Imagine this. Someone writes a detailed brief for a new campaign. It's handed to a copywriter who writes the copy. That copy then goes through multiple reviews and changes by a minimum of two other people, each one making changes based on their experience and "gut feeling." Finally, it's approved, and it goes live. Can you predict how it will perform?
I'm not suggesting that copy will automatically fail to perform well due to the complicated process it went through and the number of people who touched it. It could quite possibly perform very well. Or it could perform mediocrely, or it could completely fail. It's a process, and if not backed by reliable data, a guess. Can AI improve this process?
The answer is yes, it can, and there are a few solutions available that have been able to show that. I wrote about one of these types of solutions several years ago (remember Albert?), so they have been around for a while. But they haven't hit the mainstream yet. I had an opportunity to see another one - Phrasee - this past week. Here's what I found.
Phrasee was founded by Parry Malm, Neil Yager, Ph.D., and Victoria Peppiatt in 2015. Malm started his career in marketing, writing, and testing copy. He tried to define a rule-set for good copy but always failed. He joined with Yager (who developed the Phrasee AI engine) and Peppiatt to work on a solution to solve the problem.
Phrasee went to market, educating marketers that there was a better way to create copy, suggesting that AI can augment existing talent.
Malm showed me how Phrasee helps improve marketing copy. He talked about how hard it is to predict what language will be effective at scale, which why the approach described above is challenging. "It's a human problem, not a machine problem."
With Phrasee, you can enter a short brief into a web form, and the system will generate a set of options for your copy. You can pick the one(s) you want to use and export them into your marketing campaign tools (Salesforce, website, marketing automation, email marketing, etc.). The results are then fed back into Phrasee, enabling it to learn what works progressively and focuses on that language and copy.
Malm said Phrasee democratizes access to high-powered copy because anyone can use it to help build copy for their campaigns.
A lot is going on behind the intuitive user screens. Phrasee has a Natural Language Generation (NLG) system that, with the help of a team of computational linguists, generates a language model unique to each brand that uses the system. A brand can have more than one language model depending on their requirements. This is the part that takes the most time to get set up, but it's worth it when you realize it's designed to match your brand voice and guidelines.
When you enter a brief, Phrasee generates thousands of potential copy ideas - whether it's a headline, a subject line, or something else. It then takes all those copy variants and puts them through their Deep Learning Engine to rank the best ones in order of predicted performance. It then shows a sample from that ranking for the user to choose from.
The user can remove the ones they don't like, and have them replaced by new variants. All copy follows the brand voice, none of them sound robotic. If they want, the user can select the variants they want to use and put them through an additional review for brand voice or legal/compliance. Still, Malm said that doesn't happen very option, mainly because the language model already considers these things.
AI doesn't replace humans for all content
In a piece on Marketing AI Institute, Yager wrote about the misconception that using AI in marketing will put all copywriters and content writers out of work.
Make no mistake, AI can do a lot. However, modern AI systems excel at narrowly defined tasks. In the case of Phrasee, our system specifically generates short-form marketing copy. In some situations, it is better at this task than humans. However, we can't flip a switch and use the technology to write a blog post or an ebook.
Artificial intelligence and NLG have evolved and improved dramatically in the last two decades. But, today, most AI tools do just a couple of things well. They're not general systems that beat humans at everything.
We all want better copy
It's never a bad thing to leverage tools that help marketers create better campaigns; AI is no exception. But the tools have to be built well.
One of the benefits of Phrasee is that it doesn't create language models by scraping the web, so there is less chance you will introduce language into your copy that could damage the reputation of the brand. Also, because it is continually learning at a speed that is much faster than we can learn as humans, the copy should improve quickly over time.
You would think that a tool like Phrasee would be in every marketer's toolset. Malm believes it should be a foundational tool in the marketing stack - like Photoshop is to designers. We aren't there yet, even though there have been versions of this kind of application available for awhile. I suspect it will come in time as more marketers look for opportunities to speed up the time to get campaigns out the door, allowing them to focus their efforts on other projects.