How to Think About When to Use Gen AI: Where is the value? The Process Itself, or the Outcome?
Gen AI presents a difficult question about the act of creation. Luckily, philosophy and common sense have already provided some great answers already.
I talk and think about AI a lot - in fact probably too much for my own sanity. One of the recurring topics I encountered in my own inquiry, and in discussion with friends, clients and basically whoever is interested in talking about it, is a deep concern that can be summed up into “if xyz AI model/tool can do this, what are people going to have left to do?”
There are a lot of drivers behind this thinking, including ethical considerations, economics, philosophical, emotional, etc. - but frankly I believe the most pressing one on everyone’s mind is the practical consideration of how this technology will impact their lives.
Focusing on this pragmatic consideration also has the benefit of messily rolling all these drivers into a workable package - none of those drivers ever exist in a silo, both in our day to day lives and thinking and the broader problem space. Everything is always related-to, informing, connected and considered together in our subjective experiences of our lives.
Unless you are an AI Ethics researcher, or otherwise have your own self-driven interest in the field, you don’t give a shit about AI Ethics as a field in a vacuum, you care about the ethics of AI in your life.
With this consideration in mind, I’ve created a simple framework deciding when to use Generative AI, and when to refrain and do the thing yourself.
A Means to an End, or a End in Itself?
In Kantian Ethics (bear with me please), Kant argued that people have innate value in themselves, and therefore are an ends in themselves and it is morally wrong to use a person solely as a means to an end. He specifics that some objectification is not problematic, but solely treating a person as a means in to an ends is.
A practical example of both:
Permissible means to an end: You go to a restaurant, you are served by a waiter. You eat your food, pay the restaurant, and tip the waiter. The waiter was a means to your end of having a dining experience, and you were the waiter’s end to those tips. This is fine.
Solely as a means to an end: You want money. You manipulate someone (lying, gaslighting, whatever) to get them to give you their money even though you have no intention of returning it or investing it. You used this person as a means to an end without regard for their intrinsic value as a human being.
You can also think of these as categories of intrinsic value (the thing itself for its own sake) vs instrumental value (the thing, for the sake of doing/getting something else). (We are now done with Kant thanks for sticking with me)
How Does this Relate to AI?
First let me clarify, I am not suggesting you should think of Gen AI models as worthy of being considered ends in their own (this would require subjective experiences/consciousness and is connected to a much larger debate that I have opinions on, but are simply not relevant here.)
Rather, the point I am trying to make, is to think about the purpose of the job to be done in this framework. Is the value of the job to be done found in the outcome (with the job being a means to an end)? Or in the process itself (with the job being an end in itself).
What is the point, the goal or reason why you are spending your time in this specific way? To use our two categories again with some of my own examples (that are broader than just a job to be done to make the point):
Spending time with my girlfriend/family/friends: this is an end itself, there is no value gained by using an AI to “spend time” for me. It is the thing for its own sake.
Doing my taxes myself: I am not an account. I am not trying to get better at doing taxes. I am paying my taxes so I don’t go to jail, do my civic duty because I like having a country, and minimize my taxes as much legally possible. This is a means to an end. It is the thing for the sake of doing something else.
In the specific context of deploying Gen AI models and tools to do some work, is the job to be done valuable in itself, or is a means to an end?
Jobs to Be to Done for Their Own Sake.
For something to be valuable in itself, is it helping you clarify your own thoughts, build a relationship or just fun and pleasurable to you, is it your craft?
For example, writing a memo, outlining a deck, or otherwise writing/whiteboarding and brainstorming is a means in itself. The point of writing that long email is often clarify your own thought, to flush out the details of a problem set, identify potential solutions and weigh the trade offs, it is to develop insight and truly create novel some novel idea. You know your intentions, your why for working on this in the first place, an AI cannot “know your intention” (nor even “have intentions” as again, subjective experience and self - but different topic).
