Despite the theoretical power of artificial intelligence to transform the customer experience, many AI projects fail at the first hurdle. Henry Jinman from EBI.AI outlines the five most common mistakes and how to avoid them.
While AI promises a new dawn of efficiency, performing tasks better, faster, with fewer people, at lower cost and on a far larger scale, it also holds the key to transforming the customer experience (CX). Chatbots are already a common phenomenon in contact centres while millions of people interact daily with virtual assistants such as Google Home and Alexa.
For those organisations who haven’t yet invested in AI, many are experiencing a fear of missing out. As a result, plenty of businesses are rushing in and too many AI projects are failing – what’s going wrong?
Why do AI projects for CX fail?
AI technologies are transformational but they can be complex to scope out, build, deploy and operate. Here are the five most common mistakes organisations make:
- Unrealistic expectations – It’s common for users to have inflated expectations of new and emerging technologies. This could be because of marketing over-hype, lack of familiarity with the technology or the plain old hope that they have found a solution to some of their problems.
- Addressing the wrong challenges – ‘Trying to boil the ocean’ is a familiar term to describe companies who try to fix everything with a single project or, at the other end of the scale, spend 18 months writing an AI strategy paper that delivers nothing.
- Lack of training data – Many people say the more data you have the better. Yes, you do need data (ideally, lots of it) but it must be relevant.
- Lack of stakeholder engagement – The people who will make or break the project are those responsible for deploying the technology and the leaders of that department. Remember to involve the budget holders from the very beginning.
- Misunderstand the technology – Many AI projects fail for the simple reason that they are not really AI projects. AI technologies for customer contact need to be three things: digital, intelligent and automated.
What should I ask?
Don’t rush in – here are the top five questions to ask before you begin:
- Where do I start? Most organisations will want to achieve and demonstrate some quick wins but where do they begin? Should they be ambitious and try everything at once or run a mini-pilot to test the waters and find out what works and what doesn’t before going live?
- How do I measure success? Whether you are starting out big or small, the budget holders will want to monitor your progress and see that they are getting a return on their investment. Which goals should organisations focus on – should they look at their competitors, what customers want or what the business wants?
- How do we overcome our fear of the unknown? If you are normally conservative and play safe, could you change this approach to become bold and experiment even if you fail? It’s a great way to learn and there are even greater ways to share that learning before the all-important go-live.
- How do I test in a real-world environment and, crucially, while maintaining business as usual? How do organisations ensure the new solution can integrate with the production environment, provide the required functionality, and deliver a return on investment?
- How do I ensure a successful roll-out? There are various options to consider including the two most popular methods, known as ‘incremental improvement’ and ‘applying the learning’. What are the benefits of each and which one is best for my organisation?
Henry Jinman is the commercial director of EBI.AI.
