RPA and conversational AI (chatbots, voicebots etc.) are two technologies often associated with one another, both in conversation and in mindset. This is understandable given both are pursuits of automation, they both entered the mainstream at similar times, and can provide significant value when combined together. Many organizations on their automation journey begin with RPA before later expanding into conversational AI. But while this makes sense on some levels, it can also be detrimental to the overall impact and value of the chatbot project.
The anchoring effect of RPA
I’ve seen firsthand the difficulty of shifting from RPA to conversational AI initiatives.
One client I worked with, a multinational retailer, successfully established an in-house RPA factory. With a centralized army of trained and skilled developers, they were successfully automating standardized, high volume processes across the organization. The progression to conversational AI was challenging, however, as the glow of success and embedded approach to RPA was difficult to shake. “Conversational AI is just another form of automation”, they said, “our approach was successful with RPA, why don’t we just replicate that?”
This sentiment is not unusual. To understand why RPA and conversational AI should in fact be approached differently, it can be useful to understand the Three Lenses.
Three Lenses of Innovation
The Three Lenses of Innovation is a framework to ensure innovation teams are heading in the most promising direction. At a high level, the framework provides three key areas that must be assessed to reduce the overall risk of failure, and associated rework cost of the project.
To be successful, an innovative idea must be:
- Desirable: customers actually want it;
- Feasible: building it is technically and organizationally possible, and;
- Viable: it has a robust business case with a quantifiable ROI.
Structured activities to test each lens should be incorporated into every conversational AI project. Unfortunately, many organizations pursue development with such enthusiasm that one (or all) lenses are not assessed, causing the project to fly blind, and pinning success on a metaphorical coin toss. Without assessing and understanding each lens in detail, conversational AI solutions will most likely fall short.
RPA vs Conversational AI
RPA focuses on back-office automation of highly manual business processes. For RPA initiatives to be successful, it is important to have the technical and organizational capabilities to execute at scale (Feasibility), and a strong business case to justify the investment (Viability).
Compared to conversational AI, there is one notable difference: RPA bots automate manual tasks while chat solutions automate human interactions. This means a significant shift in focus is required towards the end user. As I’ve written previously, end-user adoption of the solution is critical for success. Without users interacting with the solution, a chatbot will not deliver on the value promised. As such, I would argue that, unlike RPA, the Desirability lens is perhaps most important of the three for conversational AI initiatives.
We have established that the Desirability lens is arguably the most important to consider in any conversational AI project. So how can we test whether our solution is Desirable? Of course, specific tests used will depend on your organization, your customers, your culture, and your scope. But some useful activities to consider include:
Mapping the customer journey
This activity requires both a good understanding of your customer (through customer personas, for example), and the steps they take to complete a task (process from a customer perspective). This will include touchpoints with your organization, likely questions they will ask, and channels they will use. Journey mapping can be a useful way to shine a light on the customer experience, where the pain points lie, and specific areas where chatbots could enhance the customer experience.
Bring customers into the design process
What better way to ensure your solution is Desirable than getting customers to help design it. Inviting customers to join an interview or focus group can provide deep insight into customer preferences and behaviors. Very quickly, design considerations will emerge that will help drive adoption and customer value. Better yet, hosting workshops with customers to specifically discuss the design considerations of your solution will make two-way ideation with your project team possible.
Iteratively test your solution with customers
Building user tests into your development plan can extend timelines, however, this will allow course correction early, when the cost of doing so is still relatively low. Some useful tests include:
- Wizard of Oz testing drives insight into copy and conversation flow design. How do customers engage with the tone of voice? Are the bot’s questions clear enough? Is it intuitively driving towards the outcome the customer needs?
- Design testing focuses on the visual elements of your solution. Customer feedback on the visual design will provide insight to refine and improve engagement and usability through visual elements.
- Hallway tests can be useful for testing the usability of your solution with a random sample of customers. Any issues impacting the progression towards desired outcomes will soon emerge and should be incorporated into the backlog.
Once your chatbot is released into the wild, the consideration for customer Desirability continues. Many organizations have understood the need for analytics and performance metrics to continuously improve the solution. This data is a promising step in the right direction, however, only tells one side of the story. Qualitative data points in the form of customer feedback should be collected and analyzed on a periodic basis to validate Desirability test results thus far, and evolve the solution in step with changing customer expectations.
Critical to the success of chatbots is the adoption by end-users. Automation goals cannot be realized if no one uses the chatbot, thus the importance of ensuring it is Desired by your customers. This is the key difference with RPA and necessitates a distinct approach to development. It is not a case of ‘build it and they will come’; conversational AI teams must ensure they are building the right solution for the right customer problem. Desirability hypotheses regarding your chatbot should be tested before, during, and after development to ensure customers want it and use it. The success of your solution depends on it.