AI automation: navigating from proof of concept to production
The field of customer experience automation is littered with dead proofs of concept that never could make it to the production floor.
Anxiety about using AI in customer experience environments is, to some extent, justified. Deep-learning algorithms are extremely complex and do not lend themselves to intuitive human understanding. This makes it difficult to predict how they’ll respond to exceptions outside the controlled laboratory environment.
Did you know?
AI driven operations
77. 4% of customers believe customer operations will be positively affected by AI and CX robotics.
Automate CX activity
84.6% of users say less than 25% of CX activity is being handled by AI and/or robotics.
Measure AI performance
Just 32.1% of users say AI/robotics are meeting or surpassing expectations.
Not always a case of plug-and-play
Be wary of claims that automation tools can be configured in production overnight, without the need for a technical team. Even at the simpler end of the scale, with robotic process automation (RPA), you may need to temper your expectations. Consider:
- how long it takes to grasp all the nuances of that tool
- whether the target is process-ready or needs to be reengineered
- how you will approach user buy-in and adoption
- how to secure systems and protect data that’s gathered and processed across your firewall
- if the present platform and process will be around long enough to get a positive return on investment from the automation
While not to be ignored, advanced digital technologies such as AI and machine learning can’t form the basis of a near-term plan. A gradual path that starts with robotic process automation (RPA), moves to rules-based voice agents and finally to AI-powered automation, should become the framework of a long-term plan. Use assisted-service, self-service and automation appropriately to drive productivity and reduce effort that enhances CX and does not put it at risk.
Senior Vice President, NTT DATA Americas