Writing with AI: What it is good for and how to use it (if you dare to)
What is AI?
Let’s start with the basics because you have to know how it works in order to understand what it’s good for and how to use it.
AI is essentially a pattern recognition and regeneration machine. It sorts through data to identify patterns. And then it replicates those patterns or identifies outliers in those patterns.
So, it’s being used in radiology, for example, because it can read a whole lot of x-rays, learn what a break looks like, and then use that identified pattern to predict and forecast breaks in new x-rays. In fact, it’s had a successful diagnostic rate of over 90%, so it’s pretty good at learning what a break looks like on an x-ray and then finding those same markings on new material.
That’s awesome! I love how AI can help us with tasks like this–repetitive, knowledge-based, pattern recognition and regeneration.
Here’s the thing, AI learns based on averages. The average spine looks pretty much the same across humans, so finding anomalies in the spine is looking for deviations from the average and sending those to a human for review. Then the human has more information to determine whether the deviation is related to something about the human’s biology or a potential injury.
But language is creative and cultural and emotional. Sure, there are basic patterns and building blocks and systems, but the potential combinations of words to create meaning are infinite. So, something like predictive text has learned the average use of a word, or the average combination of words in a particular phrase, or the frequency of a phrase in human writing. It looks at what the likelihood is that the next word is x. And that’s just not what language is *for*. See, the system is a system, but it is also infinite. And words and phrases and sentences can all be combined in unique ways.
So, would you rather sound like the average and predictable every other human? Or is your language informed by where you grew up and what cultures you participate in and your choices about language? I’m hoping that you don’t want to sound just like everybody else. I don’t.
And this is relevant for business because we don’t work with just anyone. We work with people we know, like, and trust *because* we know them, we like them, and we trust them. We work with people. And people have unique ways of using language. Whether that’s how we use our voices (very distinctive cadence, musicality, accents, phrasing) or how we put together actual words.
Our language use is as unique as our fingerprints.
In writing, your language gives people information about who you are, where you’re from, (as long as you love me–sorry. Anyone else?) what your level of expertise is, how much you care, etc. Your words and phrases and sentences translate not *only* meaning but lots of other bits of information about you and your relationship to others.
What is AI good for in writing, then?
Ok, so the basic premise is this: AI is *good* for repetitive writing, templated writing, writing that isn’t really for or about humans. You might use it to write reports or sections of documents that are similar to other documents. I’ve seen it used for statements of work, contracts, summaries of company information, reports, meeting minutes, and other kind of formulaic content that isn’t really *for* or *to* a human.
Brainstorming
What it’s really good for is brainstorming. It doesn’t have good ideas because it regenerates ideas that are already out there and presents the average of all the information it’s been trained on. But it has “patience” in that you can give it prompts and ask it for responses over and over again. It will never stop having ideas (even though they are not good). And the value of that is that you can move forward with your *own* ideas faster. No friend, family member, or coworker would put up with you for that long. But AI will. And it will keep regenerating as long as you want it to.
Basic research
It’s also good for basic research. It can help you find out basic information on pretty much any topic. The thing is that it is still an average. So it will tell you the big picture of information but it does make things up, so you have to make sure to check your sources.
Gaps
One thing that AI is really useful for is finding gaps in your knowledge or perspective. You can put in all the things you know or think about any topic and ask it what you are missing. Since it has access to so much data, it can point to where your own experiences are preventing you from being able to see other paths or solutions.
Proofreading
AI has been programmed with grammar rules, so it can correct spelling, grammar, and formatting. However, language changes over time and the “rules” are not 100%. Language is infinite and flexible. So, use AI to draw your attention to *potential* errors, but don’t just accept all of its corrections. It doesn’t *know* more than you about how language actually works.
And what kinds of things can it write?
Well, the best use of its writing skills is for repetitive, standard, or templated stuff. For example, if you have to write a monthly report to document what you’ve done, you can put your notes into an AI tool and have it turn those notes into a report. Since the report is for the record but not for humans, this is a good use of AI writing.
If you have to summarize a lot of information, you can tell AI what information to pull together into a summary.
If you have to write a contract, AI is pretty good at identifying the standard contract language associated with different types of agreements and recreating that for your circumstances.
You might also think of its writing as useful for documents that are mainly objective. What “objective” means is that they contain a lot of information that can be shared with people, but they are not documents for/to a person. Generally, you can recognize “objective” documents because they don’t use words like “I’ or “you” or “we.” Documents that use “I” or “you” or “we” are speaking/writing from a person’s perspective, so they are subjective, and AI is not (by definition) a subject.
So, AI writing is good for
Brainstorming
Basic research
Gaps in knowledge or perspective
Reports, summaries, contracts, and other standard, repetitive, or templated documents
Documents that are mainly objective (no I, you, we; not *to* a person, but *about* information)
Proofreading to correct and perfect grammar and spelling
Example: You’ve written a 200 word statement of work for one client and now you need the same content with slightly different details for other clients.
What isn’t AI good for?
Basically, since AI writing is an average of all writing and doesn’t have a point of view or subjectivity, it is *not* good for any writing that you are doing that is for building relationships and connections with people.
