Adapting to an AI-Driven Writing Culture
Are some companies better equipped to handle the challenges of using generative AI for product development writing?
Back when I was in law school, it was common practice that if you weren’t sure what to write on a paper or an exam then you’d be better off just writing as much as you can. The belief, which often materialized when scores came, was that the TA tasked with reading your essay would be less likely to grade it poorly if you created the atmosphere that you knew what you were talking about (even if it had almost nothing to do with the question asked).
As we enter the professional world, the strategy of writing as much as possible regardless of clarity or relevance is no longer effective. Instead, we must strive for clear and concise communication. brevity is king. The number one reason - no one is tasked with reading your stuff. It doesn’t matter if you have that one idea that is going to change the world. If no one reads it, then what’s the point?
Even at a company like Amazon (where I work), with its rigorous doc writing culture, if what I write is a mess, then the only feedback I’m going to get is that I need to make it “crisper”.
With the rise of text generative AIs like ChatGPT, there is an abundance of tutorials on how to effectively utilize these new technologies for writing tasks. Already, there are whole courses on prompt engineering. This post is not going to be another tutorial.
Instead, I’d like to focus on why I believe companies with a set doc writing culture (like Amazon) are better positioned to benefit from employees leveraging such tools, and what changes I expect to see in my day-to-day work as a product manager.
Amazon’s doc writing culture is well-known for its emphasis on clear and concise communication. As a product manager, I am constantly (literally, constantly) writing and reviewing documents that outline product decisions, specifications and requirements. It’s pretty incredible to realize that almost every meeting I attend starts off with everyone shutting their video and audio to read and comment on a doc that was shared in the chat just moments ago. Only once feedback is rendered, and all attendees are done, we start discussing.
By now, it’s etched in the company’s culture that this practice is good for several reasons. First, it ensures that everyone is on the same page before a discussion begins. If you simply ask people to read ahead of time you risk the whole meeting if one person didn’t get sound to it. Second, it levels the playing field by giving everyone the same amount of time to read and digest the material before discussion starts. If they didn’t get it the first time, then there’s work to be done. Third, it emphasizes narrative over charisma - no longer would the charismatic speaker win the day with their low voice and majestic hand waving.
As a side note - this made a huge impression on me when I first joined the company. Coming from Israel, I grew up in a culture that valued the opposite. Heated debates and hand waving were considered a national sport.
With the ability of AI tools to rapidly generate complex product requirement documents, I started considering how this might integrate into Amazon's doc writing culture. First, I must acknowledge that as of now the company has a strict policy ordering employees to never share confidential information with 3rd party services (such as ChatGPT). However, I fully expect this to change in the not so distant future, and for enterprises to find a way to benefit from such tools. Once they do, I believe the company culture wouldn’t need to change much as most of its doc writing practices are very well positioned to support AI generated content. Regardless, I believe that even without sharing confidential information, product managers (and many other roles) can already benefit from ChatGPT as an assistive writing tool (see Lenny’s recent Twitter thread for some very creative ideas specifically in the product realm).
As generative AI takes a more central role in helping product managers write their docs, I believe we’ll see 3 main changes:
Insisting on seeing the data. Even today, in every doc review I attend, the easiest thing to notice is whenever someone makes a claim that is not backed by data. As it becomes easier to write, it will also become easier to generate baseless claims. Therefore, it’ll become increasingly important to question claims writers make and challenge conclusions made based on the data that is provided.
Enforcing word count limits. Writing less requires more time and energy. And with a potential inflation in the number of words written, we’re likely to see more scrutiny on lengthy writing. In my experience experimenting with ChatGPT, it excels in articulating concepts individually, but struggles to produce a coherent condensed narrative that incorporates multiple concepts seamlessly.
Calling out superfluous writing. Unlike emails sent from iPhones, I can’t imagine someone adding “Written by ChatGPT” at the bottom of their doc. However, as the use of similar tools will increase in popularity, I expect a lower level of tolerance for non essential wording and repetition. I wouldn't be surprised if a year from now I start seeing people comment “too ChatGPT” on redundant blocks of text.
As generative AI tools become more popular, they will likely become prevalent in all areas of writing. To handle this change, companies will need to implement review mechanisms for generated documents. Companies with experience in facilitating doc reviews will have an advantage over other companies. Still, only by adjusting to these new tools, and acknowledging their potential drawbacks, can we find a way to communicate more effectively. After all, Writing = Thinking, and it seems neither is going away.
What changes do you expect to see as these new tools find their way to our day-2-day work processes?
Yes, writing = thinking, but my hope is that in light of AI that writing = personal experience will start to trend more and those of us who tell real life stories and write from authentic human experience will actually stand out even more. This one line in your article hit me as a revelation. "No one is tasked with reading your stuff." It's absurdly obvious that this is true, but it put the challenge of gaining readership into a stark frame of reference. No one on earth is getting paid to read what I write, while everyone is trying to get paid for creating content. The trick is to get someone to task themselves with reading what you write, which of course comes back to exactly what you're talking about. Brevity. Relevance. Value.