Until recently chatbots were viewed within enterprises as little more than digital receptionists and out of hours customer service back up, operating within predefined scripts and seldom delivering responses aligned with human inquiries.
But thanks to advances in AI, and innovative new applications, interactive agents are evolving into a technology that is beginning to transform customer service-facing businesses.
This evolution has been underscored by the presence of dedicated conferences like Chatbot Summit – located this year at the ExCel Centre in East London. The two-day event, which took place earlier this month, featured key delegates sharing knowledge, experiences, and future predictions.
The event highlighted how chatbots are providing round-the-clock customer support, answering queries promptly and efficiently, improving customer satisfaction and engagement.
Unsurprisingly, AI assistant success stories took centre stage during day two of the summit.
During his keynote, Franz Weisenburger, SVP innovation, Deutsche Telekom, discussed the decade-long journey behind Frag Magenta (Ask Magenta), the German telco’s acclaimed AI-powered chat and voice bot that helps with customer service queries, from internet faults to contract extensions and bills.
“As you can imagine, 10 years is a long time to learn about how to interact with customers, how to use technology in the right way,” he said. “So, three years ago we decided to scale massively, not only in chat, but also in voice, using our partner Rasa.”
Rasa, the technology partner that Weisenburger namechecks, offers an open generative conversational AI platform capable of building text and voice-based bots to scale.
Rasa claims to work at Level 3 of conversational AI, meaning that the bot can understand context – handling things like the user changing their mind and dealing with unexpected queries.
According to Weisenburger, AI assistants continually improve, yet like any tech, they have constraints. The telco exec for instance, noted a 60% transfer rate of Frag Magenta’s cases to human agents. Hence, the need to work on enhancing its offerings with partners like Rasa.
The next challenge for DT, Weisenburger said, was to offer conversational interactive voice response (IVR) to its customers.
“Voice is much more complicated than text,” he explained. “Customers don’t accept latencies, they don’t accept written dialogues, translated invoice dialogues, so we have to solve all this by the beginning of the journey,” he said, adding that the solution is likely to be active by “mid-2024”.
DT also has ambitious plans further into the future. Imaging a bot that is completely informed and aware of an individual customer service worker’s clients, as well as that worker’s responses and decisions. Enough so that, if that person goes away on holiday and a customer makes contact while they are away, the bot can step in seamlessly to help with enquires.
“We envision a digital twin concept — leveraging technologies like robots, avatars, and LLM technology — where we can seamlessly step in for that worker when they are away on vacation,” Weisenburger explained.
Using himself as an example, he added: “Imagine, if a customer typically contacts Franz, they can reach out to the chatbot during my absence, any time, day, or night. In about a year, the chatbot will possess the same comprehensive understanding of Franz. It’s a forward-thinking idea, perhaps in two or three years, but this represents our vision of the future.”
Kevin Lee, chief digital officer at BT Digital, which launched its own AI chatbot, Aimee, last year, predicted at the conference that AI chatbots would soon offer insightful recommendations for business processes on top of their traditional customer service roles.
Lee illustrated this with the case of Aimee, which currently serves diverse functions within the telecommunications giant.
According to Lee, since its inception, technology has been continuously advancing in sophistication by leveraging insights from each interaction. Looking ahead, he claims it will intelligently anticipate specific customer needs, even those that may not currently be met.
“Because Aimee has been harvesting its large language model across millions of customers a day, she will start to know what features we actually need to build for that particular customer,” Lee noted.
Lee also elaborated on Aimee’s ability to leverage insights and patterns from customer interactions to prompt the development of new features. BT envisions Aimee achieving a net promotion score [which benchmarks customer satisfaction] exceeding 80 (which is in the top percentile) by 2025, based on over 400 million customer conversations.
Despite the improvements made to recent iterations of AI assistants and chatbots, there are still trust issues which continue to remain a barrier to adoption. Even the conference’s founder – Israeli tech entrepreneur Yoav Barel, acknowledged this.
“Some companies fear the hesitancy and slow progress in adopting, combined with consumer trust issues which deters them from achieving product-market fit,” he said.
Ethics and bias are pivotal discussions in AI, and chatbots are no exception, according to another speaker, Elizabeth Stokoe, a professor at the London School of Economics and Political Science.
Stokoe noted that with the proliferation of chatbots in service-based encounters, especially with human-sounding voice bots, “there are bound to be user concerns”.
“[It’s] not just about who you’re actually talking to but also what is happening to the data you are giving – and how accurate the tools are in either understanding the written messages you supply to the chatbot or in converting your speech to text – as well as what happens with the data you have shared,” she said.
Stokoe also highlighted that chatbots have demonstrated biases and tendencies to reflect assumptions or propagate stereotypes. “For instance, when referring to a doctor, LLMs may subsequently use ‘he’ pronouns, assuming that doctors are men,” she added.
“This bias stems from the language data they learn from, reflecting human biases present in the text. Consequently, it’s crucial to consider the implications of the information AI tools are learning from on the resulting biases in these tools,” she stressed.
Bot enthusiasts such as Barel, meanwhile, envision a future where everyone has access to their own personalized AI coach, aimed at improving every facet of their existence.
“I believe the next Uber equivalent will be a digital concierge, guiding us in specific aspects of our lives,” he predicted.
This platform, he added, would facilitate various interoperable communication systems, rather than being monopolised by a single company.
“In each of these systems, you’d have the freedom to customise your own AI assistant and these assistants could be provided by different brands,” he explained.
“Just like how Uber, a concept that was nonexistent before, is now a simple app, you could envision a similar scenario for services like a mental health coach, tailored to you and offered by different brands—essentially, your personalised life coach.”
To read more stories on AI click here
Subscribe to our Editor's weekly newsletter
Share
