Since we first adopted ChatGPT as a content marketing team at Wrike, a lot has happened, to say the least. Not only has the tool itself evolved, Wrike as a company has embraced AI — using it internally to increase productivity and fusing it into our product as well, with a whole new set of generative AI features.

As a manager in the content team at Wrike, when first asked how our team planned to multiply content production using AI, my first thought was well, I have no idea. When I recovered from my initial panic, I had two more thoughts. One, AI was really here, and two, why not embrace it and add it as a tool in our toolbelt to see what we can really do? So we dove in. 

Here’s what we’ve done, what we know now, and where we plan to go.  

How we’re using ChatGPT now

Those I’ve talked to in other content marketing teams outside of Wrike have mostly said the same thing: Their company has adopted ChatGPT and their instructions have been to use it in all content they create to produce work faster. This mostly seems to come as a blanket instruction, without much education or strategy regarding the best ways to use the tool.

After having these conversations with others in the industry, I definitely appreciate the approach Wrike’s leadership has taken on generative AI so far when it comes to content creation. We’ve been given the opportunity to test and understand the tool, and be the leaders when it comes to setting and evaluating our AI content creation strategy. 

So what have we been up to? Here are some of our top initiatives using ChatGPT for content creation:

  • SEO: Our approach to SEO content creation has totally changed with the introduction of generative AI. Why? Number one, SEO content is largely programmatic and an AI model lends itself well to something like this, something with a formula to how it should be created. Since SEO content is always a part of our strategy, it’s also a great place for us to test different methods of content creation using AI. We’ve been running several tests involving increasing our new SEO blog production and we’ve seen an exponential increase of our SEO blog production by experimenting with both AI and ChatGPT. 
  • Gated asset production: With the use of AI, we’ve also been able to multiply our gated asset production. These types of assets are often more human-led since the goal is to go in depth on a topic and provide value to our readers, but we’re finding ChatGPT can be very helpful when it comes to repurposing existing content into different formats, providing an outline, or creating draft content for sections of larger assets.
  • Promotional kit automation: For a long time as a content manager, I’ve wanted to create promotional kits containing pre-approved promotional copy stakeholders can use in their ads, emails, and social campaigns. But, as is a common pain point for a lot of content teams, we’ve struggled to find the time to do these on top of our regular tasks and requests. With the addition of ChatGPT to our toolkit, we’ve been able to start testing the creation of these promotional kits with AI, taking much less time from our writers than it would have before.  

What we’ve learned

While I could maybe write 10 blogs on what we’ve learned so far, I’ll keep it to a few top points.

  • Test, but don’t forget to evaluate: Don’t just stick with one or two ways you use AI, but test and retest, and then don’t forget to evaluate the results of the content you’re producing. You want to know what’s working and you want to be spending your time in the most valuable ways. 
  • It’s probably not the right time to lean all the way into one method of content production: Based on our tests and retests, we found that human-written blogs still largely outperformed AI-generated blogs. However, we were impressed with how many clicks AI-generated blogs were able to generate without any human intervention. So, while it’s not the right time to go all AI as it doesn’t compare to the performance of our human-led content so far, we’re finding that it’s been a nice augment to the performance of our content overall.
  • Don’t make your writers use AI in everything they do: As discussed above, it’s not time to totally cut out the human voice. Instead, we’ve been thinking about content production in terms of tiers, ranging from tier one (mostly human-written) to tier three (mostly AI-written). As our data has also shown us, we’re currently finding the highest value in producing a mix of these tiers. This also allows our writers to still flex their very valuable creativity to provide the thought leadership voice to Wrike, while embracing innovative technology to supplement their work in other areas.

Where we go from here

This year, I attended the AI 4 conference in Las Vegas. When it comes to using AI in marketing, the big topic of the day is personalization. How can you use AI to create the exact right piece of content to deliver to the right person, at the right time? Using AI to scale content creation gets us one step closer in that journey, as we’ve been able to multiply production essentially overnight, giving us a much wider sample size to see what’s resonating with our audience better. 

From here, our plans are to continue to test and measure. Our ultimate goal is to find the right balance of AI content that augments what our team does, doesn’t dilute the Wrike brand, and ensures we get the most value out of working with our new robot friends.