Without the code, Conversational AI is another “Black Box”

black box

Most people reading this will be familiar with the term “black box”. But for those that aren´t, it´s simply a phrase used to describe a scenario of obfuscation. One where the internal workings of a technology system are hidden from those that want to examine it´s inner workings.

Funnily enough, in the world of conversational AI this obfuscation is often presented as beneficial to buyers, though this only to proves to be detrimental over time.

Machine Learning is a great example. We all rallied around the hype of these seemingly magical systems that, when trained properly, could make highly accurate predictions. When they worked, that is….And when they didn’t? Well, that’s when the “black box” effect reared its ugly head.

That´s why the team at Artificial Solutions has focused on supplementing ML-based approaches with a highly sophisticated, linguistics-based modeling language. It’s essentially a coding language made for conversational AI language processing and it’s a critical complement to machine learned NLU. Without it, your natural language understanding engine is relegated to the never-ending and highly fickle tuning of the aptly named “hidden layers”.

But Machine Learning isn’t the only black box out there. In the world of conversational AI and intelligent process automation, there are many such systems, with a huge cohort now masquerading themselves under the banners of “No Code” and “Low Code”.

This is by no means to say that no code and low code interfaces don’t have value. In fact, for many functions within CAI, they are the absolute ideal. Take conversation design as an example.

While there exists a growing number of impressive platforms in the market that cater to this burgeoning field, each with their own unique take on the modern conversational design canvas, a very select few take a code-first approach. The industry has proven this to be one area where low code and no code are ideal. There are unquestioned advantages to a graphical palette for this type of work, and the most productive CAI platforms today lean into this GUI model.

That said, the CAI platforms that start and stop at a graphical UI, the so-called pure “no code / low code” types, will ultimately present the challenges of becoming a black box.

Even worse, the black box effect will likely show up at a key inflection point in scaling the CAI solution and present a chasm too far to cross. Solutions become stuck, and the lack of a pro code interface creates a huge barrier to scale.

Another area where this black box effect can prove damning is on the integration front. In particular, the integration of a conversational AI engine into the modern contact center.

Let’s be honest, contact center technology ecosystems are complex to say the least. In most cases, we’re talking about global infrastructures, of different architectures, platforms, interfaces, and applications. High variability and high customization. Years of organic growth and an ever-shifting technology landscape have fueled this paradigm.

The idea that an off-the-shelf, turn-key, connector or adapter is just one plug-and-play away from integrating your call center with advanced CAI is highly suspect.

Furthermore, the notion that a purely “no code” integration can scale to the levels of most large contact centers, where interactions are at an order of magnitude in the millions, if not billions, is laughable.

Bottom line, when it comes to integrating CAI into the contact center, you’re going to need some code. Frankly, you’re probably going to need a lot of it. Building scalable, efficient, and resilient integrations across voice and messaging channels is achievable, and with truly impactful results. Just look at what we’ve done over at Swisscom and Telefonica.

But the work isn’t going to be done by your average business user or “citizen developer”, it’s going to be done by your enterprise architects and your technical developers. And if there is one sure-fire way to break those teams’ productivity and efficiency, it’s to lock all their tools away inside of a black box.

So, the next time you find yourself browsing a heavily touted connectors library where the bragging points are “how many” and “how easy”, thinking to yourself “Hey look! They’ve got our system on their list, this is going to be a walk in the park!”, think again….

Bring your technical teams in and start to peel back the onion layers. Odds are you’re going to uncover some serious limitations and shortcomings and come to the realization that whatever exists out there today is, at best, a kick-start to your own development journey. A journey that will rely on classic application development methods and actual coding skills.

So, what are contact center leaders to do when evaluating the growing marketplace of CAI technologies?

Your best bet is to look for a platform that gives your experts the interfaces and tools that allow them to be their most creative and their most productive. Building high-scale, high-powered integrations into your contact center means arming your experts with pro-code interfaces, giving them deep, programmatic access to the underlying technology, a rich API and SDK catalog, and modern, enterprise-class CI/CD and DevOps features.

We’re proud to consider Teneo at the forefront of this niche CAIP community, but we encourage you to consider any conversational AI platform that can deliver on the critical needs of the global contact center. Just make sure you do your homework.

Oh, and most importantly, don’t be fooled by “easy buttons” that in time just prove to be another frustrating “black box”.