Proven Strategies for Successful ConversationalAI Initiatives
Why Did Early Chatbots Fail?
Chatbots hit the scene around 2015 and early adopters experienced moderate success. Back then the technology used very little artificial intelligence and relied heavily on scripted answers. If the bot developers didn’t manually keep updating the answers, the bot would perform badly relative to the user expectations. These poor experiences proliferated with many early projects and created a waning reputation for the nascent technology.
Enter AI and NLU. Since 2015, many advanced technologies have greatly improved the performance of the current generation of chatbots/intelligent virtual assistants(IVAs). Given the improvements, some companies are still struggling to successfully roll out their initiatives. There are many reasons for this, however the major factors are typical of any innovation or business transformation project.
Top 3 Pitfalls For AI/Chatbot Projects
- Lack of defined use case
There are thousands of potential use cases for chatbots/IVAs in just about every business. Knowing how to identify and define the most pressing use cases is key. Typically, by understanding where customers engage and how much it costs for resolution is the place to start. - Lack of resources
If a company or department has identified a compelling use case, stakeholders frequently have issues allocating enough resources to support the project. By identifying a champion who can understand the project goals and drive the project forward is very important. - The “big bang” approach
Once the key stakeholders get their heads around the power of AI and chatbots/IVAs they tend to start thinking big. Of course there is nothing wrong with this, however in order to properly integrate any new technology, there needs to be a pilot program with smaller initial goals to more easily accomplish.
4 Steps To Chatbot/IVA Success
- Build a Roadmap
Once the benefits of AI/Chatbots are understood, many stakeholders will create a large wishlist of use cases. While this is great and all potential use cases should be captured, a product lead should organize those use cases and build a roadmap. - Define a Pilot or “POC”
A POC or pilot project is the perfect place to start with ConversationalAI. As a newer technology, there needs to be a low-overhead project to prove out your use cases to the organization. Typically, a pilot project consists of one or two use cases with 4–5 intents built into a chatbot/IVA. Also, it’s extremely important to measure your results of the pilot project and show how each “turn” of the conversation is improving customer experience and lowering costs. - Build Consensus on KPIs
Key Performance Indicators are paramount with ConversationalAI. As with any innovation project, the ability to measure performance and outcomes is critical to the success of a project and longevity of the solution. An Enterprise-grade platform for building chatbots/IVAs should also have a customizable dashboard for business users to monitor performance. - Monitor and Train the AI
Launching a chatbot/IVA is really just the beginning. Since the core component of any chatbot/IVA is rooted in Artificial Intelligence, there needs to be a mechanism to continuously train the IVA. As data is collected from the ongoing dialog, additional training should be simple to promote.
The End is Just the Beginning
By thinking through these four strategies, you will begin to formulate a solid plan for success. Our final recommendation is to assign a Product Owner for the chatbot/IVA and charge them with publishing a roadmap of use cases and intents. The solution should become exponentially more valuable over time with proper care.
