Futureproof Summary: How to Make Your Job AI-Proof

A manager asks you to write down your job, step by step, for the automation team. You do it carefully. Then you start to wonder what that document is for. Is your job AI-proof?

Futureproof by Kevin Roose does not give a list of safe jobs. It gives a way to look at how you do yours, and what to change.

What is Futureproof about?

Kevin Roose is a technology writer for The New York Times. He published Futureproof in 2021. For years he doubted people who said AI would destroy jobs. Then he met people building AI software, and he heard executives talk in private. His view changed.

His main point is simple. There is no such thing as a robot-proof job. What you do matters far less than how you do it. He gives nine rules. This page covers the main ones.

Why your degree may not protect you

Many people expect that skilled, well-paid work will go last. Roose argues the opposite can be true. Fields such as finance, law, accounting and medicine are full of planning, prediction and process, which is where AI does well.

He gives examples. A system at JPMorgan Chase now reviews contracts that once took people over three hundred thousand hours a year. In a test, twenty top lawyers raced an AI on five confidentiality agreements. The AI was more accurate and finished in 26 seconds. The lawyers needed 92 minutes.

A study by the Brookings Institution found that workers with a bachelor’s or higher degree were much more exposed to AI than workers with only school education. Exposed does not mean doomed. But a degree is not a shield.

Here is a first check. If most of your day is planning, predicting or following a set process, you are standing where AI is aimed.

Jobs rarely disappear with a loud bang

We picture a robot walking in and a person walking out. Roose says that is rare. Usually it is quieter. When people leave, nobody hurries to replace them. The remaining team absorbs the work and the targets go up.

Think about the last colleague who left your team. Was the chair ever filled? That is the second check. Watch the empty chair, not the robot.

He compares it to a spreadsheet program. Excel removed whole groups of data entry clerks, yet nobody called it a killer machine.

Roose also notes that many “boring bots” now do routine office tasks, such as processing invoices. They do not look like robots. They are just software working all night.

People decide, not machines

At a business meeting in Davos, Roose heard a senior leader at an Indian IT firm describe how clients had changed. They used to want to keep most of their staff and automate at the edges. Now they ask how to do the work with a tiny fraction of the people.

The point is that executives decide how many people stay. This is not a story about machines taking over. It is about choices in meetings.

A one-month test for your job

Roose offers a hard test. Imagine your job fully written down. Give the notes to a stranger. If that person could do your job as well as you within a month, you are likely to be replaced.

Ask yourself how long a new joiner would need to learn your work from your notes. A week, a month or a year? The answer tells you something.

Be surprising, social and scarce

If your job can be copied, what is left? Roose names three qualities.

• Surprising. AI does best in stable settings with fixed rules. People do better with messy rules and new situations.
• Social. Machines meet material needs. People meet emotional ones. Roose expects lawyers to become more like advisers, and doctors to be chosen for how they treat patients.
• Scarce. This means work that is rare, high-stakes or an unusual mix of skills. A simple test: would anyone pay to watch you do it? We still cheer for swimmers, even though boats are faster.

Ask three questions. Does your work meet surprises? Would anyone miss the human doing it? Is your mix of skills rare?

Do not try to outwork the machine

Roose says you cannot outwork an algorithm. If you try, you lose, and you give away the human strengths you had.

His answer is short: elbow grease is out, handprints are in. Working harder at the routine part does not set you apart. What shows how much of you is in the work does.

Do not be the link between two machines

A designer once wrote that humans are quickly becoming expensive connection points. In software, an endpoint is the spot where one program passes data to another. Roose noticed jobs like this everywhere: people who mainly take directions from a system or carry data between two systems that cannot talk to each other.

This is the fourth check. Count how much of your day is moving information from one system to another. If it is most of it, you are standing between two machines. When they learn to talk, that gap closes.

The way out is to move toward work that needs judgement. Handle the exceptions the software cannot. Build relationships with clients. Learn how both the business and the tools work. Someone has to watch the software, and Roose compares unchecked automation to an army of chimpanzees. It is useful but can do damage without supervision.

How to use this at work and at home

1. Write down the steps of your own job. Be honest about how long a new joiner would need.
2. Mark every task that is planning, prediction or fixed process. These are most exposed.
3. Look for the work the software cannot do: odd cases, rejected items and unclear requests. Ask to take those on.
4. Spend more time with clients and colleagues on the human side of the job.
5. Learn the tools your team uses so well that you can check their errors.
6. Do not approve without reading. A rubber stamp is easy to replace.

Our take

The strongest part of Futureproof is its honesty. It does not promise safety. It gives practical tests, such as the empty chair and the one-month notes. The idea of being surprising, social and scarce is easy to remember and use.

There are weak points. Parts of the book are written for an American reader, and some examples are from before the latest AI tools. The advice to “be human” can be vague for someone with a repetitive job and no room to change it. And not everyone can ask their manager for new work. The book also does not give much help to people who lose their jobs anyway.

Still, if you want a clear way to look at your own role, it is a good starting point.

The animated summary walks through each idea with stories. It takes about 20 minutes.