Can you use Gen AI to help accelerate this process? Absolutely, if done carefully and with a strong core of your intention - it can help. But:
The process is the point, and there is no substitute for the process.
Trying to replace this process with generative AI is not just pointless, it is actually counterproductive, it presents the illusion that you have an understanding of the problem, when really you have just told it to vomit "looks close enough" that lacks intentionality.
Generating a memo with no true intention behind it as a replacement for a well-considered, intentioned one will only present further problems down the line, when you attempt to actualize whatever information is in that memo and suddenly you realize it lacks any sort of coherence or broader understanding of your business environment, goals, and general intuition.
So This is Just the F@#$@ “Its the Process not the Outcome” Argument?
Kinda. The idea of prioritizing the process above the outcome has been a bit run through recently in all manner of self-help social media, despite its deeply meaningful origins, but that doesn’t negate the argument itself. Its still a great idea.
Prioritizing the process is critical to acts of creation, no matter the medium, and this has been true regardless of Gen AI.
The problem space I am addressing here really is inherent to the act creation, and common sense and some faith traditions/philosophies has already provided the answer.
Have you ever rushed to get something out the door without deeply considering the question and problem state? You probably have, because of a time constraint, tiredness or just not caring - you know when there were more questions to ask and answer and when you skipped over them.
The same sort of problem will occur, to the 10th degree, during overzealous and frankly, lazy use of Gen AI, did you use it to help edit down an original idea or did you ask it to do your thinking for you?
This is just a core aspect of the act of creation, and has existed in human-only contexts forever. You can tell when something was phoned in, and when there was intent and consideration put into it. You get out what you put in.
When to Use Gen AI and How:
Any time you have an original, well-considered piece of work, that has human intention behind it, you can use it as your starting context for creating downstream works.
Think of your original document as contenting all the context you'd need to answer questions, it can answer first order questions (questions that are clearly answered in the document), and it can fake understanding by combining two first order answers to create a second order answer, but third order answers (anything outside of the original intentioned-context) it will miss on because it cannot have intention.
Have a detailed brand guideline, a few case studies and a sales playbook? It can answer cold emails with a high degree of contextual-correctness. You will need to manually review each one, but you're saving 95% of the time per email and this adds up over time in real time savings.
Or if you have a FAQ for your Shopify shop, take that, your shipping and return policy, and brand guideline and RAG (Retrieval Augmented Generation) for you orders and it will be able to answer customer service inbound for you without a major problem. Again you'll need to verify correctness, but the time savings and ability to detach from the tedium will add up in extra time and less stress.
Have a gigantic mess of notes for your book? Tell it to review and semantically relate everything by idea groups that you care about (essentially mind mapping it for you) and then see what finds.
But remember, it cannot write those notes for you, in fact no AI or person can ever create what you alone can create.
I do not mean this in the literal, physically-produced end object sense, but rather in the fact that for there to be a creation, one must want to create in the first place - to have a drive, a will to make something specific must exist prior to the act of creation, which exists prior to the creation itself. And it is in this process of creation you clarify, iterate and eventually end up at the end product.
All the above examples require intentionality, a human being with a why, as the starting point, and there is only ever one of each of us.
You → Will to Create → Act of Creation (the Process) → End Creation
Only you can create what you want to create, and there will never be another you. You can use Gen AI to help you. You amplify and maybe speed up that process, but there is not replacement for it.
Gen AI is a tool, in the toolbox of your process, to be respected and cared for, but only used when appropriate - and to know when it is appropriate for you to must engage in the process. How else can you know?
No one else can have your subjective experience and vision.
You are the source of intention, and are the answer to your own questions, more often than not. It just takes time in the process to find the answers.
So regardless of your beliefs about what AI is, what it is not, what it can and can't do, use it to facilitate your acts of creation, not replace it, because nothing truly can.
Thank you for reading. I appreciate your time. I am still trying to figure out what this is going to be, but for now I am finding a lot of joy writing up articles like this when I have the time - so stay tuned :)