As you’d expect, we use AI extensively in our writing. The way we use it varies between team members, who come from a wide range of backgrounds—humanities, law, business, maths, computer science and so on. AI helps us express ourselves more clearly, but we also stress the importance of maintaining our individual voices and value human input.
Although our blogs and articles usually have a single named author, they are always the result of feedback and reviews from colleagues. We’ve incorporated AI into our workflow to complement, rather than replace, human input.
In the ideas stage:
We use AI to help kickstart our thinking and shape our ideas, whether for blog posts, events, or sessions.
A typical prompt might look like this:
I want to write a blog post about ‘X’. My readers are in the tertiary education sector, so it needs to be relevant to them. I’d like to focus on ‘X’ and ‘X’. Can you suggest some ideas to include?
We also use AI to collate and organise information. For example:
Can you identify 3-5 themes and group these articles for a newsletter update?
In the research stage:
We don’t rely on generative AI as a primary source of information, but we use it in ways similar to how students tell us they do.
There’s a lot to read about AI! In our team, it’s a collective responsibility to find and share relevant materials. Just for background, we’ll share how we do this. We use a Teams channel called ‘Interesting Articles, Podcasts, and Videos’. We bookmark articles using Raindrop, which has an IFTT (an automation service) flow to post them into the Teams channel.
The content ranges from newspaper articles and blog posts to academic research papers, and there’s a lot of it. Our use of AI varies between team members, but typically:
- We use it to summarise key points, helping us decide whether to dive deeper into a source.
- We use it to explore unfamiliar topics. Like students, we value the ability to have a conversation about a new subject rather than just Googling it.
In the writing stage:
As we mentioned, our team members come from varied backgrounds. It’s fair to say our team members from a humanities background tend to write better than AI, but some of us appreciate how generative AI can help make our writing clearer. We might drop a paragraph into ChatGPT with a simple request like:
Proofread this: <text goes here>.
Interestingly, when you do this, AI often does more than just proofreading—it typically refines the text for clarity, usually in subtle ways.
We also use AI to check for issues without making changes:
Proofread this, don’t change anything, just point out any issues: <text goes here>.
This has had two side effects:
- Some team members report that using generative AI has improved their writing, almost like having a personal coach.
- It saves time during the review stage, allowing us to focus on the finer details, as GenAI handles the initial proofreading.
We also use Generative AI when we are working collaboratively on content, to ensure consistency of style. For example:
I am writing a document about the capabilities of Gen AI, please rewrite this in the same tone as <example text from colleague>
In the review stage:
Previously, once an article was finished, it would be passed to a colleague for review and feedback. The reviewer would look at:
- Mechanical issues (spelling and grammar)
- Accuracy
- Whether the article is compelling and engaging
Reviewers have begun using generative AI to assist with the first two tasks, but at the moment, we believe that a human review is still best for determining whether the article is compelling.
We’re moving towards a process where the writer uses GenAI for a self-review of mechanical issues and includes AI feedback when passing the article to a human reviewer.
A typical prompt here might be:
Review this for accuracy and point out anything incorrect: <article text>.
While this isn’t a thorough fact-check, it does catch typographical errors (like a missing “not”) and often suggests ways to improve clarity. It also provides positive feedback, which is always encouraging.
We also do a final check:
Proofread this, don’t change anything, just point out any issues: <text goes here>.
Images
At the moment, we rarely use AI-generated images in our blog posts. We usually prefer to illustrate our blogs with human photos, which AI, though improving, still struggles with. Most of our images come from Adobe Stock. However, this is starting to change as we now have access to Adobe Firefly, which has been trained on Adobe Stock, and Adobe Stock itself now includes AI-generated images. We’ve found that Adobe Stock’s image descriptions make great base prompts for Firefly when we want variations of existing images.
We try to avoid using generative AI to create images of AI itself. These tend to be stereotypical—blue backgrounds, glowing white robots, brains, and circuit board/neural net imagery. Our approach is inspired by the work of Better Images of AI.
The human role
Human review is still vital to us. I’ve mentioned that the role of the human review is to ensure the article is interesting, compelling, and achieves its purpose. But there’s something new—now, we also check if the author’s voice has been lost. It’s always a balancing act, but if the writer’s voice (and we know each other well!) seems absent, we’ll suggest rewrites to restore it.