Will AI Take My Job? An Honest Answer From a Non-Coder.
I sat there. Type and remove. Type and remove. Twenty minutes in, I still had nothing - so I asked AI to write the email.
The email I couldn't write
A customer was agitated about a delay. The delay wasn't from my side, it was from his own IT department, but he wasn't ready to understand that. I had to write the reply. Senior, calm, no blame, move things forward.
I sat there. Type and remove. Type and remove. Type and remove.
Every sentence had a second meaning. I would draft a line, read it back, hear it land badly, and delete it. Twenty minutes in, I still had nothing.
So I asked AI to write it.
30 seconds. Boom. The response was as if AI could read my mind. It understood the dilemma, the tone I needed, the part I could not say. I edited two phrases and sent it.
That was the moment AI won me over. Rest is history.
I'm telling you this because I am not a developer and I am not 25. I run services delivery in pharma IT, twenty years in. Eighteen months ago I was the skeptic this post is written for.
Why I almost never got there
I first heard about AI in 2022. A friend back from the US mentioned the US military was taking up AI. I was taken aback. I wasn't aware until then that a technology called artificial intelligence existed.
I did what most curious people do. I read articles. And I hit a lot of conjecture. AI will take away jobs. AI will take away thinking. AI will take away the things that make work meaningful. I read the negative sentiment first and I stopped there. The phase faded out. I got back to work.
If you've been doing the same-heard the noise, read the doom, decided you'll get to it later-you are not behind. You are in the majority. A 2024 Gallup workplace survey by Kate Den Houter: Nearly seven in 10 employees say they never use AI, while one in 10 say they use it at least weekly (Gallup). Seven in ten. Almost everyone you work with is asking the same question and waiting too.
But the curve is moving faster than the curves we lived through before. A 2024 NBER working paper by Bick, Blandin and Deming: 23 percent of employed respondents had used generative AI for work at least once in the previous week, and 9 percent used it every work day (NBER 32966). The same paper notes that work adoption of generative AI has been as fast as the personal computer (PC), and overall adoption has been faster than either PCs or the internet. I lived through the PC ramp and the internet ramp at work. This one is faster.
That's not "AI will take your job." That's "your colleague is starting to use it on Tuesdays."
The two things I got wrong on day one
In January 2024, AI started hitting my ears again. The curious buff side of me started reading articles again, and this time I read about AI identifying problems in code and suggesting probable solutions. That hit home. I'd been a developer for a short period. I know the feeling. Sleepless nights, deadline approaching, you cannot figure out where the problem lies, you are scratching your head.
So in February 2024 I finally tried it. And how stupid I was then.
I used it as a Google search.
I typed the way I would type into Google. Three or four keywords, terse, no context. AI gave me back a paragraph that didn't match what Google would have given me. I concluded "AI is not as good as Google." I closed the tab.
That conclusion was wrong, and I'd like to save you the months I lost on it. Apparently, Google has its own charm; AI has its own forte. Google is a librarian who hands you ten books. AI is a colleague you describe a problem to, who writes you a first draft. Say "tax India 2024" to that colleague and you'll get a confused look. Tell them who you are, what you want, what you've tried, what would help-and they will. AI works the same way.
The second thing I got wrong was deciding AI was biased. The framing I'd use now is that AI responds to the prompt you give it. The biases at the user level are mostly the prompts. The training data isn't perfectly neutral either, and I'm still working that out. But at the level of you-and-the-screen, prompting is not magic words. Prompting is context.
If that sounds like an excuse to skip the prompting course, it is. Ethan Mollick, Two Paths to Prompting (November 2023): The lesson is that just using AI will teach you how to use AI. (One Useful Thing). Nobody has displaced that line.
Friend, subordinate, not foe
Once I stopped expecting Google, the picture changed.
The mental model I use now: AI is like that child in class who is friendly with everybody. Front-benchers, back-benchers, doesn't matter. The child isn't manipulative; the child just knows how to deal with each individual. AI is a friend you can speak to about anything. A subordinate who helps you do your task better. You stay the boss.
A document landed on my desk. 120 pages, dense, normally 2–3 days of work alongside my regular load. I handed it to AI. 5 minutes. There were a few false positives, which I'll come back to. But I had a head start, and I wrapped the whole task in a few hours. That difference made me sit up.
This isn't just my Tuesday. Anthropic's analysis of millions of Claude conversations (vendor data, flag the incentive) put the split at 57% of tasks being augmented and 43% of tasks being automated (Anthropic Economic Index). People use AI to do their job better, not to hand it over.
The most striking number is from Mollick's RCT write-up The Cybernetic Teammate. This past summer we conducted a pre-registered, randomized controlled trial of 776 professionals at Procter and Gamble, the consumer goods giant, to find out. The headline finding: Individuals working with AI performed just as well as teams without AI, showing a 0.37 standard deviation improvement over the baseline. (One Useful Thing). One person plus AI matched two people without it. P&G professionals, not coders.
