When It’s Time To Boot The Bots

Five years ago, Cornell University’s Creative Machine Labs almost broke YouTube with a video of two bots talking to each other. Whether you found it hilarious, mesmerizing, curious or even uncomfortable, many cite it as the first computer program to demonstrate artificial intelligence.

Indeed, we humans are full of error. Try as we might, we just aren’t perfect. That’s precisely why science fiction novels may have considered and introduced the idea of bots as another employee option in our world that might be systematically flawless. Some argue bots are quickly changing the face of marketing and all sorts of industries, simply because of enhanced communication, learning opportunities and targeting of consumers. Marketing leaders opine on if businesses are ready, while others question if we as a society are even ready for bots at all, despite their current existence. This notion of bots as efficient and perfect would be only limited by human error in programming.

As a result, some people and companies believe that perfection means efficiency. Thus, focused on the bottom line, companies with less human error would get more things done and be closer to perfect.

Chatbots are becoming lawyers that expunge parking tickets, personal stylists at H&M, makeup consultants at Sephora and recipe finders at Whole Foods (complete with emoji replies), among a slew of other professions. But how successful are they, and do they know their place? Or, have we not yet discovered the right place for them?

Max Robinson of the U.K.’s Aceworkgear believed in that notion as well and considered a chatbot for his website, which sells construction safety gear. That said, it was a short-lived endeavor that lasted only six months due to efficiency and not human or programming error. He’s since ripped the bots from the site.

Aceworkgear is not alone among companies that have had bumps with bots. Real estate site Trulia has had difficulties, and of course, there’s the infamous and potentially racist Tay representing Microsoft.

“We were experimenting with ways of helping our customers find what they were looking for on our website,” said Robinson. The business is a small one, with only two other staff members besides Robinson working in the office. “We were receiving countless phone calls every day from frustrated customers who hadn’t managed to find what they were looking for on our website.”

After learning about the growing popularity of chatbots, Robinson decided to push the button on adding a chatbot to the website. 

The chatbot joined the Aceworkgear website in Dec. 2014 with the help of a chatbot creation company, which Robinson still speaks very highly of.

“They were very responsive, offered us a free trial and constant support. They provided us with a custom bot, too, and worked with our site seamlessly,” said Robinson, who remembers being charged just $200 per month.

The goal was simple: efficiency and freeing up more time for human employees to be more efficient themselves. Was the goal met? Not really, Robinson said.

The chatbot helped the Aceworkgear team to spend more of their time dealing with larger customers, rather than smaller ones that were simply looking to purchase a single or specific product on the website.

As for transactions that happened between customer and chatbot, they were few and far between. Robinson said the chatbot was mainly helping customers who were first-time visitors to the site, rather than repeat consumers, and as a result, very few — the largest was a set of power tools costing around $300 — made purchases through the bots.

“We found that our customers were either becoming frustrated and bouncing off the site or were simply calling our office to speak to a member of our team. The chatbots weren’t answering customer queries sufficiently, as we were still receiving calls from customers,” said Robinson. “They were probably even more frustrated because they’d wasted time with the chatbot!”

Robinson has delved into the analytics of the bots to see why things didn’t go as planned. The average traffic for the Aceworkgear website was 15,364 unique visitors per month when the company had the chatbots in place. The highest number of interactions made with the chatbot was 97 in one day. However, that’s where the inefficiency appears. In the same day, Robinson says his team received over 80 calls to the office — all from customers looking for help using the website, most of them saying they had first tried with the chatbot.

After six months, Robinson booted the bots. And he’s back to human help.

“I’ve hired an extra member of staff for the office so we can deal with customer queries slightly easier,” said Robinson, who adds that he’s considerably changed the layout of the site to make it easier for customers to navigate more efficiently on their own.

Lessons from the experience? Robinson said the chatbots weren’t a complete failure, but the bots were simply prolonging the time that customers had to spend on the site without finding what they were looking for, which is obviously very frustrating.

Robinson has not sworn off chatbots — or the chatbot creation company for that matter. “Shop around before deciding on a company to work with, and be prepared to adapt it constantly until you find what works for you,” said Robinson. “We, unfortunately, didn’t have the time to do this properly, so we didn’t get the most out of our chatbot.”