Yet, research now emerging on the value and impact of chatbots suggests it’s not this simple. While chatbots can have benefits, there are also potentially hidden downsides, including their limitations and their impact on human staff. So it’s time to have a serious chat about the value and role of chatbots. Below, I’ve explored three recent research studies on chatbots to get the conversation started.
Chatbots require a multi-faceted approach
As research on chatbots emerges and evolves, we are learning that how a chatbot works is only one factor (albeit a crucial one) in determining the impact it will have for consumers and, therefore, retailers. For instance, a stream of research has examined how making a chatbot appear more human – such as through the way it responds or giving it a name – can make consumers more comfortable using it. Others have examined how chatbots may work better in some situations than others (more on this below). As a result, there is a lot of research emerging about specific factors that influence when and how chatbots do or don’t lead to positive outcomes for consumers and brands.
A recent paper1 summarised this research and created an overarching framework that explains how chatbots can lead to positive consumer outcomes like intention to use the chatbot, satisfaction, and even willingness to pay. This framework shows getting value out of a chatbot comes down to a combination of four factors:
- Chatbot design, including things like its appearance and interactivity
- Customer characteristics, such as desire for control
- The nature of the service, including whether it is transactional or more relational
- The level to which the chatbot facilitates agency (control over outcomes) for customers
The paper’s authors found that these factors work together in various ways. For instance, customer characteristics influence the degree to which different design factors lead to agency, and therefore positive outcomes. In other words, some customers will have a better experience with chatbots that are designed in certain ways (such as being very human-like and interactive), while others will prefer different types of chatbots (for example, not human-like).
From this you can see that getting the benefits of chatbots is not as simple as turning one on or even designing it well. Careful thought is needed in terms of when, why, and how chatbots are implemented to get all these factors right. The complexity doesn’t end there either. In fact, complexity itself is highly relevant to the value of chatbots, as you’ll see next.
Chatbot effectiveness depends on the complexity of the task
Artificial intelligence, and by extension chatbots, are becoming increasingly sophisticated. In fact, AI is exceptional at parsing large amounts of data quickly and returning logical or pre-planned responses. For example, AI can easily show which products are commonly bought together and make recommendations based on that data, and even perhaps a consumer’s purchasing history. A good chatbot could turn that data into a response if a consumer asked what shoes they should buy to match an existing outfit.
AI and chatbots aren’t well equipped, however, for more complex tasks, particularly those involving human feelings and emotion. For example, a chatbot might struggle to respond if a consumer asked for shoes that are nothing like ones they’ve purchased before because they want to switch up their style and turn some heads when their office opens again. These types of questions require more nuance and human-to-human interaction than chatbots currently being used are well suited to handle.
The result is that while chatbots are great for some service interactions, humans are better at others. This is supported by academic research into chatbots and AI more broadly. For instance, one study2 found that consumers preferred chatbots for simple problems, but preferred humans when they thought their problem was more complex. The authors argue that chatbots don’t reduce the need for human service agents. Instead, they should merely change the way human service agents work. In essence, chatbots should be used to handle simple queries, while humans should be focused on the complex, emotional, questions that AI can’t yet comprehend.
Chatbots can change expectations of human service
The link between chatbots and human staff goes even deeper. I’m supervising PhD candidate Anh Tran at Swinburne University of Technology, along with Professor Lester Johnson. For her thesis, Anh uses Twitter data to explore consumer sentiment towards chatbots and the impact this has for retailers. It’s a fascinating research project, and we recently co-authored a paper3 together exploring some of the findings.
In that paper, we tracked how consumer sentiment towards chatbots, and towards service from humans, changed over time, across multiple retailers, before and after they implemented their chatbots. One of the intriguing findings is that after a retailer implemented their chatbot, consumer sentiment towards human service decreased. In other words, consumers got more negative about the service they were receiving from real people after being exposed to the chatbot. This trend generally increased the longer the chatbot remained active.
When we investigated this further, we discovered that the things consumers were saying about human service agents also changed. For instance, we started seeing a lot more references to the speed of service than before the chatbot was first launched. This tells us that as consumers are exposed to some of the strengths of chatbots – such as quick responses and being on all the time – they start to expect this everywhere, even from other humans. When humans can’t live up to these new expectations, consumers respond poorly and voice those frustrations publicly.
For retailers, this means the trade-offs that need to be considered when implementing a chatbot involve more than the technology itself. It’s important to consider how a chatbot might affect customer expectations for service overall, and the impact this could have on human staff. Retailers may need to think about how to mitigate these consumer expectations or potentially make it clearer when service is automated or being provided by a real person. Additionally, staff may need training to mitigate these effects themselves, and prepare for the new expectations consumers will have for them.
What this means for retailers
Like all retail technologies, chatbots have a lot of potential but also risks if not used effectively and thoughtfully. To summarise, I’ll refer to the SAMR (Substitution, Augmentation, Modification, Redefinition) model I discussed in a previous article. The basic idea is that technology has little value if it’s simply a substitute for humans, with no functional benefit or change. It has slightly higher value when it augments or improves existing people or processes. However, the real value of technology is unlocked when it is used either to modify (redesigning existing tasks or processes) or redefine (creating totally new tasks or processes).
Most chatbots are currently substitutes for existing customer service agents, or at best provide slight augmentation. However, the emerging research suggests they have the potential to truly modify and redefine customer service and the role humans play. Consider how the role of service employees could be totally redefined because of chatbots. Could their time be spent becoming true concierges, stylists, or product advocates? Freed of having to answer the basic questions – which chatbots can answer quicker and more accurately – how could their time be spent further enhancing the customer experience? That is where the true value of chatbots will be realised. Not in replacing human staff, but in unlocking human potential to redefine what a great customer experience is like.
1 “AI-chatbots on the services frontline addressing the challenges and opportunities of agency” in the Journal of Retailing and Consumer Services, led by Terrence Chong
2 ‘AI Customer Service: Task Complexity, Problem-Solving Ability, and Usage Intention’ in the Australasian Marketing Journal, led by Yingzi Xu
3 ‘Exploring the impact of chatbots on consumer sentiment and expectations in retail’ in the Journal of Retailing and Consumer Services, led by Anh Tran. [BIO]Dr Jason Pallant is a senior lecturer in the department of managing and marketing at Swinburne University of Technology.