If you think nobody around you is doing this, you may be wrong. Mollick, in Secret Cyborgs (March 2023): We are seeing the first controlled experiments on the use of generative AI, and they are demonstrating that the disruption of AI is already here, just not everyone knows it yet. From his informal Twitter poll in that post (anecdote, not statistic), over half of generative AI users reported using the technology without telling anyone, at least some of the time. Ask around your office anyway.
But what about hallucinations, leaks, and my regulated industry?
Now the worries. They are real, and the answers are not "don't worry about it."
Hallucinations first. AI confidently makes things up. The dangerous part is that the made-up part looks correct. Mollick, in 15 Times to use AI, and 5 Not to (December 2024): When very high accuracy is required. The problem with AI errors, the infamous hallucinations, is that, because of how LLMs work, the errors are going to be very plausible. And: Hallucinations are therefore very hard to spot, and research suggests that people don't even try (One Useful Thing).
This is the everyday version of my 120-page false positives. The output came back convincing, and a few flags were wrong. If I hadn't read every flagged item myself, I'd have shipped a worse document than I started with. So stay behind the wheel. Use AI to draft, summarise, accelerate. Verify anything that has to be right.
Now the data leak. This matters most for readers in pharma, finance, healthcare-anywhere the data is sensitive. The UK's National Cyber Security Centre is direct: the query will be visible to the organisation providing the LLM (so in the case of ChatGPT, to OpenAI). Those queries are stored. And: queries stored online may be hacked, leaked, or more likely accidentally made publicly accessible (NCSC guidance).
People are already doing it. Per Cyberhaven, a data-loss-prevention vendor (flag the incentive): Since ChatGPT launched, 4.7% of employees have pasted sensitive data into the tool at least once. And The average company leaks confidential material to ChatGPT hundreds of times per week (Cyberhaven).
One rule, non-negotiable. Ask generalised queries. Do not paste sensitive or customer data into the cloud. "Help me phrase a difficult email about a delivery delay" is fine. "Help me phrase this email to this customer about this IRB submission ID" is not. Pharma IT is stringent for a reason. That still leaves a wide lane for AI on the parts of your work where the data isn't sensitive-most of your work.
How to start using AI at work this week
The "AI helps debug code" anchor was mine. What's the equivalent for someone who doesn't write code?
Four delegations to try this week. They map to what people are reporting in 2025. HBR's 2025 ranking by Marc Zao-Sanders: There are 38 new entries in the top 100 use cases. This reinforces what we know: that there is still so much change going on. (HBR). The biggest theme of 2025 is no longer technical troubleshooting; it's Personal and Professional Support. Non-technical professionals are the dominant story now.
Draft the email you don't want to write. Mine was an angry-customer reply. Yours might be a status update, a no-thank-you to a vendor, a difficult message to a teammate. Tell AI who you are, who the recipient is, what you want, what tone. Read what comes back. Edit. Send your version, not the AI's.
Summarise the long document. My 120-page proofread. Your 60-slide deck. A regulator's 40-page guidance. AI gets you a head start. You read what matters. You stay accountable for the final output.
Plan the messy week. A user from HBR: I just asked it to create a timeline for me to clean and organize my house before we have guests staying. Same trick for "three deliverables, one sick parent, four days." Tell it the constraints. Get a rough sequence. Adjust.
Explain the thing the course glossed over. Another user, same ranking: I use ChatGPT as a study guide to explain some stuff that the course kind of glosses over, which I then add to my notes. For someone like me-picking up new domains to map customer requirements to product features-this is gold.
You don't need a prompt-engineering course for any of this. Tell it who you are. Tell it what you want. Read what it gives back. AI also learns about you while you learn about AI, so the second time is easier than the first.
Still wondering whether normal people use this stuff? A Pew Research Center survey from June 2025 reports that Today, 34% of U.S. adults say they have ever used ChatGPT, according to a Pew Research Center survey. Roughly double the share two years earlier (Pew). One in three. The room is not empty.
Stay behind the wheel
People walked to commute. Then vehicles were invented and people travelled in vehicles. But people did not lose their ability to walk. Manual car, automatic, or self-driving-doesn't matter. You are expected behind the wheel. You cannot leave the car by itself.
Will AI take your job? The honest answer is the one I started with. It is not the AI that will take your job. It is the people who adapt to AI. The exposure is real. A Brookings analysis from October 2024 puts the upper bound starkly: more than 30% of all workers could see at least 50% of their occupation's tasks disrupted by generative AI, while some 85% of workers could see at least 10% of their work tasks impacted. (Brookings). Read that twice. Notice it says tasks, not jobs. People who change with their tasks keep their jobs and outgrow them.
Final decision-makers are us. Don't power AI too much. Don't lose your ability to think because AI is around.
Open the tab tonight. Pick the email you don't want to write. Type the context: who, what, why, what tone. Read what comes back.
Adapt.