Most of us have been there — frustrated on a hellish customer support call with seemingly infinite hold times, repeatedly explaining the problem to multiple agents who, like robots, thoughtlessly read from a script.

Chatbots, which are in the process of going mainstream, promise to help make some customer service interactions more seamless, but they have their own quirks. Companies developing chatbots for Facebook’s Messenger had a 70% failure rate when dealing with customer service queries, according to Harvard Business Review.

But the chatbots of the future will be smarter, leveraging Internet of Things technology and advanced artificial intelligence to overcome consumers’ biggest gripes, said Kirk Borne, principal data scientist and executive advisor at Booz Allen Hamilton, speaking at an Amdocs press event. And as they grow smarter, IoT chatbots could help telecom companies grow more intelligent at diagnosing service problems, helping them cut down on sending service trucks to consumers when it isn’t necessary.

Contrary to what many people think, most communications service providers aren’t looking to simply replace human customer service agents with bots. According to Forrester research commissioned by Amdocs, 19% of such companies are hoping to replace human customer service jobs with AI. Only 10% believe AI can help them replace a sizable number of workers. Most of these companies are hoping chatbots can help them improve customer experience, using human agents for more sophisticated problems and having bots help out with simpler queries.

Chatbots, however, have a ways to go before they win over the majority of people. At present, 83% of the public would opt to speak to a human agent, based on the Forrester research. It’s only a matter of time, however, before chatbots and digital assistants become more adept at handling complex requests. Their inability to do that is currently their biggest shortcoming, according to the Forrester research, which surveyed 7,220 consumers internationally. In addition, improving such bots’ ability to grasp context could help offer custom interactions while also understanding human emotion, which are their second and third biggest problems.

A conversational bot today, as well as many human customer service agents, start from zero, asking people for the same information at the beginning of an interaction. “That is very frustrating,” Borne said. But as today’s Internet of Things advances to what Borne terms “the Internet of Context,” such bots hold the promise of closing the gap with low-level customer service agents.

[IoT Data & AI Summit demonstrates how the next generation of IoT and AI technologies will converge to unlock the intelligence of things. Get your ticket now.]

This IoT-driven context includes additional information surrounding an event, a person or an activity. “It could be where you are located, the weather or the news,” Borne noted. All of those factors could influence how an IoT chatbot interacts, in addition to historical information about the person communicating with the bot. “Has the person already called 10 times with the same complaint? … There are lots of things bots can skip [in an interaction] once they understand the context.”

Borne notes that the holy grail of IoT chatbots is not just being able to help people solve complaints but being able to delight them with high-quality service. An example could be a chatbot using data to intuit why a customer is reaching out in the first place. “You build delight into your customer experience by giving customers more than they were anticipating,” Borne said. “That helps build loyalty. That comes back to having a little more information in the conversation.”

Borne acknowledges, however, that there is a fine line between bots that wow users and bots that creep people out. “Bots don’t need a lot of information,” he said. “[The interaction] should be personalized but not personal.”