There is no shortage of current conversations about artificial intelligence (AI) – both its advantages as well as potential concerns. There are also many more unknowns than definites when it comes to the world of developing computer systems capable of performing tasks typically associated with human intelligence.
In fact, the two-letter AI acronym can be favorably compared to a three-letter predecessor in WWW. The early days of the World Wide Web were often a wild, wild west scenario with few people really knowing what they were doing and limited options to obtain any answers.
Additional similarities between the early days of the internet and AI: Tremendous optimism about the capabilities combined with a clear degree of skepticism as well as a wide range of unanswered questions that fall into the “to be determined” category.
The AI world has been thrust upon the business community in the past few years. The long, somewhat tedious recovery from the COVID pandemic was jolted by the introduction of ChatGPT (a chatbot and virtual assistant developed by OpenAI) in late 2022 and its widespread expansion ever since.
While technology changes were commonplace over the last few decades, the speed and scale of this development, along with its widespread adoption, was shocking to many. It was a very large paradigm shift and flipped the democracy of information access on its head.
This simple fact underscores the impact. Nvidia, which produces the graphics components inside computers, surpassed Microsoft and Apple in mid-June to briefly become the world’s most valuable publicly traded company.
Setting the stage
Let’s look at a few terms before diving into practical uses and concerns (both unfounded and realistic) in the world of customer experience:
- Generative AI and large language models (LLMs) are what most people are referring to when they talk about AI. These systems work by predicting tokens (words, or parts of words) iteratively, one after another based on context and prior predictions.
- Prompting is a way we can interact with the AI systems. By tailoring our input, we can get the LLM to exhibit desired behaviors, such as responding in certain regional prose or taking actions like an agent.
- RAG (retrieval augmented generation) is a way to provide documents or outside information to AI systems. This information can be used as supplemental material for the LLM to use in its response.
- In-context learning is a specific application of RAG. An example would be providing a list of best-practice survey questions to an AI system in advance, allowing for more precise responses to future inquiries about generating survey content.
One of the frequent conversations about AI revolves around its impact on jobs. Extensive research reveals the current capabilities of AI to be on the level of the average high school sophomore. While our young people are more capable than ever, I don’t think we’re ready to turn the future of our organizations over to computer systems with those competencies.
What are some of the AI limitations? Today’s AI falls short in completing complex tasks. If a project is broken down, AI can contribute strongly to processing meaningful units. But putting those individual pieces back into the big picture proves much more difficult.
Other AI shortcomings, at least to this point, include decision-making, statistical analysis, running regressions, and filling in gaps in data. In other words, AI can play a role in producing data for a report, but human expertise is required for verification, analysis, and determining action steps based on the information gathered.
While we again note the unknowns of the future, it is not anticipated that AI will reach this degree of proficiency for a very long time.
Practical uses
So, how can – and should – we be using AI today? Again, a few examples:
- AI can give structure to unstructured data. Your restaurant chain offers a survey after a customer visit. Part of the feedback is that “Bob was an excellent waiter” or “Jill was a great server.” If you have 1,000 employees at over 30 locations, finding the correct Bob or Jill may be difficult.
Providing the system with an employee list and store numbers will allow AI to deduce which server the customer is talking about. AI is taking data and making sense of it. This is a process that can be done manually, but at the expense of time and resources. - AI can be valuable in not only summarizing but moderating content. A service representative in your organization is dealing with a very upset customer. Instead of subjecting individuals to treatment that can be detrimental to their mental health, AI can absorb the information and deduce what needs to take place to address the customer’s concern.
AI is not responding or fixing the problem but serving as a guardrail between the employee and a hostile experience. Human response is still required to resolve the situation. At the end of the day, people are required to make decisions and execute solutions.
And while AI can prove valuable, it also must be approached with a degree of skepticism. Reviewing raw data to ensure results are not being misrepresented is an advisable best practice.
On the horizon
Government regulations are on the way. They will impact how AI systems are built and where they are located.
In addition to the job concerns noted earlier, many worry about privacy and data security. Use of customer data as part of an AI system, however, is no different than putting that information in a database. Leading customer experience companies have already put extensive measures in place to safeguard data – whether it is information stored in traditional methods or used as part of the AI process.
AI may be new but measures to protect data are not.
Currently, some companies are all in on AI while others are ignoring it or treating it as a fad. Taking the latter route is putting the future of your organization in peril.
Go back to the wild west days of the internet. The list is long of those that didn’t adapt and either no longer exist or do so at a dramatically decreased level of influence. Think Sears, Kodak, Blockbuster, and Xerox as just a few examples.
Not getting into the game now could result in a company being left behind a few years down the road.