Here we are, another year into the Great AI Revolution. Since I last wrote about how we’re using AI, much has changed and little has changed. In 2024, many people experimented with using generative AI for creating content — and, depending on what you asked it to do, often found it to be a poor substitute for a good human writer.
We spent the last year digging deep into understanding the tools and how GenAI works, from retrieval-augmented generation (RAG) to structured content to fine-tuning. These efforts helped us better understand how AI can best be integrated into our work and help our clients.
In general, our approach remains the same as it was a year ago: we think AI can be a tool to improve efficiency and save our clients money. It augments our process but doesn’t replace it — or our people.
But it occurred to me that the better question isn’t how to use AI, but when to use it. That answer gets complicated.
When Should We Use AI?
Recently I had an interesting conversation with a computer scientist about an aspect of using AI that I don’t think enough people consider: the energy required with each AI request. Every time we ask AI to do something, it uses a lot of computations to deliver an answer, and as a result, uses a lot of energy — much more than a simple web search.
Energy is a finite resource and using it impacts the world. That makes using AI an ethical decision. Since that conversation, I’m much more cognizant of when and why I use AI. If my brain can do the work, I try that first; then a Google search; and as a last resort I turn to AI.
It’s getting harder to make that choice because tech companies have embedded AI into almost everything. Microsoft has baked Copilot into Office 365. When I do a Google search, I first get the AI overview. Apple Intelligence is summarizing texts (with little benefit). There are ways to turn these tools off, but they don’t make it easy.
Beyond my ethical concerns about energy use, I also worry about losing essential skills. Remember when we knew phone numbers by heart before cell phones? Now I couldn’t tell you any of my family’s phone numbers without looking up their contact info on my phone. If I consistently use AI for tasks, I worry that I may lose the ability to do those things well.
If I let AI write a first draft of an article, it saves time. But the process of starting an article from scratch is the process of analyzing information and deciding what is important. Will I have the context needed to best edit an article if I don’t go through the initial steps?
I also think about the work I did when I was young that helped me develop the skills I have today. If I hadn’t had to write a thousand blurbs, would I now be able to judge whether an AI summary is good?
I don’t want to become dependent on AI for everything — only the things it can do to help me be better at what I do.
What We Use AI For
With those thoughts in mind, here are some ways we’re now using AI:
Alt text: By far the most common use we have for gen AI is writing alt text for images. Put in a good prompt, upload the image, and ChatGPT does an excellent job of describing what’s there (Gemini, less so). It’s especially useful when it’s a science image that we’ve been provided without a caption.
Grammar and proofreading: AI can be helpful in proofreading content. While I still think I’m a better judge of grammar in context, it’s good to have another set of “eyes” on content before we decide it’s ready. This is especially the case in print publications, where no one wants a typo to mar a piece forever.
Analyzing articles: AI can quickly tell me if and how a source is relevant to a topic I’m researching. It can also identify the most relevant sections or data, which is a big time saver for scholarly publications.
Formatting: While we don’t use AI to write physician bios, it can be helpful in formatting long lists of data, such as publications or awards.
Keyword research: When fed a topic and some descriptive text, AI can generate a list of strong keywords along with ideas for headings. It can also point us toward the long-tail keywords.
Because of the pace of change with AI, it can be hard to decide when is the right time to use it; something it wasn’t good at last month, it may be good at now. We’ll continue to stay on top of the latest advances in AI, always with an eye for how we can better serve our clients and what we think is an appropriate use.