See, humans like to work with other people, especially people they know, like, and trust. We get to know people by meeting them, whether that’s in person or online at a conference or networking event or through our family and friends or through groups of shared interests like gaming or shopping. Once we interact with people, we have the opportunity to determine whether we like them and want to keep interacting with them or not. As we get to know people, we typically get a better understanding of who they are, what they care about, what their skills are, and even what their limitations are. We build trust with each other by following through on what we say, showing a knowledge of and general commitment to basic social standards, behaving ethically and kindly with each other, and showing respect for each other. Deep trust is typically formed over multiple shared experiences where more superficial trust can be developed through appropriate attire, a good handshake, and confident body language.
The thing is, AI doesn’t have a body or a subjectivity or a point of view or any common experiences. You can’t trust it in the same way that you might trust (or not trust) a person.
AI writing is reflecting the average of other people’s writing, but that can’t replace your unique language, voice, experiences, and trust with others.
So, AI writing is not good for
Connecting with people
Building relationships in writing
Creating stories
Using emotion
Using AI successfully for writing
If you are going to use AI to help you write (and I’m trying to face that fact that you probably are), this last section is to help you create good prompts and understanding how to evaluate and edit the AI output.
Input
To create a good prompt, you need to give the AI tool the same information you would need to craft a message. You and the AI need to know
What type of document you are writing–is it an email? A contract? Meeting minutes? A commercial script? A social media post? Every type of document that exists has different qualities and parameters. It uses different language, different sizing and formats, different expectations around formality and emojis, etc. You have to be clear about what you are writing so the AI can look at similar types of documents as basic templates.
What is the content of your message?--you’ll need to explain in depth what you are writing about; otherwise, the AI will simply generate the average of all the other content available on the internet and that will sound like no one and accomplish nothing.
What is the purpose of your message?--get really clear about what you want to happen as a result of your message. Do you want people to sign up for something? Join something? Read something? Fill something out? What is the goal? You need to know so that the AI can target that goal.
Who is your target audience?--we use different types of language with different people. Some language is formal (legal writing) and some is informal (texting your best friend). The expectations change depending on all the circumstances, so you have to make it clear who your message is for and what your relationship with them is. You might even need to say how many times you’ve communicated with them before or how often you interact with them to get the right language. Our language also changes depending on power dynamics, so you might want to explain who is the decision maker related to your message. Are you telling someone else what to do? Or are they making a decision based on the information you are providing?
What are the circumstances in which people will see your message?--Are they scrolling through their social media feed? Are you sending the message directly to them? Have they chosen to see your message? Are they expecting your message? Is your message related to their job responsibilities or not? You’ll want to think about and tell AI as much as you can about how the message will reach your audience and what their investment will be in it.
What is the best organization of your message?--You might want to ask a question first or explain something first. You might use bullets or headings. You’ll definitely want to think about how to arrange the message to make it easy for your audience to read it.
Then, once you’ve put all that stuff into the prompt, your going to let AI loose. And it’s going to come back with the average of all the relevant writing it can find related to that type of document about that content for that purpose and audience in those circumstances with that organization. But it’s going to be average, not good. The difference is that AI is combining all the data it has access to in response to your prompt. But what makes good writing is that it connects with people and gets them to do what you want them to do. And remember, people want to work with people, so your AI writing is probably going to feel weird and flat to most people because it doesn’t sound like you or anybody. It sounds generic, empty, point-less, and probably inconsistent.
So, your task is now to revise and edit the AI generated content.
Lots of people will keep tweaking the prompt and asking AI to produce more and different stuff. You can do that.
But you could also just work on it yourself.
Because the thing that will connect with people is to have a clear sense of who the writer is and how the writer cares about and connect with the reader.
Output
In reviewing, editing, and reworking the output, make sure you are checking these things:
Is the message complete?--Make sure it has all the information and says exactly what you want it to say. Remember: AI doesn’t know what’s important to the audience. It knows what *other documents* emphasized to *other audiences*.
Is the content clear?--Make sure the language will make sense to the audience. Remember: AI has access to the entire dictionary. If you don’t know what the writing says, your audience won’t either.
Is the content concise and easy to read?--Humans have short attention spans. We want short writing, few words, arranged in ways that are easy for us to look at. Remember: AI doesn’t have eyes. It doesn’t care about formatting.
Is the content nice and unique?--Your writing should sound like you. Not like someone else. And not like no one. Remember: People do business with people they know, like, and trust–not AI.
Overall, I don’t like writing with AI myself. I think AI is an incredibly useful tool and it can help us regular humans accomplish a lot of amazing things.
But being literate is not the same as being a good writer.
A typical 3rd grader is pretty much literate: they can read, write, spell, communicate about pretty much any topic that they need or want to.
But I wouldn’t let a 3rd grader write my blog posts because that 3rd grader doesn’t have my expertise, skill, or witty personality.
Sure, AI can be useful for some writing tasks, especially repetitive ones that are meant to document information more than motivate people. But when it comes to writing content that people connect with? That’s where I think most of us should still be writing our own work.