Two years ago I feared generative AI could eventually drive me out of my profession. Now, I am using it – carefully – as a tool for content creation and data-driven content strategy. Before I explain how I have used generative AI in my writing, I would first like to explain why I decided to use it at all.
If AI-generated content could endanger my career, should I still use it?
In April of 2023, four months after ChatGPT was publicly launched, I wrote a note to myself:
“Generative AI is new to the writing profession. So, its true potential, application and impact are unknown. Yet, when I think about generative AI, I see parallels to pre-prepared meals – an easier, faster, and ultimately less fulfilling way to have dinner – and to social media – an easier, faster, and ultimately less fulfilling way to stay connected with my friends. I believe there’s a place for these technologies and products; however, for a sustainable future, their application must be used prudently."
Then, as now, I worried that outsourcing some of the most challenging, creative and time-consuming parts of the writing process to generative AI would ultimately jeopardize my career as a writer. Truly generative work takes time, stamina and creative decision-making, which are often as stressful as they are rewarding (the term “pulling teeth” and “vomit draft” are colorful indicators of the early writing stages). In contrast, generative AI can “generate” words in seconds because it leapfrogs the thinking and understanding part of writing.
I’m also concerned that increased reliance on AI-generated writing, while attractive, could be counterproductive for the clients I work on behalf of. Generative AI can be used to create passable content very quickly and affordably. And, while it doesn’t have to displace career writers or the human intention behind writing, it has that potential. We live within an economic environment in which speed, efficiency and growing profit are highly advantageous and valued. These priorities can come at the expense of outcomes most organizations want to prioritize: social benefit, long-term sustainability and health.
Rewarding communication – which I define as communication that establishes or reinforces a beneficial relationship – takes time, and is deeply human. Particularly for organizations in highly technical fields, effective communication requires multiple layers of nuanced understanding: the technical topic and its language, what you want from your audience and what you can offer them, and who your audience is – their interests, how to translate your message, and how to earn their trust and influence their behavior.
The psychology behind content strategy has corollaries in interpersonal interactions. You are more likely to fulfill a single, well-timed request from a friend than respond to a solicitor who rang the doorbell over and over again. Building a rewarding relationship with someone takes time, attention, empathy and continuous effort. Tenets of relationship building are at the core of what makes me identify as a writer and take pleasure in writing: the creative challenge of connecting people to each other through the synthesis and translation of ideas and desires. AI has no identity as a communicator. It is only a tool – one that can, theoretically, be used to replace writing and the people who do it.
I’ve now had a few years to think through how my job as a communicator differs from what generative AI can be used for, and I’m no longer afraid that my career in communication is in jeopardy. At the same time, I’ve had the opportunity to explore what generative AI can do that does not replace, but rather complements, strategic and rewarding communication.
How I’m using AI as a tool in my communications job right now
Recently, for an article on AI’s applications in medicine, I was thinking about what AI is uniquely good at doing – how it’s able to complement, rather than compete, with human creativity and interpersonal relationships. In my opinion, these are some of AI’s high-level skills:
- It’s able to generate relatively uniform results from complex or copious information
- It can do a monotonous task completely and quickly
- It makes some tasks, which previously relied on extensive experience or specific knowledge, more widely accessible
Based on this list, I’ve found a few applications in which generative AI is able to extend the reach of my intuition and capabilities as a communicator:
1. Transcribing and summarizing interviews
I often conduct interviews with subject matter experts to draft communications on their behalf and in their voice. Even though I take copious notes during these interviews, I often miss key details. And, when I’m juggling multiple assignments, I forget things that weren’t captured in my notes.
Written transcripts from recorded interviews help preserve all the information in a lasting format. However, the raw transcript from an hour-long interview can span upwards of 20 pages – or more. To address this, I use a combination of an AI transcription platform and a secure AI tool – that generates results only from sources I provide – to create condensed versions of the interview transcripts, while maintaining key details and the interviewee’s voice.
But, a word of caution when relying on interview transcripts. The handful of “stand-out moments” I remember most vividly from a conversation are often the most powerful for storytelling, because they’re most likely to also resonate with others. But these moments are hidden within a transcript, and easy to de-prioritize if I read the transcript too soon after conducting the interview. To balance the value of my memory with the value of an interview transcript, I will often start by writing a “vomit draft” or outline based on recall, and then fill it in with key details from the transcript and my notes.
2. Finding references from a library of primary sources
I often write about nuanced scientific topics in which I am not an expert. Yet, because I have good research skills, I know where to find relevant and trustworthy scientific literature. But, I often end up finding more relevant sources during the research phase of a project than I am able to remember or reference when I am writing.
Generative AI has been a useful tool for streamlining my citation management and reference recall system. Here, I use a generative AI platform that generates results based only on sources I’ve vetted and provided.
After uploading the sources I plan to reference in a project, I can prompt the platform to:
- Generate possible ways to categorize or sort the sources I provided
- Identify sources that share specific terms, topics, or perspectives
- Create tables that categorize my sources by terms or topics provided in a prompt
This is helpful for understanding common themes among my sources, and for identifying specific sources to reference when I’m crafting a topic-specific idea. I’ve found using a generative AI tool alongside a traditional citation manager helps minimize the chances of overlooking relevant information in my citation library. It’s especially useful when developing very long-form content with 20+ sources.
That said, generative AI doesn’t actually think, and its responses can’t be relied on in isolation. Although the accuracy and uniformity of responses can be improved with more specific prompts, I recommend only using generative AI platforms that are able to reference back to the original source so results – which are often partially wrong or incomplete – can be fact-checked. And, unless an AI platform is verifiably secure, only upload sources that are available to the general public and lack sensitive information.
3. Generating more uniform quantitative data to inform communication strategies
HDMZ’s approach to high-level communication strategy relies on collecting quantifiable data. For example, we might categorize all the content published by an organization by topic, audience and modality to understand, at a high level, what is being communicated, how, and to whom. Or, we might categorize all the media articles on a specific topic (such as obesity) by sub-topics to determine which specific subjects are being emphasized, and which are reported on less frequently. Generative AI tools can pull and compile content from the internet to help expedite this process.
In addition, we’ve found that if a secure generative AI tool is provided with enough information on each source, it’s typically pretty good at preliminary categorization, though it needs to be double checked for accuracy. Using AI tools, alongside human insight, makes it possible to categorize qualitative and data-rich sources quickly and with more uniformity, allowing us to spend more time on drawing insights out of the quantitative data.
So, how do I feel about generative AI today?
Two years ago, I bristled whenever generative AI was brought up in conversation. I felt as if we were in competition, and I didn’t see how AI would make my job easier or improve the quality of my work. At the time, I also had more frequent conversations with folks who immediately compared me to generative AI when they learned about my job, or suggested how I could use it to make writing faster and easier, but not better.
Now, I feel as if AI and I have reached a more comfortable truce. I’m not sure whether I’ll be glad for the invention in 50 years, but I have found ways to use it to extend my insight instead of leap-frogging it, and that does not undermine my goals as a relationship builder. While undeniably useful, I still worry about the social and environmental impact of using these tools. But, I can say the same about cars and packaged meat from the grocery store.
"Your scientists were so preoccupied with whether or not they could that they didn't stop to think if they should." – Jurassic Park
About the author:
Miriam Hyman is a Senior Science Communications Strategist for HDMZ. She draws on an analytical background in the life sciences and a passion for storytelling to communicate complex ideas with creativity, precision and regard for her audience. After work, you can find her pulling weeds at Rhythm Seed Farm in Portland, Oregon.
