Libraries: Don't Use a Virtual Contact Assistant, For the Love of God

I spent a couple hours this weekend trying to work through two very fun processes: medical insurance and banking. I figured since I spent the previous week doing jury duty, I might as well stay in the horrid tunnel of impenetrably bad processes and get a couple things done.

I needed help with a couple things, simple things. With the bank, I was supposed to get a new debit card in the mail, and I never did. So I needed them to cancel the first one they sent out and send a new one. Pretty simple, right? You understood what I just typed and, if you worked for the bank, would probably be able to get that rolling?

The newer wrinkle on the ballsack of customer service is the Virtual Assistant, usually powered by…what I guess we’re calling Artificial Intelligence, but I would like to refer to as Artificial Doltishness. Because this thing not only can’t replace a human, it can’t replace a really, really stupid human who doesn’t really know the basics of what’s going on.

What Is An AI Assistant?

You’ve seen this. It’s a little chatbox, usually with a short, human female name to give it that soft touch, retro-futuristic names like Athena or Diana, never names like Agnes or Irma or Milfy.

It sort of pretends to chat with you, but really it’s just a search engine for help topics on the website.

And they’re useless. We all know these things are useless, right?

It may be the case that Claude or ChatGPT can do some amazing, useful things, but these lil’ fuckers just have nothing for you.

The AI Assistants Are Coming To Libraries

If you haven’t already, I can just about guarantee you that you’ll get an offer to add an AI assistant to your website or catalog within the next couple years. It’s happening everywhere, from medical portals to universities to package delivery services.

And I’m begging you to NOT utilize one.

Let’s talk about how a company is going to try to convince you to pay for one.

The 90% Argument

The person trying to sell you on the AI assistant is going to tell you that 90% of the calls, emails, and messages you get are very simple. Renewals, retrieving lost library card numbers, shit like that.

And this is almost certainly true.

But let’s consider the two most likely types of library users who will call.

One is the type of user who is probably not very tech savvy, not going to learn, and just wants to do something as simple as renewing their books. For this person, an AI assistant is useless because that involves just as much typing and computer shit as logging in and renewing their own stuff, things they are unwilling or unable to do.

The AI Assistant does not solve their problem.

The second type of person likely to call or message with a question is someone who is reasonably tech savvy and tech willing, but is trying to do something they can’t figure out.

This is the kind of person I am most times. I know that I need to have the bank send me another card. I know that I need to inquire with the post office as to why we’re not getting a bunch of mail.

And it’s been my experience that, despite knowing exactly what I need, and despite being pretty good at digging through menus on websites, I often need help locating a simple contact form or doing a specific sort of thing. And it’s been my experience that AI assistants are not much help in these scenarios.

When my problem is that I’m not receiving all of my mail, I’m directed to lots of help topics related to not receiving a specific piece of mail, or not receiving ANY mail, but there’s really nothing about receiving some mail.

AI Assistants are not helpful for something that has some nuance to it.

While 90% of your questions might be simple, that doesn’t mean they will be adequately answered by the AI Assistant.

Saves WHO Time?

Someone trying to grab a budget line for an AI Assistant will tell you that it will save the library tons of time.

AI assistants save companies money because they reduce the volume of people who call. But the cost is paid in time by the consumer because it almost always takes longer to get where you want to go with an Assistant.

If time is saved by the staff, but if that time is saved by passing the work back to the users, the community is not saving time, it’s just redistributing whose time is wasted.

In the library world, this means that we’re passing the time costs to users. Which means A) Our community, as a whole, isn’t saving any time, and B) We’ve shifted work from paid staff whose job it is to perform these duties to library users, who are not on the clock when they engage with the library and should not have to spend more time than necessary to accomplish simple library tasks.

Consider The Reasoning

Are you getting an AI Assistant because it’s the answer to a specific problem, or are you getting one because it’s a thing you can get?

Unless you have been having a serious problem returning digital messages, you probably don’t need an AI assistant.

It’s like this: When I buy a new phone, I should be buying a new phone because doing so solves a problem, such as my phone being a completely spiderwebbed piece of junk that can’t even handle being dropped out of a car window and run over several times.

I should NOT buy a new phone if the new phone doesn’t solve a problem, but it’s just something I could buy, and that’s reason enough to buy it.

Technology should, generally, solve a problem. It should not be adopted because it’s neat. It should not be adopted because it solves a problem that SOMEONE is having, but you’re not. It shouldn’t be adopted because it solves a problem you may have in the future if you grow a lot or start seeing a higher call volume.

It shouldn’t be adopted unless it’s solving a specific problem you’re having right now.

People Arrive At The Same Place, But They’re Pissed Off

In my experiences with AI assistants, I get where I want to go eventually, but by the time I arrive, I’m pissed. I’m annoyed. I’m frustrated. I do my best not to take it out on anyone, but I’m human (at least until some servos come in and we’ll see what we can do about this whole “being human” thing!).

I have my doubts that AI assistants are good and effective, but I have NO doubts that they are super annoying.

The One, Correct Question To Ask:

If we do this, when it comes time for voters to decide whether the library should have more or less money, will this make them want to vote for the library to have more or less money?

Maybe Another Good Question To Ask:

I we don’t use an offsite call center now, why would we adopt an AI assistant.

Some Good Questions To Ask The Company Pitching The AI Assistant:

We hold privacy in very high regard in the library. What happens to the data, like the questions people ask, after they end their session with the assistant? Is that retained? Does the AI assistant train on questions it’s asked?

Do you have hard data for us regarding the percentage of satisfied customers who feel the AI assistant adequately answered their questions?

Can you direct us to some sites already using your AI assistant so that we can ask it questions and see how it performs?

What’s the advantage of your AI assistant over someone choosing to use ChatGPT to ask the same questions? In other words, what sorts of things could your assistant do while integrated into our catalog that ChatGPT could not?

If You Feel You Must

Do me and you a favor: When you implement your AI assistant, when it pops up, give patrons the option to immediately be switched over to a human assistant. Track these numbers, see how many people select that option and how quickly.