In conversation with Giuliana Lucchesi, Inbound Sales Development Representative at Wizeline
When you think of professions requiring employees to be “on call,” your mind probably wanders to doctors or lawyers who often must be available at all hours to help with urgent patients or client situations.
As a Sales Development Representative (SDR), I am also in the position of being “on call” for Wizeline customers. While I may not be saving lives, maintaining after hours and on-call duties helps save the company money by retaining and engaging customers.
On-call doctors might agree that there are some nights where being on-call is pretty relaxed, and others where it’s a never-ending stream of urgent requests. My team decided to implement an AI-powered chatbot to alleviate inbound requests to our website and make my on-call life a bit easier.
24/7 bot support to triage customer inquiries
My fellow on-callers can imagine my initial joy when I received the news that Wizeline would be implementing a chatbot to help manage inbound website traffic. It had also always been a bit awkward to explain to customers why a company that builds chatbots did not have a bot on its own site.
We named our bot “Wizebot” (Wizeline’s corporate chatbot) and determined it would have two goals: 1) improve inbound customer support efficiency and 2) serve as an example chatbot for potential clients scoping our site.
I was optimistic that implementing Wizebot would mean that I would have a trusted 24/7 virtual assistant supporting me on off hours, during meals, in the middle of meetings, and on weekends.
Friend or foe?
Unfortunately, Wizebot became more of a frenemy than trusted ally. As I watched the conversations come in through Intercom, I was unpleasantly surprised by the relatively basic capability of the bot. Small spelling errors or awkward sentence structure would trip it up. The chatbot followed programmed conversation flow fairly well, but the moment a user took it out of its traditional flow, it went mute and could not respond.
Wizebot was creating more work than assistance. Instead of simply engaging with users manually, I was now monitoring and documenting the bots’ behavior in addition to jumping into conversations.
Ownership of this bot was also a problem; the intern that had initiated the project returned to university. A ship without a captain left us without concrete ways to improve Wizebot and make it more useful.
Our patients (customers) felt the pain
In the months following the birth of Wizebot, I noted a steady decline in the number meetings I was booking because Wizebot struggled to hold basic conversations. In early October we hit an all time low when a lead called to cancel a meeting because he had a poor interaction with Wizebot. “Is this what they call AI?” he asked, “I don’t understand how this quality of interaction could help my business.”
Flooded with embarrassment, I realized it was better to explain why Wizeline did not have its own chatbot versus trying to defend Wizebot’s poor performance. Despite always adhering to a formal processes for our clients, our homegrown chatbot was built without much structure. We had a tight deadline and needed to get the bot launched before our product manager intern returned to school.
I experienced increased stress trying to mitigate the high quantity of poor bot conversations at the expense of quality interactions with customers. Like a mediocre health care provider, I addressed patient symptoms but not underlying illness.
Investing in a “good” bot drove results
Our Wizebot was sick, but we knew what medicine to prescribe to get it healthy: a commitment of time and talent. We locked in ownership from UX, product, sales, and engineering departments to tackle all areas of bot performance and programming.
Wizebot went through a series of improvements the third week of October which has ramped up our SQOs (Sales Qualified Leads). In just one week, we tripled our conversions.
We learned from our mistakes
Confession from a company that builds chatbots: If you cannot dedicate the time and team to thoroughly design a chatbot, it is better to not do it at all. When done correctly, chatbots can have a transformative impact on your company’s customer support efforts.
Our experience has underscored how much dedication and work is required from creation and programming, all the way to post-launch optimization. Chatbots, like any member of a customer support team, need to be trained.
My father is a physician and takes great care of his patients, but sometimes neglects or ignores his own symptoms. Like an overworked doctor, Wizeline builds incredible conversational UIs for customers, but had ignored the symptoms of a badly-made internal chatbot.
I’ve had a front row seat to working with a hastily-built chatbot as well as a much more complete and maintained solution. With the re-designed Wizebot, my role as an SDR has been meaningfully supported and improved. I am free to focus and connect with customers, and work on projects throughout the day knowing that I have a reliable first line of defense.