During COVID, corporations tested chatbots and conversational AI systems extensively as they looked for ways to continue operating despite lockdowns. The technology performed better than anticipated, and it is now poised to make a significant breakthrough in 2023 as businesses seek to build on previous successes and achieve new levels of office automation.
Before 2020, most conversational AI deployments were in the pilot or proof-of-concept stages, according to Hayley Sutherland, a research manager for IDC who monitors the market for conversational AI tools and technologies. But because COVID offered a “extreme test case,” businesses were able to apply conversational AI with success.
She tells Datanami, “I believe that conversational AI has earned this place in the previous couple of years as a popular corporate application. “Conversational AI is now really enterprise-ready.”
With hundreds of suppliers creating a range of tools, technologies, and platforms for everything from first-generation chatbots to the most advanced conversational AI systems, the market has expanded swiftly. Over the past few years, thousands of successful deployments have demonstrated that conversational AI can provide 24/7 support in addition to a favourable financial ROI.
The early days of rule-based chatbots, which many users found annoying, have long since passed, according to Sutherland. The development of massive language models as the backbone of conversational AI deployments, she notes, including OpenAI’s GPT-3 and BERT (open sourced by Google).
Some individuals are suddenly recognising, “Wow, this is smarter than I thought it was, this is better than I imagined it would be,” as technology advances, according to Sutherland. The development of deep learning and machine learning in these fundamental large language models, as well as the open sourcing of those huge language models from some of the largest vendors and research teams worldwide, have played a significant role in this.
Many conversational AI deployments today are supported by large language models, but there are many more tools and capabilities that allow businesses to provide a finished product. A big team of developers would have been needed to create a functional conversational AI system before the epidemic.
However, many low-code and no-code conversational AI platforms have since appeared, and Sutherland claims that these can be useful in assisting businesses in implementing conversational AI without having to make a significant investment in highly qualified data scientists. That does not, however, imply that businesses can successfully implement conversational AI without any qualified personnel.
Data scientists may be a part of the teams creating successful conversational AI applications, according to Sutherland, who notes that conversational AI providers are increasingly releasing these tools. However, even if they do, they also need to include a line of businesspeople who are aware of what constitutes a good conversation. What details must be known by the bot in order for it to respond to queries?
She says, “I think there are solutions out there that enterprises can utilise to quickly deploy conversational AI even without data scientists.
A terrific moment to invest in conversational AI is right now because there are so many possibilities available to businesses. It’s crucial to understand that there isn’t a universally applicable solution, and what works for one business might not work for another, according to Sutherland.
Analyzing the level of data science talent present at the organisation is the first step in determining the best course of action, according to Sutherland. Companies may require a different calibre of skill than businesses that decide to collaborate with a vendor to develop the application if they wish to construct the entire conversational AI system themselves. The availability of pre-built templates that can kickstart a project is also influenced by the company’s industry, according to her.
Sutherland continues, “I think those are factors to take into account because there are a variety out there. “Some platforms offer a wide range of testing and monitoring features, which may be preferable for a company with a significant developer presence. Some people might concentrate more on those low- and no-code solutions and how to make them operate seamlessly with the rest of the business workflow.
In conversational AI, the availability of training data is another key factor. According to Sutherland, some vendors may provide training data and pre-trained models for particular sectors, while in other instances, the customer may need to provide their own training data to adjust the large language model to function in their particular business.
According to Sutherland, “Low- and no-code tools, in combination with pre-trained models—which some vendors are offering, that are essentially pre-trained for certain industries—can provide those quick starting points for organisations with greater accuracy out of the gate without necessarily having to hire a whole team of data scientists or even one data scientist.”
Conversational AI has a wide range of possible applications; it is not just restricted to substituting for or enhancing human customer care employees. How much further training and tuning will be required will depend on the variety of use cases and the industry’s specifics.
According to Sutherland, a conversational AI system being developed by a biotech company to aid in the production of novel substances will likely require far more detailed data than, say, a mattress retailer.
According to her, “I think there will always be some level of tuning to get to a particular level of precision.” “I think the question is to what extent does an organisation do that in house vs having the vendor assist with it, and how automated is that?”
Companies should also be aware that while some conversational AI platforms are made for voice and can be integrated with IVR systems used by human agents, others are made for digital channels like the web and mobile, and some can even do both.
New issues and challenges have surfaced as conversational AI becomes more widespread. Choosing the metrics to measure the conversational AI’s success and the effects it is having on the business is one of them, according to Sutherland. Another issue is making conversational AI accessible to businesses with limited resources because it consumes a lot of processing power, especially some of the most recent large language models.
Finding the correct balance between conversational AI and humans will become increasingly crucial as adoption grows and businesses learn more about its advantages and limitations, according to Sutherland.
“AI won’t necessarily be able to take the position of every single human I have. That won’t make all of my troubles go away. In fact, she claims that it might even produce new ones. As a result, the ideal way to employ AI might be to understand how to use it to supplement human labour. We can then use this to our advantage so that AI does what it does best while humans do what they do best. And I believe it is a balance that we are really starting to see businesses achieve, particularly in the past year as they deal with what is being referred to as the Great Resignation.
While some vendors of conversational AI platforms merely offer technology, others purposefully include human operators in the mix. For instance, the conversational AI company Simplr may aid a business in implementing conversational AI while also providing a staff of human agents to help with customer service. With its “human cloud,” the company hopes to improve the human operator’s capacity for providing excellent customer service while also providing real-time training for conversational AI algorithms.
“I believe that it will be crucial to consider where AI functions best. Where does a person function best? says Sutherland. And I believe that Simplr has the ability to offer that to its clients by combining the provision of AI and human services as well as the enhancement of their own human network with AI.
