By David Baum
Last fall the fast-growing package delivery company Hermes launched a digital assistant named Holly to help customers get answers to their questions about their packages. Just seven months later, this chatbot is handling 38% of all chat volume and Hermes has reduced the support interactions that require a human by 30%. Holly helps customers track their shipments, change delivery orders, update account preferences, and handle many other tasks.
“A software bot can manage routine interactions quickly and effectively,” says Chris White, director of customer experience at Hermes. “Holly automates a huge volume of transactions with our customers and with our network of couriers.”
Founded in Germany in the 1970s, Hermes is one of Europe’s leading parcel delivery services. The company delivers approximately 340 million parcels over the course of a year, both for consumers and as a partner of 90% of the UK’s top retailers.
Hermes created Holly to help the company contend with a familiar problem: rapid growth. With year-over-year growth exceeding 20%, Hermes’ customer service staff was working overtime to field a mounting volume of inquiries via three main contact channels: telephony, email, and chat.
“On any given day, we have roughly a million parcels running around in our network—and about twice that many during the busy holiday period,” White says. “As our physical parcel volume continued to grow, we started to outgrow the elements of our services infrastructure.”
To create Holly, Hermes used Oracle Digital Assistant, a platform and set of tools for building AI-powered assistants. Developers use this cloud-based software environment to create natural conversational interfaces, through text or speech, to enterprise systems. It enables them to build and train digital assistants without the need for specialized AI skills. Developers don’t have to worry about creating algorithms to understand natural-language processing or classify inputs.
“We got Holly up and running fairly quickly,” White says. “Most of our time was spent establishing the flow or ‘journey’ that consumers follow during conversational interactions. We wanted to make sure that these exchanges were as straightforward as possible.”
White says the important thing with this type of project is to carefully model the workflows. “If you put nonsense in, then you’re going to get nonsense back from the bot,” he continues. “We put a lot of thought into creating a decision tree. The branches represent the ways a conversation can go, based on the supplied information and context. Within these conversations, there isn’t much that a live agent would do differently from the bot.”
Although artificial general intelligence remains a difficult nut to crack, chatbots programmed to respond to questions within a circumscribed domain have become extremely accurate and consistent.
We’ve taken away the mundane activity and freed up the human agents to handle the tricky cases—especially those instances that require experience and sensitivity.”—Chris White, Director of Customer Experience, Hermes
“The more you increase your mastery of an ever-smaller domain, the easier it is to instantiate that knowledge in a computer program,” explains Byron Reese, a futurist, entrepreneur, and author of The Fourth Age: Smart Robots, Conscious Computers, and the Future of Humanity (Atria Books, 2018). “If you want an AI to play chess or detect email spam, then you just teach it that simple thing. You don’t ask it to generalize, contextualize, be creative, or anything else. That is what we call narrow AI.”
Intelligent bots such as Holly use AI to improve the conversational experience—and boost efficiency—in the call center. “Using a bot that can operate 24 hours a day eliminates friction, by allowing us to accommodate a steady exchange of data,” says Chris Ashworth, CIO at Hermes. “We welcome Holly for its efficient, speedy, and consistent performance, but it also injects some self-deprecating humor into what could be cold transactions.”
But make no mistake: These transactions drive the business, and there is nothing funny about the information systems that give Hermes an edge. Just as Uber and Lyft transformed the taxi industry with an app that connects drivers with riders, Hermes created a logistics platform that connects parcel senders, a network of couriers, and parcel recipients, and its macro delivery infrastructure facilitates parcel delivery through its network of self-employed couriers.
“We are feeding 15,000 small business owners the information they need in order to efficiently complete our deliveries,” Ashworth says.
Hermes decided to use Oracle Digital Assistant partly because it is easy to integrate with other databases and cloud services. Finished bots can be exposed through standard chat and voice channels, a custom mobile app, or a link on a website.
Hermes integrated its Holly chatbot with Oracle Service Cloud and is establishing an interface to Facebook Messenger. All of Hermes’ data is stored in Oracle Database, which will soon be migrated to Oracle Cloud.
“We have certain automated transactions that write back to Oracle Service Cloud,” White says. “The bot handles them directly.”
There was an initial concern in Hermes’ customer service department that Holly would eliminate some jobs. In fact, the total number of jobs hasn’t changed, but customer support workers are engaged in tasks that are more interesting and challenging. “We’ve taken away the mundane activity and freed up the human agents to handle the tricky cases—especially those instances that require experience and sensitivity,” White notes. “The one thing we are not asking Holly to do right now is deal with the elements that require an emotional response.”
As Holly takes over more and more of the routine chat volume, Hermes’ customer support team can devote more time to the company’s email and voice channels, through which the more complex support cases tend to come. “The agents feel less rushed, because Holly is dealing with a good chunk of the caseload for them,” White says.
The effects of Holly on Hermes employees collectively are broader still. Thanks in part to Holly, Ashworth observes, Hermes has already seen improved customer satisfaction (CSAT) scores on overall outcomes, and it expects to see reduced employee churn.
Holly has become quite capable in handling the simple tasks that have been entrusted to it. According to Ashworth, Holly occasionally receives thank you notes from appreciative customers. “I look forward to getting to the point where Holly can inform consumers if something is amiss, to mitigate problems before they even contact us,” he says.
LEARN more about Oracle Digital Assistant.
Photography by Ian Gavan/Getty Images