it and improving the result.

That starts to look like an intrapreneur: someone building and running a part of the business from within it.

For example, a person responsible for lead follow-up could own the journey from a new enquiry to a booked appointment. Their job would include choosing the approach, setting up the automation, reviewing conversations and fixing the points where customers drop off. AI could handle much of the routine execution.

For this to be real ownership, people need decision-making authority, training and pay that reflect the responsibility. Otherwise, "intrapreneur" becomes a nicer title for an overloaded employee.

There may also be more businesses for these people to build and support. LinkedIn's 2026 research found that 18% of Gen Z entrepreneurs surveyed said AI made starting their business feel possible. That does not establish how many lasting companies or jobs AI will create, but it illustrates how the tools can bring entrepreneurship within reach. ¹¹

If more people can start viable companies with smaller teams, I expect opportunities for people who can own a function across one or several of those businesses. Whether that creates enough work to offset what disappears remains an open question.

This is the thinking behind why I started Sage Kite: to build an ecosystem around AI, spanning products, workflows, insights, services and workforce.

GHL Scale Up has given me a close view of the implementation side. With Sage Kite, I want to connect that work with the other pieces businesses need: understanding what is changing, choosing useful applications, building the systems and finding people who can take responsibility for them.

I do not know exactly what the job titles or team structures will look like a few years from now. I do believe there is an opportunity to help businesses and people grow into this broader form of ownership.

When everyone has AI, I think businesses will still pay for a function they can hand over with confidence. The tools may do much of the execution. Someone still needs to understand the business well enough to make that work count.