When we started tapping into the power of AI and ChatGPT a few weeks ago, we admittedly didn’t know where to begin. Like, not at all. The directive came down to do something with AI and do it fast.
So we did.
Our AI Assist beta program is off and rolling and just today we deployed the feature I mentioned in our last post, where the tool will take your title (or any subject/prompt) and lay out appropriate headers for the post—giving you a nice little head start. Just fill in what goes in-between.
This falls in line with what’s starting to solidify as a mental roadmap for where we go with AI and where it can help on our platform.
And where is that? We’re going to build from the inside out. But that’s likely best understood as groups of users, so let’s dive in.
For the bloggers–the writers themselves
The post editor is at the heart of our product. You can’t use LexBlog as a paying client without using, in some capacity, the post editor.
What’s already done
To build things fast, we started with what we call Scratchpaper, a ChatGPT-powered module off the side that can call on the model using prompts provided by the user or a few suggested ones like “Suggest a headline for this post” or “Write a tweet for this post” and so on.
With that done (quickly), we moved on to a couple purpose-driven blocks.
As mentioned above, we launched one today to our beta users—we call it Sections.
The potential uses for AI in a writing setting are almost unlimited, and that’s why it can be helpful to provide training wheels or laid-out paths to users who are completely new to the technology. But at the same time, you don’t want to inhibit folks, either.
Coming soon—a Freestyle approach to AI Assist
In nearly identical sizzle reels released by Microsoft and Google last week, showing how AI can work in their office suite products you start to get a sense for where the UI around artificial intelligence and content creation is heading.
There’s a lot of overlap in how Microsoft and Google build their tools.
- Design and the associated UI is light—it’s not in your face.
- The input UI doesn’t take up a lot of real estate—it’s there if you want it, it’s not if it isn’t.
- AI text is clearly labeled as such, with the user having the ability to confirm, undo or change.
- Even within this tool, there are guided options to change the text, like making it more whimsical or shorter or more professional.
This type of tool is where most of our engineering time is being spent now, on a block we roughly call Freestyle.
We want AI-powered assistance to be as subtle and streamlined as italicizing a few words. The full power of AI is there, but only if you need it.
This is where we are right now. What’s next?
Up next: For the editors, for the digital marketing managers
Not everyone writes their posts directly in the post editor, in the WordPress-powered interface. A lot of lawyers don’t touch the CMS at all, instead sending a Word document or something similar to a marketing professional and having them upload it.
How can we help these folks? We need to help these folks because they can be our biggest proponents within the firm.
Maybe it’s coming up with post ideas? That’s something I want to figure out. Can we have AI examine your practice area, or your posts to date, and come up with blog post ideas that we can send you and your team on a regular basis?
Or, it could be editing the posts for online consumption. What if we could come up with a tool that took your standard largelaw post with no headers, big blocky paragraphs and stodgy language and made it a bit more accessible?
What about organizing a blog as a whole? Are you using the best categories and tags you could? Are there posts from colleagues that lawyers should be citing when they provide an update on a case?
On the horizon: For the Archive, for communities, for bar associations
We started with a single post’s worth of content (or none at all!), moved towards a publication’s worth of content…and now we get a little greedy.
LexBlog and the Open Legal Blog Archive are home to as vast a body of secondary law as you will find anywhere. Our goal, one we work on everyday, is to compile and archive the insight provided by every single credible law blog.
Naturally, we have a card on our Asana board (kind of our Product task list/forum) labeled “Feed the Open Legal Blog Archive to ChatGPT and see what happens.”
What if we made an AI-powered search that was able to call on and cite the knowledge provided by tens of thousands of law blog authors?
You could then theoretically tailor that to specific locations, perhaps even having a great deal of relevance to an entity like a bar association.
At the very least, we could use the help of AI to identify hot topics and potentially spin up syndicated and curated publications highlighting the blog content devoted to those subjects.
And then, whether it’s within those niches or with something as big as LexBlog.com, we could train the AI to identify what’s a good and relevant post, surfacing those automatically every minute of every day.
Could be pretty cool.
As I mentioned above, our AI Assist beta is off and rolling. If you want to be part of it, let us know. We’d love to have you join us on the journey, have you test what we’re building and help us build this into something great for the law and law blogging.