We had a great time at ACA Annual, and we were super excited to talk about generative AI on the Innovation Stage.
It's cute that Chat GPT can write funny limericks, but you need real help - with staffing shortages, FTC Safeguards Rule compliance, complaints tracking and management, and cutting costs to stay competitive as agencies combine. And you've been burned by technology overpromises before - with faulty speech analytics and solutions that don't understand finance.
But what if AI could actually be helpful?
We asked our Events and Community Manager, Katalina Dawson, to lay out a few AI basics and then talked about how AI can make a difference in your business today.
You can check out the video, but the ACA floor was *hopping*, so you might just want to read on for the transcript.
If you guys read the session description today, what we're going to be going over is I'm going to give a little crash course in AI. In particular, we're gonna go over generative AI. And then we're going to go over the past, present and future of how it has been used in the industry.
Now, before we get into that, I want to see a show of hands. Yes, this is gonna be a little bit of participation. Who here is very familiar with AI? Who has used it, has had experience with it - anybody?
Okay, so keep your hands up if you've used ChatGPT, and typed in something like, "Okay, tell me a limerick about a lamp." So several of you are familiar with that.
What about something like, DALL-E? There's a bunch of programs, but DALL-E does images. You can say, "Show me a cat jumping on the moon." Okay, several of you.
And then there's other ones, there's videos like Synesthesia, where you type something in and you have an AI person in a video, anybody use those? Okay, so still a couple of people, so everybody's relatively familiar. And if you're not, that's okay, we're gonna go over everything, we're just gonna review it.
Those are all very fun. But what AI can actually do for you at work is totally different. And that is generative AI, all those things I mentioned are forms of generative AI and generative AI is just like it says: It's AI that generates something new.
And let's talk about how that works today in the collections industry. First is ML - there's gonna be a lot of acronyms. Whether you're familiar with this, if you're not familiar with it, we're just gonna go over them make sure everybody's on the same page. ML is machine learning, which is algorithms that learn from data. And I know it's technically under generative AI on this slide. But it's not a part of generative AI, it's actually its own thing.
Machine learning is what generative AI completely leans on in order to do everything it needs to do, In order to generate anything new, the machine does need to have that ability to learn. And that brings us to the next part that matters to this industry, which is Natural Language Processing, which will take, analyze, and process language. Very different than what's been used in the past. It's trying to break down the nuances of our speech, our syntax, all those sorts of things.
And this is why I have the umbrella. So under machine learning is NLP. And under NLP is what we primarily use. And that is LLM, which stands for Large Language Models. It is a subset of that NLP, like I said, and the way it works is we take the machine, and we say, okay, machine, read a lot. And then when you're done doing that, read about two times that, and then when you're done doing that, read about 10 times that.
It is taking in massive amounts of data. And what it does from there is it breaks down the data, to recognize patterns - big patterns, small patterns. And all of these things weave together so that it can understand the most basic terms, and "If this, then that." But it gets so much more complex than that, because all of the patterns are weaved together. And that is the basis of machine learning in order to create something new.
A great example of this is like ChatGPT creating a poem. It goes, "Okay, I know poems, most poems, rhyme all of the time," and it goes on from there and just creates a whole poem like that. So.
Why generative AI? Why are we using this in the industry? Let me get something clear. First and foremost, we still need people. As much as AI has advanced, crazy amounts in the past couple of years, people are still super important. We can do things that machines absolutely cannot do.
Let's have a machine do what a machine is good at doing, things that humans aren’t as good at (or don’t like), and let humans do more of what we’re good at (empathizing, problem-solving, connecting).
So an example I like to use for this is a machine - like our machine at Prodigal has listened to over 300 million phone calls. I cannot imagine having a single person sit there and just listen to 300 million phone calls. Instead, AI does that work: listens to it, pulls those insights, and then from those insights we can use that data to improve.
