Challenges in Enhancing Conversational Assistants | Uniphore

Challenges in Enhancing Conversational Assistants | Uniphore

Kumaran ShanmuhanBy Kumaran ShanmuhanVice President, Industry SolutionsUniphore
2 min read

Numerous discussions with Artificial Intelligence and Machine Learning experts have revealed why making enhancements to conversational assistants is such a challenge. Well-known companies have shared, and this is what I’ve learned from the largest and most innovative.

Even companies with endless resources, innovation labs, and highly skilled teams can’t seem to solve the enhancement conundrum. Conversational Assistants require a lot of development and resources depending on the approach your organization takes (I am familiar of an enterprise who required 1,200 on-site developers dedicated to their bot project). This is a route to consider, but I think there are more efficient paths to success.

Next Level of Conversational Assistants

The truth is that customers have high expectations. They want a truly interactive experience with a machine that understands their intent, solves their problem, and offers the next best action. Most bot deployments today start with the regurgitation of FAQs based on a string of inputs. This approach is straight forward and the main reason why we see so many bot vendors in the market. The next level of Conversational Assistants is one that is integrated with back end systems, continually learning from the data it collects, improving over time and personalizing the experience.

Once data has been gathered from customer interactions, what a company does with that info is crucial. This is the driver of customer-journey mapping as well, but now a conversational assistant is in the mix that’s using the most prolific Natural Language Processing and Understanding (NLP and NLU) on the market.

Conversational Experiences with Dialogflow

To create these new interactions, the key is to work with a vendor that has a low code, drag and drop designer. If your bot vendor doesn’t have a designer, it takes significantly longer to develop a conversational assistant. This would be the case if a company were using cognitive services like Watson or Dialogflow on its own. These cognitive services do a great job of intent recognition and replying with static text. However, to effectively service a customer, companies need more than that.

Companies need integrations to back end systems, branching business logic, etc. to personalize the conversational flow, and adding this to Watson or Dialogflow or Microsoft Luis, etc. requires code, and lots of it.

Some companies have a designer, but that means using its own proprietary NLP. This is risky because while an incumbent vendor may have its own NLP, the odds of the NLP dealing with the volume of interactions compared to Watson, Dialogflow, Facebook, or Microsoft and receiving the same recognition rates is unlikely.

The truth is, no small company is going to keep up with the innovation of cognitive service providers like Google, IBM, Microsoft, Amazon, or Facebook. Less data and development resources make a big difference, especially when customers have complex needs.

So, what’s the key to successfully enhancing a conversational bot to the point it becomes a full fledge Customer Assistant? The solution is two-fold: 1) work with the most renowned cognitive services company and… 2) utilize a vendor that has a designer for developing enhanced conversational assistants.

[About the author]Kevin headshot Kevin is an advocate for autonomous customer experiences and quick customer resolutions. He is driven to help enterprises realize their full potential, by educating themselves on what’s possible with technology. He has a background in marketing, public relations and advertising, and has a firm belief in the mission of Uniphore. He is an Atlanta native who loves competition, and is passionate about his family, his work, and his dog Peaches.

Table of Contents