Let’s start with what most of what the collection industry may think AI is – Intelligent chatbots. Dogs are four-legged animals, but all four-legged animals are not dogs. By the way, there is an old test for AI and its called the ‘Turing Test’. It’s simple. Put the expert and the machine in separate rooms and ask them the same questions. If the machine gives the same answers as the human expert, you have AI. Chatbots behaving like agents (even-mid level agents) would be AI but is AI the chatbots that are advertised today?
Let’s try the Turing test.
- You can call an agent and talk to them. Does your chatbot answer the phone?
- Is your chatbot allowed to speak to the consumer? If the state of Tennessee insists that only specially licensed collectors can talk to a Tennessee consumer, do you take a chance by allowing an unlicensed chatbot to talk to that consumer?
- This consumer has written a bad check. Is your chatbot going to take the check option away when a payment plan is offered?
- When an agent takes a check by phone for today, the checking account information is immediately encrypted, stored, and masked for future inquiries. If checks are submitted for processing 5 minutes later, these transactions should be picked up. Does that happen with your chatbot and collection platform?
- You have two companies – Bad debt and early out. Is your chatbot intelligent enough to play different greetings for each company?
- Consumers famously do not want to give an agency their credit card or checking account details. They will also call the day before their payment is due and ask to delay their payment by 2 days. Does your chatbot recognize these situations and offer to take the payment or change the next payment date in those two examples?
- Does the chatbot ask for payment in full and then go to the payment arrangement options, and finally, offer a settlement? How is the settlement offered? Does the bot ask the consumer for their best offer to settle their total balance, as an experienced agent may do?
- Can your bot trigger a text message to management indicating that it just took a $1,000 payment?
- Can you chatbot wear different hats? For example, can it become the interface for someone who clicks on your website link “Make a payment”? Regardless of the way your chatbot was accessed, can the consumer be quickly authenticated without their account number? No? Agents do that, don’t they?
Think about that list. There is much more. There is a great deal of experience a human agent uses in its interaction with a consumer. How close to a human agent’s behavior is your AI?
FACT CHECKING AI CLAIMS
AI is not new. Along with Robotics there have been practical uses of AI since the 1950’s. Long before Elon Musk introduced the Tesla to us, rich kids drove electric cars in their driveways – Probably cars bought at the famous toy shop FAO Schwarz in New York. Why did it take Elon so long to create a practical electric car? Because a toy car was not the same as electric vehicles for commercial use!
A chatbot that does not do what a human can do, will not pass the test for AI. Talking about AI and making use of today’s great computing power is a great first step. But we must not close our minds believing that AI in collections is large language models.
- Is ChatGPT AI? Chat GPT is language model that can generate human-like text based on context and past conversations. To use ChatGPT in collections and call it AI, we must create a product that uses the technology to mimic human behavior. In a practical example, our software will send a consumer-agent conversation to ChatGPT, quickly get back an accurate summary of the call and allow the agent to update the account. In the best case, not a single note was entered by the agent. That’s AI!
- Does ‘Genius Technology Inc.’ that builds chatbots to help collection agencies, become an AI company by changing its name to ‘Genius Technology AI’? The answer is ‘No’.
- A tech company says, “After 6 months of hard work, we have converted our original system into an AI-based product”. Is that possible? Depending on the scope of the product, a complex product could take years to rewrite. AI is not a simple programming change. It is not even a complex programming exercise. It is a different architecture that has to be designed ground up. Speaking from experience, we deployed an AI-based collection platform in 1991. It ran on powerful IBM midrange computers. When we say that creating AI can take years, we mean it!
THE TIME FOR AI IS NOW
We live in an era of tremendous computing power, at a time when a comprehensive AI-driven technology solution will not break the bank. We are well-qualified to say that, and to also add that most companies that adopted our technology enjoyed greater productivity at lower operating costs. Beware. “Sunk-cost fallacies” and unprofitable efforts to extend the life of aging software platforms, can squander great opportunities for change and growth in our industry. This could be a period of auspicious change for the collection industry, but hype and talk will not get us there. Responsible, innovative vendors, an honest search for knowledge and proactive action, will.
_________________________________________________
Quantrax Corporation is a technology company that created an intelligent collection software platform over 30 years ago. They believe the ARM industry has been poorly served by collection technology that has not kept up with the great potential of computing power or challenging industry changes. Self-funded, Quantrax has continued to successfully develop and deploy technology that offers modern solutions to old problems.
Ranjan can be reached at ranjand@quantrax.com
www.quantrax.com – (301) 657-2084