Essentially, AI is assisting where humanity lacks. And I think a great quote for this is something that our CEO actually said. He said, "Anything you are doing several times a day or that you spend several hours in a month doing is something you should figure out how to automate." And that's from Shantanu Gangal.
So, now we're going to get into the bulk of the rest of this. Where have we been with AI in the industry? Where are we now? And where are we heading?
Starting with hiring. Because we've seen a huge growth in this space, and we're using AI to streamline things like repetitive tasks, where they're going over resumes, and they're making sure the people you want to see are the people that come in front of you when you're hiring.
And then humans come in at the back end, so that we can analyze, we can actually make that human connection and find out okay, do they fit within the culture? And that's a great example about how AI is used in a very human aspect of the industry.
Training is the next one. And we have moved far beyond things like scripts trackers or like a ball bouncing along your script. Now the AI can recognize patterns, and what it can do from there is it can simulate calls to help train agents. And even further than that, it can recognize patterns from all of that data is taking up all the successful agents.
And now it knows, "Okay, the most successful agents do this, so we need to take that information and help train your agents to be better."
It's gotten even more advanced than this but before we get into that, I do want to mention something real quick, there is a bit of a "catch" with current AI. And I use the quotes there for a reason. So you've likely heard from either your phone systems or voice analytics tools that they have added an AI enhancement to their tools.
You want to be careful anytime you hear something like that, if it's like "a small AI enhancement." That's just a bolted-on enhancement. So it ends up being not as much like artificial intelligence, and more like simple automation.
True artificial intelligence, true AI, is built from the ground up, and it's trained on large, large quantities of data to actually learn and become better, faster, and continuously change. That is true artificial intelligence. And that's a very important distinction.
So now we're gonna get into a little bit more complexity. What we're doing now.
Operations. Day-to-day in the collections industry, you've got calls, you've got emails, you've got texts. And when you bring in the AI aspects of that machine learning, large language processing, these can drastically help you unlock new levels of efficiency.
Things we're seeing: If you have custom AI mine the data - you already have the data, but you're not going to sit there spending hundreds of thousands of hours mining it. So again, pass it on to the machine- let it mine our data. Fantastic. From there, you extract highly accurate insights that deliver things like stronger email and text message open rates.
You can also have real-time, context-aware agent prompts. And what I mean by this is when an agent is going through a call, they're not just hearing, "Okay, this is what worked in the past for another agent." As you're going through the call and something happens, the context of the situation is understood by the AI that is listening in. And based on that it can say, "Okay, this conversation is going this way, here are the next steps that are gonna make it the most successful."
But then it continues on to work with you. So then if the conversation changes again and the person you're talking to throws you some other curve balls, it has learned so much from different patterns and sequences, it can know, "Now that the conversation is going this way, the next best path to go through is these steps."
It's listening to the entire conversation as a whole. Knowing from all those different patterns and guiding you through. It's actually a really amazing thing.
And as these models get more and more accurate, we're seeing what's coming next.
What is coming next is things like strategy. All segmentation and treatment strategies will be adopting these tools in the next coming years. And all these changes - agent and agency assignments, settlement and legal strategies, risk management policies - will all feel improvement in both efficiency and effectiveness as a result of utilizing AI.
The next one - my favorite - is forecasting. So forecasting, I'm going to use an example.
By looking at borrower and agent behaviors, we can design highly predictive models, which recommend things like faster or slower assignments. And it can also do things like legal and debt sale strategies and help figure that out. It's really incredible with these models.
Then when they gain access into the aggregated data across the creditor agency and debt buyer agency networks, the entire system is going to get better, faster, and cheaper for everyone.
It's a really, really exciting place we're at right now in the advancement of AI. At Prodigal we're very excited to do things like this with clients. And I'm super excited to see this many people in the audience learning about AI, because you guys want to level up and be part of that next, huge advance, which is amazing, and understand that things like the difference between a bolt-on enhancement and true artificial intelligence.
.svg){kind=icon}
-min.png)
.jpg)
-modified.png)
.png)
.webp)
