Skip to Content
Artificial intelligence

The pandemic is emptying call centers. AI chatbots are swooping in

Covid-19 is accelerating job losses in an industry that was already automating work at a rapid pace.
May 14, 2020
switchboard operator
Getty

Brian Pokorny had heard of AI systems for call centers before. But as the IT director of Otsego County, New York, he assumed he couldn’t afford them. Then the pandemic hit, and the state governor ordered a 50% reduction of all government staff, forcing Pokorny to cut most of his call center employees. Meanwhile, inbound calls were rising as more residents began seeking reliable covid-related guidance and medical information.

So Pokorny picked up the first solution that landed on his desk: Watson Assistant for Citizens, which IBM had started offering to governments, health-care organizations, and research institutions in a 90-day free trial. Within days of his signing up, the Watson team helped him deploy a chatbot to address callers’ most common questions, such as how to identify symptoms or how to get tested. The software also made it easy for him to update and expand the chatbot’s responses as queries evolved.

As the coronavirus crisis has dragged on, understaffed government agencies, grocery stores, and financial services have all scrambled to set up similar systems for handling a new influx of calls. IBM saw a 40% increase in traffic to Watson Assistant from February to April of this year. In April, Google also launched the Rapid Response Virtual Agent, a special version of its Contact Center AI, and lowered the price of its service in response to client demand.

While call centers have long been a frontier of workplace automation, the pandemic has accelerated the process. Organizations under pressure are more willing to try new tools. AI firms keen to take advantage are sweetening the incentives. Over the last few years, advances in natural-language processing have also dramatically improved on the clunky automated call systems of the past. The newest generation of chatbots and voice-based agents are easier to build, faster to deploy, and more responsive to user inquiries. Once adopted, in other words, these systems will likely be here to stay, proving their value through their ease of use and affordability.

IBM’s and Google’s platforms work in similar ways. They make it easy for clients to spin up chat or voice-based agents that act a lot like Alexa or Siri but are tailored to different applications. When users text or call in, they are free to speak in open-ended sentences. The system then uses natural-language processing to parse their “intent” and responds with the appropriate scripted answer or reroutes them to a human agent. For queries that can’t be answered automatically, the algorithms group similar ones together to show the most commonly missed intents. “The nice thing about the technology is that it somewhat learns what types of questions are being asked, so we can plug them in; we can program it after the fact,” says Pokorny.

The platforms are proving popular among a range of organizations, especially those with limited technical resources. Small and midsize government organizations, including the city of Austin, Texas, and the Czech Ministry of Health, have used Watson to build chatbots that provide information about covid testing, prevention, and treatment. The Oklahoma Employment Security Commission has used Google’s virtual agent to help field over 60,000 daily calls related to unemployment claims. Health providers like the University of Arkansas for Medical Sciences and the University of Pennsylvania’s medical school have worked with both platforms to develop patient triage tools that help them administer timely care.

The goal of the systems is to get as many queries as possible answered by the virtual agents before any calls are handed over to humans. This reduces the burden on call centers and the wait times for users. It also reduces the need for expensive human agents.

A week in, Pokorny was impressed. “I see us using this technology well beyond the crisis,” he says. Right now, his county is using the chatbot only to field covid-related questions, but he plans to deploy a voice-based version that becomes the first point of contact for anyone dialing in about any issue. Watson’s pricing after the free trial is over is also “inexpensive,” he says. He demurs about what this means for laid-off staff and whether they will regain their jobs after the pandemic. “We’re fighting this from two angles,” he says. “One is obviously the health crisis. But once that’s over, we’re going to have a tremendous economic crisis. I see some artificial intelligence maybe helping us.”

Many other organizations will be facing similar pressures to keep their operations lean, so it’s possible that call center jobs already lost may not come back. But perhaps the good news is such jobs won’t be eliminated entirely: algorithms are still far from understanding the full context and nuance of conversation, so rarer cases will have to be reserved for the more capable hands of humans. Some matters are also too sensitive to outsource to AI.

Around the same time Pokorny started using Watson, Children’s Healthcare of Atlanta in Georgia began working with IBM’s system to develop a chatbot for helping parents assess their children’s symptoms. The health center turned to automation after people began inundating their phone lines, cranking up wait times to 30 minutes or even an hour. Like Pokorny, CHOA was impressed by how fast it the bot was to deploy and how easy it was to customize so that it replicated the assessment flow that would’ve been performed by a pediatrician. In just its first week, the application logged 1,000 unique conversations a day.

But as much as CHOA sees the technology’s role expanding beyond covid symptom assessment, the center doesn’t see it fully replacing call lines or becoming the first point of contact. “It’s worked really nicely during the pandemic time when people really just want some information,” says Daniel Hirsh, a physician there. “But I work in pediatric emergency medicine in the ER. When parents are nervous about their kids, people want to talk and have a conversation with a human.”

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Providing the right products at the right time with machine learning

Amid shifting customer needs, CPG enterprises look to machine learning to bolster their data strategy, says global head of MLOps and platforms at Kraft Heinz Company, Jorge Balestra.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.