---
title: AI Is Raising Human Potential for Some and Eroding It for Others, and the Difference Is How You Use It
description: Does AI make us smarter or lazier? Evidence from doctors, students, coders and Go champions shows what builds your abilities and what erodes them.
author: Dr Marina Nani (Editor-in-Chief)
date: 2026-11-09T09:33:15.518Z
updated: 2026-11-09T09:33:38.627Z
canonical: https://www.richmanmagazine.com/article/ai-human-potential-progress-or-stagnation
image: https://cdn.nanimediahouse.com/pexels-4963001.webp
categories: Innovation
content_type: Feature
region: Global
publication: Rich Man Magazine
schema_type: Article
---

Think of the last difficult thing you wrote entirely by yourself. It might have been a letter of condolence, a hard email to a colleague or a toast for someone you love. You can probably still recall a line of it. You fought for those words, and memory keeps what you fight for.

## What AI Does to Your Memory

In 2025, 54 young adults in Boston wrote short essays over several months while sensors on their scalps measured how much different regions of their brains were communicating. One group wrote unaided, one used a search engine and one used ChatGPT. Brain connectivity was strongest in the unaided writers and weakest in the ChatGPT group. Minutes after finishing, 83 per cent of the ChatGPT writers could not quote a single sentence from the essay they had just handed in. In each of the other two groups, the figure was around 11 per cent.

The study was small and had not been peer reviewed, and its lead author has objected to headlines claiming it proved AI rots the brain. Nobody measured intelligence. What the experiment captured is simpler and more personal: words you did not wrestle with never fully become yours.

The same pattern appears wherever people have measured it closely. AI raises what you can produce while you hold it. Whether it raises what you can do once you put it down depends on who is doing the thinking.

## AI Deskilling in Experienced Doctors

At four clinics in Poland, 19 experienced endoscopists, each with more than 2,000 colonoscopies behind them, began working with an AI system that highlights possible precancerous growths on screen. Three months later, the procedures they carried out without the AI were compared with their record before it arrived. Their detection rate had fallen from 28.4 per cent to 22.4 per cent. With the AI switched on, their rate was 25.3 per cent, still below the level they had once reached on their own.

These doctors had spent years training their eyes, and three months of help was enough to blunt that instinct. You know the mechanism if you have followed satnav for a year and then tried to find a friend's house without it. When a machine reliably points at what matters, your attention relaxes, and the brain stops investing in effort it is no longer asked to spend. In colonoscopy, a missed growth can become a cancer.

## Why AI Helps Beginners Most

The picture changes at the start of a career. In a large customer support operation covering 5,179 agents, an AI assistant raised the number of problems resolved per hour by 14 per cent overall and by 34 per cent for the newest and least skilled staff. The most experienced agents gained almost nothing. The tool appeared to pass the habits of the best agents on to newcomers who would otherwise have taken months to pick them up.

This is AI at its most generous. For someone without connections or expensive training, it can turn a nervous first month into real competence, and it offers a ladder that used to belong to the lucky few.

## Where AI Makes Professionals Worse

When 758 consultants at a global strategy firm were given GPT-4 for realistic tasks within its strengths, they completed 12.2 per cent more work, finished 25.1 per cent faster and produced work graded more than 40 per cent higher in quality. On one business problem chosen deliberately as a task the AI tended to get wrong, consultants with AI were 19 percentage points less likely to reach the right answer than colleagues working without it. The consultants who did worst tended to accept what the machine produced and question it less. Their ideas were also more alike.

[Experienced software developers](https://www.richmanmagazine.com/article/anthropic-study-finds-ai-coding-assistants-speed-up-work-but-reduce-skills) show how convincing the illusion can be. Sixteen developers with years of work on their own large codebases tackled 246 real tasks, some with AI tools and some without. With AI, they took 19 per cent longer. Afterwards, they estimated that AI had made them 20 per cent faster.

[A survey of 319 office professionals](https://www.richmanmagazine.com/article/ai-weakens-workforce-judgment-critical-thinking) found the same tilt from the inside. The more they trusted the AI, the less critical thinking they reported applying to its output. The more they trusted their own expertise, the more they checked.

You may recognise yourself somewhere in those numbers. A clean paragraph arriving in seconds gives you the sensation of having thought, and that sensation is [easy to mistake for thinking itself](https://www.richmanmagazine.com/article/ai-answers-too-quickly-and-it-might-be-changing-how-we-think).

## How AI Made Go Players More Creative

The most hopeful evidence comes from an ancient board game. In 2016, the program AlphaGo beat Lee Sedol, one of the strongest Go players in history. A superhuman Go program was later used to score 5.8 million moves made by professionals between 1950 and 2021. For about 66 years, the quality of human decisions barely moved. After AlphaGo, it began to climb.

Players also started making moves nobody had played before. In 2015, 63 per cent of professional games contained a sequence never seen in earlier play, and by 2018 the figure was 88 per cent. Those new moves were linked to better decisions, which means players were not simply copying the machine. They were exploring.

The same machine produced two very different human stories. Lee Sedol retired in 2019 and explained that even as number one, he would face an entity that could not be defeated. Most professionals treated the machine as a teacher instead, studied moves that broke centuries of convention and became stronger than any generation of players before them.

## AI Tutoring and Real Learning

[In a Turkish high school](https://www.richmanmagazine.com/article/ai-in-schools-has-anyone-tested-it-on-children), nearly 1,000 students practised maths with one of two GPT-4 tools or with textbooks alone. Students with an open ChatGPT-style assistant solved 48 per cent more practice problems correctly. When the exam came and the AI was gone, they scored 17 per cent lower than students who had never used it. A second version, built to give hints and withhold answers, lifted practice scores by 127 per cent and left exam results level with the textbook group.

In Benin City, Nigeria, secondary school students attended after-school sessions twice a week for six weeks, using a freely available chatbot to study English while a teacher guided them. On pen-and-paper tests, their gains matched what normally takes 1.5 to 2 years of ordinary schooling. Girls, who had started behind the boys, made some of the largest gains.

Both projects used versions of the same technology. The programmes that protected learning kept a teacher in the room and made students produce the thinking themselves.

## How to Use AI Without Losing Yourself

Every one of these findings points to choices you can make tomorrow morning.

**Write the first draft yourself.** In the Boston essay study, people who wrote unaided first and then turned to ChatGPT showed a strong rise in brain connectivity. Your own thinking gives the machine something worth improving.

**Ask for hints, not answers.** Tell the tool to question you, point out your weak step or give you a clue. The students whose AI refused to hand over answers kept their exam results.

**Keep one skill manual.** Pick the ability that matters most in your work and practise it without help every week, the way pilots keep hand-flying to protect what autopilot would otherwise erode. The doctors in Poland lost ground in three months.

**Distrust the answer that sounds best.** Polished output hides errors well. The consultants who stumbled were the ones who stopped interrogating it.

**Measure instead of trusting the feeling.** Time a task with AI and then without it. The developers felt 20 per cent faster while working 19 per cent slower.

**Use AI to find the moves you would never make.** Ask it for the option you dismissed, the opposite view or the unconventional route, and then judge it yourself. Go players turned a machine that beat them into a source of new ideas in exactly this way.

Whether this becomes an age of progress or stagnation for human ability rests with you more than with the machines. You make that choice dozens of times a day, in the small moment between a blank page and a prompt box. If you have children or grandchildren, they will grow up with machines that write, diagnose and calculate faster than any person. What they learn from watching you is whether a person still does the thinking when the screen goes dark.