This post is about good old delegating outcomes versus delegating tasks and why it matters even more in the AI era.
I tried a few different prompts to generate a header for this post. All of them were pretty much what you would expect. All the drivel you find all over LinkedIn, marketing materials or paid tech journalism. “Strategic, harnessing, empowering, future of…”
So if I were to delegate the task to an LLM, a person using AI without any context, quite possibly the header would be vanilla LinkedIn, which is not interesting, with the only purpose that I need a header there.
“You just don’t know how to prompt it properly. If you won’t learn it, someone who knows will take your job.”
… every LinkedIn comment ever of every self-proclaimed thought leader or any kind of consultant.
So what’s the problem with delegating tasks in the upcoming era of people using AI/LLMs?
LLMs are really good at giving ok results for a lot of things.
- Do you want to write an article on the dangers of AI in the financial advice industry? Sorted.
- Write a generic return-to-office mandate? Done.
- IT security policy? Sure thing.
- Create a one-pager on why your product is the best, whilst you supply only a sentence description, and it somehow makes sense? Then you question your life choices and wonder if anyone ever reads those? Boom, done!
Any of the above saved you quite a bit of money. Hundreds of pounds, possibly.
Have you got any good leads based on your generic article? Quite possibly not. Are people coming back to the office after the most bland, generic mandate? Not really. Are you, all of a sudden, compliant with the IT policy? No, you are not. And there might be parts which you might not need or want to be compliant with. Do people understand the nuanced difference why your product is better than your competitors? You wish.
“Maybe someone else than bots will read our AI-generated article, which is exactly the same as everything else on the internet and will convert to a qualified lead eventually.” Oh, you dreamer…
“Wait a minute, this has been happening all along, so why is it different now?”
A big part of the delegating outcome is learning and finding new and different solutions. Sometimes you reinvent the wheel, sometimes you don’t know about the best practices, and you just create an abomination.
But the learning is there. It might look something like this. “I don’t know what to do! Wait, I could try this… No, that won’t work… Maybe something like this? YES YES YES! No, I just fucked it up… No, I didn’t, it actually could work!”
Through the process, there is anxiety, a bit of hopelessness and the joy of figuring something out. And that’s what makes it stick!
And if you progress this step by step, year after year, you understand what not to do and what is better. And you will be able to write a nuanced and very niche article on the dangers of AI, which will catch an eye. You will be able to explain why your product is better very clearly, or hand held your company to comply with your IT security policy.
When we get stuck with delegating tasks, it is so easy to just throw something LLM-generated over the wall. Task done. However desired outcome was nowhere to be found, and very little was learned.
We still need people to want to get to the outcome, we still want them to learn, and we want them to use the AI to complement their knowledge and experience. If we take that away, there isn’t much meaning left.
It’s the sole reason why I write this blog. To learn and hopefully share something remotely interesting. Not to just post and tick a box.
To finish with food for thought. You can have a free liability contract written by Chatgpt or by an experienced lawyer, who will be litigating if something goes wrong. Regardless of whether they used AI or not.
Doing tasks can be cheap as chips; however, achieving outcomes is a different beast, and I think we conflate those two.
Just above average 