How much AI is ‘real’ in CallMiner? All of it.

At its core, artificial intelligence (AI) is merely a tool to achieve a result or insight. Just like a broom, Swiffer or a Roomba. A broom is great at sweeping everything, and especially the hard-to-reach corners. It takes work, and is not AI, but tested technology. A Swiffer is more efficient than a broom, takes less effort. A Roomba does an excellent job, takes far less effort, but there are places it can’t reach like the stairs or under my radiators. I need a different technology like a broom or a Swiffer for that. So, if we demystify AI, it is just a tool to do something a person can do with less effort or learn something a person can learn by using the technology. Machine learning (ML), deep learning (DL), natural language processing (NLP), etc. are nothing more than tools with cool names, but like any tool, there is the right one for the job. That is the key distinction.

Because people mystify and anthropomorphize AI, it allows companies to tell you their “broom” is a Roomba, because when you look, the floors are clean. No one knows they were swept by a person, not AI.

Also, it allows companies to point at other AI companies and say “we are more Roomba than they are” because how would the uninitiated know? Both of the floors are clean. If you hear language like, “the AI learns as you go and makes the product better,” it’s worth taking pause – it's likely there is a human behind that somewhere. 

AI solves problems and provides insights for humans. AI’s advantage is a faster way of doing the tasks than human can or might be impossible due to scale. Because it can chew though more data than human in the short time, it can provide useful information. Especially when looking at a more data than a human ever possibly could. So we use AI to abstract a large amount of data where it finds additional useful information. 

The real question should be, “does one company’s solution do that better or more powerfully than the competition?” For CallMiner, our solution, with a logical blend of AI and heuristics, is the best on the market.

How much AI is ‘real’ in CallMiner?

Technically the answer is all of it, but detail is key here. Let’s start with old vs new.

“Old or Pre Computer Aged AI” are those tried-and-true methods from before the current computer age. Linear & Logistic Regression as an example, is a fantastic, supervised approach from last century (technically two centuries ago, but who is counting). It works, it’s powerful, it’s old, brilliantly transparent, and there is every reason to use it when you need to solve a problem that regression solves. Use it when it’s the right tool. CallMiner uses Regression; only a fool would pass up powerful old tools that still solve problems.

“New or Computer Aged AI” let’s pick on our friend BERT, merely as an example of advanced approaches. BERT is a more advanced tool where it tries to link each word to every other words using context. This creates a massive web of connections that’s difficult to interpret, even though you can use the model in a creative way. BERT is a Neural Net that can approximate the way a human thinks to achieve results. It works very well when the answer is unknown, it finds answers.  But it is not transparent, meaning it provides insight, but we don’t exactly know why because a Neural Net has hidden layers, like human neurons that are firing in a specific way based on the training.

But it is not that easy. Organizations demand transparency, so we have to answer the question: “Why?”  We can’t just say, “do that,” we need to provide a root cause or a reason. To do this, we can employ a second algorithm called SHAP. SHAP helps with the interpretation of complicated algorithm by suggesting how much each potential reason contributes to the outcome. It can effectively reverse engineer the Neural Net to expose if the importance of the features that were used for the prediction or insight. Advanced algorithms are still just tools like a Roomba. Do we use BERT, SHAP and Neural Nets? All the time. Would I ever employ them to solve a problem that Regression can solve quickly, transparently and with nearly no computational expense? I would not. They solve different problems.

CallMiner – and any company who actually works in AI – knows this. The key is the experience and knowledge to know when to use each tool.

Our ethos as a company: “Human in the loop”

The real debate here is how much human effort it takes to achieve a solution or find an insight. This is the key to using machines as a partner. We have a central company ethos that our software is a partner to humans. In our near 20-year history with more than 400 customers, we have a proven track record of using a blend of AI, ML and heuristics to provide an easy-to-use product that provides the user an unparalleled level of control and transparency. 

We firmly want you to run your software, and control its use, insights and achieve ROIs we never even dreamed of. But we will always consult with any uses. Machines are not creative at all. The best way to advance knowledge and ROI is to design the powerful components, automate the elements, but let human cognition shine.

This also is evident in our strong adherence to Ethical and Responsible AI. We have a head of Ethical AI, Micaela Kaplan, to ensure we stay true to this north star at a granular level, not just strategically. We have published a lot in this regard and are speaking at the AI Summit in December around AI Ethics.

The easy button

It is important to note, the first ROI you want to achieve has already been achieved by someone else. Thus, this first ROI can be automated by AI and presented as a mere button you push. Great sales tool and marketing teams love selling buttons. It looks like cool AI, and the user (you) learn nothing about a potential software vendor’s business. The easy button gives you what it gives you; it is not ‘your’ button, it is ‘a’ button. So, ROI one is done, dusted and sorted in the initial sales pitch, but rarely do they discuss what’s next. 

Believe us, there will be a new idea, a shift on company direction, and the need for a second, third and fourth ROI. Is your new technology partner going to build you custom AI solutions? Five new buttons to push? Charge you professional services? Maybe. If they design you a button, will they get it right? Maybe. Are they going to crack open their software and let you mess with the source code for real analytics, so you can learn why? Absolutely not. 

CallMiner’s system was designed from day one to allow you to search, analyze, refine, and home in on the ROIs you have not dreamed of yet. We want to learn the positives and negatives in your business. That is 20 years of refinement and marrying AI to human cognition as a partner. We have the experience and are the industry leader for a reason.

Be sure whatever software vendor you choose has experience in advanced ROI and use cases. As example, one of our customers wanted to use software to restructure the workflow of an agent based on previous conversations and journey to increase revenue by 10% with no effort from the operations. Another wanted to derive the insights so that an avatar could walk out to a complex medical bill and intelligently explain the charges in a VR world. Often the response to these type of ideas is bunch of head scratching. If that’s the case, call us. We will introduce you to the companies who have achieved these scenarios, and others, in our system. Use cases and ROIs we could never have dreamed of, but are readily achievable in our software.   

Taking a human out of the loop and claiming a push button solution is folly. Improving your business is work, and worth the effort.