Why Do We Need Chatbots, Anyway?

Two television advertisements are getting heavy airplay right now. Both feature Matthew McConaughey and Woody Harrelson. In both ads, Mr. McConaughey finds himself in some unfortunate circumstance.

In one of the ads, Mr. McConaughey is dining alone at an outdoor, street-side cafe ... in a pouring rainstorm. There is no canopy or awning, and Mr. McConaughey is soaked to the skin while people around him rush for cover. There are no other diners present. Despite the weather, a waiter delivers a shrimp cocktail plate to Mr. McConaughey's table and quickly retreats indoors.

Mr. Harrelson meanwhile is seated comfortably with two dinner companions across the street at another outdoor cafe. This cafe is crowded and it does have an awning. Mr. Harrelson's dinner companions regard Mr. McConaughey's plight from across the street with concern, while Mr. Harrelson himself taunts Mr. McConaughey.

The other ad features similar themes. Mr. McConaughey is being dressed by a saleswoman in the window of a clothing store. The articles of clothing are increasingly ridiculous – a loud, unbuttoned shirt, a purple velvet overcoat, a floppy red hat. Mr. Harrelson walks by on the sidewalk and taunts Mr. McConaughey through the window, quipping, "You know people can see you out here?"

It would not be surprising for the viewer in either of these commercials to assume that Mr. McConaughey has made poor choices and that Mr. Harrelson is poking fun at him for having done so.

But, that's not what the advertiser intends.

The advertiser in this case is a new software suite from Salesforce that they are calling Agentforce. The software includes chatbot functionality and is targeted at the retailer, not the customer.

I will say that the advertisements are effective. They feature well-known actors in humorous situations, and they caught my attention enough to visit the Salesforce website to find out what Agentforce is. Assuredly, that's one of the goals of advertising. To get potential customers to seek out more information about your offerings.

I'm also writing this blog post about the ads, so they must have grabbed my attention.

Presumably, what Salesforce would like the viewer to take away from these ads is that had the retail establishments deployed the generative AI functions in Agentforce, they could have better informed Mr. McConaughey about their offerings and Mr. McConaughey could have therefore avoided his unfortunate circumstances as well as the embarrassment of being jibed by Mr. Harrelson.

Critically, all of this communication could have happened without human-to-human interaction, saving the retailer time, effort, and money.

That is the overreaching "benefit" of chatbots, and most generative AI isn't it?

Do chatbots really save time and effort though, or do they more often than not frustrate both the customer and the company deploying the chatbot?

In the restaurant example, both Mr. McConaughey and the proprietors knew that it was raining when he arrived, knew that there was no awning, and Mr. McConaughey could have easily asked to be accommodated indoors (or the proprietors could have offered to do so). If there were no room to be accommodated indoors, Mr. McConaughey could have taken his business elsewhere, as he eventually did by crossing the street to join Mr. Harrelson and his party at the end of the commercial.

All of this would have required human-to-human interaction between Mr. McConaughey and the establishment, which apparently never happened. As a result, Mr. McConaughey was frustrated and subjected to ridicule. Could a chatbot really have prevented this?

Companies might be convinced to say "yes" to this proposition. After all, if they can communicate with their customer without the expense involved in human-to-human interactions, they can improve their bottom lines. Does this come at the increased expense of frustrating their customers though?

Apple Intelligence is another example. Besides waking up Siri every time I greet my cat with "Hi kitty!," which is frustrating in and of itself, Apple Intelligence now gives me an AI-generated summary of a new text message in the notification area.

Half the time, these summaries are not a good representation of what is in the text message, making them more than useless (more on this in an upcoming post). Aside from that, text messages tend to be short – often near the length of the AI-generated summary itself. It would save time for me to just read the text message compared to having to digest both the summary and the text message.

Companies aren't blind to these ideas. Yann LeCun, the chief AI scientist at Meta earlier this month told the American Mathematical Society that "the current state of machine learning is that it sucks."

A lot of smart people are thinking and building products in the generative AI area right now. They would be wise to understand the current limitations so that they avoid frustrating their most loyal customers only to save a buck here and there.