The world you'll graduate into

If you're starting college now, you'll finish in about three years. This post is about the world you'll walk into then, the one that follows two years later, and what to have ready for each.

My view: within five years we'll have general intelligence. For a business, that means AI can do anything a person can do in the back end of running one.

01The back office, department by department

Who does the work in five years

HRDone by AI: hiring screens, onboarding, payroll, policy questions.
OperationsDone by AI: scheduling, procurement, reporting, follow-ups.
MarketingMostly AI. Humans who work well with AI still win on creative, because AI may lack the taste a great human creator has.
Customer careAlmost entirely AI.
FinanceAbout 95% of the grind done by AI (my estimate). Humans oversee the critical calls.
SEO and social mediaAI.
Building the appAlready easy. By then you won't need a course: say what you want, and it runs for days until it's built.

And the price of all of it falls. The main paid AI plans cost about $20 a month today. The cost of the same level of AI capability has been dropping about 10x a year, a trend Andreessen Horowitz calls "LLMflation". All the AI you need to run a business will probably cost about what you pay for your phone.

So who's left? People for judgement and vision: at the top, and at department-head level. Very few roles below them.

02This isn't far off: the numbers today

McKinsey estimates that current generative AI could automate work activities that take up 60 to 70% of employees' time.
The IMF estimates that almost 40% of jobs worldwide are exposed to AI, and about 60% in advanced economies.
Goldman Sachs estimates AI could expose the equivalent of 300 million full-time jobs to automation.
The World Economic Forum expects 170 million jobs to be created and 92 million displaced by 2030, and 39% of the core skills employers need to change.

Those are today's estimates, made with today's AI. I think they're low.

03The sequence we have to get right

When productivity jumps this fast, the order of events decides whether we get a boom or a crash. Here's the path I think works.

If governments are ready

Productivity soarsAI makes output per worker insanely high.
Layoffs avalancheCompanies need far fewer people. GDP still holds up.
Cash is readyGovernments have direct payments prepared before the avalanche peaks.
Money landsEvery account gets a payment. Demand slows anyway, and GDP dips.
DeflationAI makes goods and services far cheaper. Prices fall in a big way.
Plenty againThe payment isn't large, but with prices this low it buys a lot. Demand picks up.
RebalanceDemand and GDP grow briskly, even though most of the world doesn't have a job.

And the path if they aren't.

If governments aren't ready

Productivity soarsSame start.
Layoffs avalancheGDP still looks fine for a while.
No cash readyPeople are broke.
Demand crashesNobody can buy what the machines now make so cheaply.
GDP crumblesTax revenue goes. Businesses go.
CrisisA terrible economic crisis, and it's hard to see how we'd recover from it soon.

The difference between the two is timing. We've seen pieces of both. When demand collapsed in the Great Depression, US unemployment reached about 25% in 1933. In 2020, the US sent most households direct payments of up to $1,200 per adult within weeks of the pandemic shutdowns. Alaska has paid every resident a yearly dividend from oil revenue since 1982. And when Finland paid 2,000 unemployed people roughly $600 a month for two years with no conditions, they reported better wellbeing and less stress, and their employment was no lower than the comparison group's. The tools exist. What's missing is having them ready before the avalanche, not after.

04What you need to be ready: in 3 years

You'll graduate into the first wave. Have these:

AI as your team. Run several AI agents on one goal, check their work, and catch their mistakes. That's the job now.
One home skill, done deeply. Digital marketing, design, finance: one area where you can tell good work from bad on sight.
Shipped work. A product with real users, built by you. A degree says you studied. Something live says you can do it.
The business engine. Lean startup: problem/solution fit, an MVP (minimum viable product), customer development, product/market fit. AI builds things. You decide what's worth building.
Sales. Writing the message that gets the meeting, and running the meeting that gets the yes.

05What you need to be ready: in 5 years

Adaptability. The tools you learn will be replaced. One simple tool that anyone can use to do everything will be commonplace. The job you want may become AI's job. What remains is what AI can't sign for.

What stays human

The ideaSeeing the business nobody has started yet.
VisionWhere this should be in five years, and why.
JudgementChoosing between two good options with half the facts. It's what lands the great job.
TasteKnowing which of a hundred AI drafts is the one.
PartnershipsGetting another company to build with you.
SalesA little goes a long way. People still buy from people.
Finding the problemThe friction everyone takes for granted. AI answers questions. You pick the question.
TrustA name people believe. It compounds, and AI can't fake years of keeping your word.
AccountabilitySomeone has to sign off and answer for it. That's a human role.
An audience you ownPeople who read what you write. Distribution beats a better product.
Learning speedHow fast you drop a tool and master the next one.
Your still pointThe chaos will be loud. The people who can get calm and think clearly make the calls.

The world you'll graduate into won't reward knowing a tool. It will reward knowing what to build, who to build it with, and when to say no.

06The short version

Within five years, AI does the back office of a business, for about the price of a phone plan.
Today's estimates: 60 to 70% of work time automatable, 40% of jobs exposed, 300 million jobs' worth of work, 39% of core skills changing by 2030.
If governments have cash ready before the layoffs peak, deflation turns small payments into plenty. If not, demand crashes.
In 3 years: AI as your team, one deep skill, shipped work, the business engine, sales.
In 5 years: adaptability, plus what AI can't sign for: ideas, vision, judgement, taste, partnerships, trust and accountability.

07Sources

Appenzeller, G. (2024). Welcome to LLMflation: LLM inference cost is going down fast. Andreessen Horowitz.
McKinsey Global Institute (2023). The economic potential of generative AI: The next productivity frontier.
Georgieva, K. (2024). AI will transform the global economy. Let's make sure it benefits humanity. IMF Blog, 14 January 2024.
Briggs, J. and Kodnani, D. (2023). The potentially large effects of artificial intelligence on economic growth. Goldman Sachs Global Economics Analyst.
World Economic Forum (2025). The Future of Jobs Report 2025.
US Bureau of Labor Statistics. Historical unemployment estimates for 1929 to 1939.
US Department of the Treasury (2020). Economic Impact Payments under the CARES Act.
Alaska Department of Revenue. Permanent Fund Dividend Division: history of the dividend.
Kela (2020). Results of Finland's basic income experiment 2017–2018.

KrishnaKrishna has worked with US customers since 2001, starting in tech support. In 2005 he co-founded Commercial Street Corp, a tele-sales and appointment-setting company, and headed its international operations. He has run Your Copy Sucks since 2013.

Comments

Loading comments…

All posts

Appointment setting3

Copywriting3

AI services2

Deep Think AI15

Get the next markup by email.

One real piece of copy per issue, rewritten line by line: a cold email, a web page or a newsletter.