Should customer service leaders fear ChatGPT?

There have been a number of articles recently about ChatGPT coming to take over your job, company, or entire industry. Many of these have been lightly disguised scaremongering and click-bait. To help cut through the noise and sensationalism, I thought it would be helpful to present a simple framework for thinking about disruption caused by GPT LLMs. 

The framework I use to build product roadmaps is called Jobs to be Done. This model was originally developed by Harvard Business School Professor Clayton Christensen to understand what drives customer behavior and the value they want from products and services. The core approach is to abstract the problem away from the solution. Separating the problem from the solution is critical because customers care about their problems. It’s what they wake up thinking about. It’s what they go to bed dreaming about. They want to get their “jobs to be done”, done.     

The trick with product development is to make sure that you’re solving the right problem in the first place and not just barreling through with a solution to a problem that is unimportant to your customer. To take this out of the abstract, how many people would want a mobile phone that could only send faxes? Or how many people would want a thousand dollar wi-fi capable juice machine that simply squeezed a pouch of juice into a glass?    

That all sounds rather logical; however, it’s human nature to get attached to a solution. They are more tangible than problems, and there can also be a lot of social momentum around a given solution, making it all the more attractive. And that’s exactly where we are with ChatGPT, we have an exciting solution with a lot of buzz, but we are still unsure of which jobs it can solve better, faster, and cheaper. We’ve got a solution in search of problems. 

Putting the ChatGPT cart in front of the customer service horse.

The GPT large language model (LLM) is like a Swiss Army knife for tasks that involve language. And since language is so central to everything we do in customer service (and pretty much any white-collar job), the question gets really interesting and vast. It’s this vastness that has all of us so excited, scared, or somewhat cautious. 

Clearly, there are some companies that are going to be significantly disrupted. Natural language processing (NLP) companies that solve the job of extracting meaning from text or understanding what customers are talking about are gazing into the abyss. Solutions like these that narrowly provide summarization, sentiment analysis, and intent detection have been supplanted. In the contact center, for example, GPT LLMs can solve these jobs less expensively and more easily than hiring a third party to provide the solution. 

But most companies and professions don’t solve problems that are so narrow. Most solve problems where GPT can help, but where it can’t deliver 100% of the solution. In these cases, GPT results in a super-boost to productivity and achieving business goals. 

Software engineering is a perfect example of a profession that seems like it might be disrupted, but where the impact is more of a super boost. Software engineers find creative and efficient ways to solve business problems by writing software. GitHub Co-pilot is a solution that uses GPT to make excellent coding suggestions. However, it can’t do most of what a software engineer does. It can’t frame the human problem that it’s trying to solve and then tackle it with a team. It just writes code in a narrow context. In that way, it’s like a calculator for mathematicians who previously had an abacus. The result is an incredible boost in productivity that supercharges the engineer who wields it. 

John Venn, you’ve done it again.

So if you want a framework for thinking about how disruption caused by LLMs will impact your company and your job, think about it like a Venn diagram that maps the problem space that interests you to the solution delivered by LLMs. 

When there’s 70% or more overlap with a company or profession, it’s likely to be significantly disrupted. If the overlap is more like 20%, those individuals and companies who are early adopters are likely to leap out ahead of the market. All companies will eventually baseline to the new solution so it won’t be a long lasting opportunity, but it will favor the fast movers. And where the overlap is nil, there’s clearly no disruption ahead at the hands of the LLM.

Using the framework above, you can see why a lot of the recent scaremongering doesn’t make too much sense. Teachers, for example, are far from disrupted. Whereas they might need to adapt their teaching techniques to the new reality of text generation, like they adapted to calculators years ago, the job that they solve is so much broader than what the LLM provides. So they can expect the LLM to supercharge their work.

This doesn’t mean that the LLM is not disruptive on a day to day basis. Of course, figuring out how to incorporate text generation into the status quo is going to take some time and is non trivial. But on the whole, the profession of teaching solves for a job which is far outside the circle of what an LLM can do. The same goes for customer service, sales, and IT. My bet is that when you draw out the Venn diagram for the problem space that’s relevant to you, most of you will fall into the super-boost category.

To find out more about how ChatGPT will impact customer service, download our new ebook, ChatGPT and the contact center of the future.