Despite the hype, AI is still underdeveloped in the area of business solutions and strategy. But research shows that if you use the right tools for the right purpose, AI can actually be a huge time-saver and provide a significant competitive advantage.
In a breakthrough study, Boston Consulting Group (BCG) partnered with Harvard, MIT, the University of Pennsylvania, and the University of Warwick to test the efficacy of generative AI on workplace productivity across 750 consultants at BCG. The findings were surprising: According to the study, generative AI creates powerful competitive advantages in areas where it has a firm competence (for example, in the field of creative product innovation) and dramatically reduces performance in areas where the models are currently underdeveloped (like business problem solving).
The study found that counterintuitively, GenAI models tend to perform better on tasks that require them to “come up with creative, novel, or useful ideas based on the vast amounts of data on which they have been trained.” In these cases, when participants used ChatGPT-4, about 90 percent improved their performance and had a level of performance about 40 percent higher than participants working on the same task without the use of GenAI.
This study also found that extant GenAIs break down and underperform their human counterparts when the models, “are asked to weigh nuanced qualitative and quantitative data to answer a complex question.” In these types of tasks, human participants performed 23 percent worse than those that did not use any generative AI tools.
So what does this mean for you? There are staggering performance gains to be made by using GenAI for the right types of tasks. These tasks should require little to no complex business analysis, be laborious to perform manually, and fall within your realm of comprehension so that you can audit your results manually. To date, the most consistent marketing application we’ve found for GenAI is the rapid generation of SEO topics and content.
If you’re interested in using a GenAI for this purpose, there are two crucial things you need to know before you start:
First, understand that most new GenAI “technologies” are really just static UI wrappers disguising the big general models like Open AI’s ChatGPT and Meta’s Llama 2—fancy bodywork around the same two or three engines. That isn’t always a bad thing: Uber is just a UI wrapper built around Google’s Geolocation API. And in general, static wrappers around AI streamline the process of developing effective prompts, which can be a laborious task in its own right.
That’s what makes a wrapped AI useful for creating highly specific SEO content: It provides a templated, paint-by-numbers approach to prompt engineering that can further speed up the process of working with a model like ChatGPT. The trade-off to this approach is that you’re paying a premium to limit the ways you can interact with a foundational AI. You’ll make more paintings, but you won’t be able to apply your best insights to make them your own.
A good example of this trade off in action is Jasper. Jasper leads you through the content creation process by asking for a series of inputs like topic, keywords, and voice.
It then turns your basic inputs into a complex prompt that it feeds to a large AI model and returns your content. Jasper can produce passable SEO articles faster and with less sweat equity than we’ve achieved using the foundational models. However, we’ve yet to get Jasper to produce a single piece of content that didn’t require manual editing.
And that leads us to the second point you need to understand: At this point in time, AI cannot be trusted to produce accurate, engaging content without human oversight. And if you use a wrapped AI like Jasper, you’ll likely find yourself doing more editing than you will when using the foundational models directly.
When you use a foundational AI on the other hand, engineering effective prompts takes more effort up front, but you’ll be able to produce more refined results in the long run. Plus, as the AI model improves over time, so will your content.
That’s why generally speaking, we recommend that you leverage foundational AI models directly, such as Open AI’s ChatGPT, Meta’s Llama, Google Bard, and Anthropic’s Claude. When you use these models, there are no limitations to your interaction patterns, and you won’t pay extra money to cap the quality of your results.
