I’ve noticed something as more people start using AI in their everyday work. They’ll use it to help draft an email, respond to a customer, prepare for a meeting or think through how to approach a particular situation, and if things don’t then go the way they wanted, it’s very easy to look back and blame the AI.
I’ve caught myself doing it too: you take some advice, send the email and, when you don’t get the response you wanted, start wondering whether the advice was actually any good. It may well have been bad advice, because AI gets things wrong all the time, but the fact that you didn’t get the outcome you wanted doesn’t tell you that.
Imagine you’re trying to win a piece of work and you’ve written an email to the prospective customer. You ask AI to review it and it suggests changing the tone, removing a paragraph and approaching one of their questions differently. The reasoning makes sense to you, so you make the changes, send the email and two days later find out you’ve lost the work.
It’s almost impossible not to wonder whether you should have stuck with what you’d originally written, but there’s no way of knowing because you don’t get to send both versions and compare the results. Your original email might have won the work, it might have put them off even more, or the email might have made absolutely no difference because they’d already decided to use somebody else.
The same problem exists when things go well. If you use AI to help with an email and then win the work, it’s equally difficult to say that AI had anything to do with it because you might have won anyway. We have always had a tendency to judge decisions differently once we know how they turned out, but AI gives us something very obvious to either credit when things go well or blame when they don’t.
I use AI a lot in my own work, although I rarely use it because I want an answer that I can simply follow. I find it far more useful for testing my own thinking: I’ll give it as much context as I can about a situation, explain what I’m planning to do and ask it to find problems with my reasoning, tell me how somebody on the other side might interpret what I’ve written, or point out something I may have overlooked.
Sometimes I’ll change what I’m doing because of the response and sometimes I’ll decide I was happier with my original approach. There are also plenty of occasions where I disagree with it completely, which is fine because I’m not asking it to make the decision for me; I’m using it as another way of looking at the problem before I make the decision myself.
Even assuming that this puts me in a better position would be going too far. AI can give terrible advice, misunderstand the context, attach too much importance to one detail or make assumptions that simply aren’t true. If you only give it half the context, it can give you a perfectly sensible answer to a situation that doesn’t actually exist, and unless you know enough about the subject to recognise that, the answer can still sound extremely convincing.
The responsibility for what I do with that advice is still mine. If I ask AI to review an email and accept every change it suggests, I’ve chosen to do that in exactly the same way that I’d choose whether to accept advice from a colleague, a friend or anyone else I’d asked for an opinion.
There are then all the things happening on the other side that neither I nor the AI can know about. A customer might prefer another company, their budget might have disappeared, somebody internally might have changed their mind or they might have made the decision weeks ago and simply be going through the motions. You can write the best email you’ve ever written and still lose the work.
For a lot of the things I use AI for, asking whether it gave me the “right answer” doesn’t make much sense because there wasn’t necessarily a right answer to begin with. What matters to me is whether it helped me examine the information I had, consider something I’d missed or make a better-informed decision based on what I knew at the time.
That decision can still turn out badly, just as a poor decision can occasionally produce exactly the result you wanted. We shouldn’t give AI the credit for every good outcome or blame it for every bad one, because all it can really do is contribute to the decisions we make along the way.
It might help you make a better decision and, as a result, it might improve your chances of getting the outcome you want. It can’t change the future for you.