Telecoms & IT

AI in Business in 2023

Language-based AI models may be about to revolutionize some aspects of our business processes—but not all!

Language-based AI models, such as ChatGPT, may be about to revolutionize some things in our business processes—but not everything!

Image credit: Shutterstock / CHUAN CHUAN

Since the launch of ChatGPT on November 30, 2022, the internet has been awash with articles and opinions about this particular form of AI.

People have expressed mixed feelings about this new technology, ranging from awe to fear or even existential meltdown.

This is not yet another article trying to explain, in layman terms, the internal workings of ChatGPT or other large language models (LLM)—a class of AI of which ChatGPT is an example of.

This article answers a more practical question: What are some applications that we can or cannot use language-based AI models for?

Specifically, we will focus on the possible uses of ChatGPT and similar AI models in business process management (BPM).

This is important because an estimated 8-10% of businesses in North America are already using ChatGPT in a commercial way, and the figure is expected to grow over the coming months.

Despite the innovation and ingenuity, however, ChatGPT is but a technology. It is not all-knowing, nor is it truly “intelligent” in the real sense of the word.

As such, there are things that you as a businessperson can outsource to a language-based AI to save time, while there are tasks that AI will spectacularly fail in, causing unwanted complications for you and your business.

Below is an alternating list of tasks that LLMs such as ChatGPT can and cannot help you with.

Making predictions (no!)

While responding to your prompt, language-based forms of AI draw on the information previously fed to them (in the form of a massive corpus of texts) to come up with plausible-sounding answers.

Simply put, they recycle the existing information on the subject and present it to you in a new wording. They cannot possibly have leaps of intuition. As such ChatGPT’s response to a question concerning future events, say, “When will the stock market go up?” may not be very useful.

What’s more, LLMs are by nature language-based and cannot be trusted to handle numbers. So, please do not outsource your finances to ChatGPT.
Do not invest all your capital in, say, the carpet trade, no matter how earnestly an AI recommends it. ChatGPT is not the Oracle of Delphi

Data summarization (yes!)

Instead, ChatGPT has proven to be an excellent tool for data summarization. Taking in a textual data and distilling it into a paragraph, at the length you desire, is exactly the sort of thing that ChatGPT is good at.

“Please summarize and annotate this 4,000-word technical report into a paragraph of 250 words,” or “Please give me the top ten most important economic news items in the Eurozone between 2010 and 2020,” are examples of a good prompt.

You can even ask ChatGPT to give you a synopsis of Shakespeare’s A Midsummer Night’s Dream! And though reading the result provided by the AI model will not be as rewarding as reading the play itself, it will be reasonably accurate.

Creating original content (not a chance!)

As said before, ChatGPT excels at paraphrasing the existing information on a subject. Although it presents the information in a novel form, the content itself will not be novel. It will not be original in the least.

Indeed, LLMs are practically incapable of coming up with an answer to a prompt, unless they have roughly 10,000 documents on the subject available in their data base. Thus, aside from lack of originality, language-based AI cannot be used to deal with new developments such as ongoing news—about which little information is available on the internet.

You are well-advised to stay reliant on your own writing skills for the time being, especially for composing original content, and even routine business letters.

Nevertheless, you can ask ChatGPT to teach you the proper outline of any type of letter or document. Just ask “How do I lay out a profit and loss statement?,” for example, or enquire about the “proper formatting of a book proposal to a publisher”. And then write the actual thing yourself!

Quick explanation (go for it!)

Telling you about the right layout of a letter is not all that ChatGPT can do. It can also be a good instructor whatever your area of work is.

You can ask it for a quick explanation of a concept you do not understand and its results will be generally more informative than a Google search: a Google search will finally guide you to one good answer, whereas ChatGPT will provide you with the distillation of many good answers available on the internet to your question.

You can go further and use LLMs to educate yourself about an entire discipline.

The prompt “Tell me about the must-knows of the science of economics” will give you a list of topics such as capital markets, cost-benefit analysis, decision-making, division of labor, competition, and supply and demand, among much else.

The bottom line is that ChatGPT is a technological tool like any other, and there are right and wrong ways to put it to use. We should also be wary of the hype that accompanies any new technological development…

Back in the 2010s cryptocurrencies were all the rage, with many predicting the imminent death of traditional currencies such as the US dollar. As of writing, the US dollar is still doing very well in the currency exchange market. Similarly, a few years ago, the potentials of “Metaverse” were hugely overplayed by some tech media.

At the same time, being overly suspicions of a new technology is unwise. The 20th century saw the advent or rise in use of electricity at home, telecommunication, television, computers, and the internet, and with each new technology taking the scene, doomsayers were warning us about their threats to society and our way of life.

Language-based AI models, such as ChatGPT, are not going to be much different: they will come and disrupt our current way of doing business, they may even cause a paradigm shift, but ultimately they will be integrated into our processes.

And then, as always, there will be business as usual.

It would be unfair to end this piece without asking ChatGPT about the “truthfulness and merits” of the information provided in the present article “in just two words”. “All good,” was the response.