Industry Leaders Explore the AI Data Challenge

Posted by Joey Greenwald


Execs in the Know gathered 75 customer service leaders and software providers in San Francisco to discuss how Artificial Intelligence (AI) and bots can improve customer experiences (CX) and share the latest advancements in the space.

The data gap 

The audience took part asking questions to the panel. A recurring theme emerged: despite advances in AI and machine learning, data remains a fundamental challenge. Organizations do not have the data needed for successful bot deployments. Bots need to be fed with questions and accurate answers they can learn from. The consensus was that too much support data is dirty or outdated. Practitioners also complained they were missing the critical mass of knowledge to enable bots to address a large enough set of use cases.

Bending the curve?

Projects end up requiring much more effort than expected. Use cases resolved by bots remain limited and, in a world where CX reigns, the impact of failure is scary. Most participants shared the conviction that AI will play a critical role for better self-service. Companies have transitioned from "if" to "how" to make their bot project successful. The burning issue has become to "bend the curve":

  • expand the scope of customer inquiries handled by bots and
  • reduce the amount of effort required to make projects successful.

AI Panel

Taming the knowledge gap

While the solution—combining AI with knowledge—is known, it is very difficult to put in place. Companies need to assemble experts having not just knowledge of all products and services but also an understanding and appreciation of customer issues. This group needs to create and curate relevant content. It must eventually ensure support information is accurate and kept current by gathering feedback at scale.

Harnessing experts

The two main talent pools available are customer service agents and external “power” users or product experts. Leveraging internal resources is limited:

  • Agents are scarce resources. Companies want them focused on complex issues resolution and handling critical moments with customers,
  • Agents are already asked to do many things and are under a tremendous time pressure. It is hard to have them to curate content or provide detailed feedback.

Several companies in the room shared how they built an expert network and incite their power users to help with customer service. Experts can be incentivized at the fraction of the cost of problem resolution to provide broad feedback on responses to support questions, curate content, and assist escalations to human agents.

The combination has shown to deliver far more accurate bot responses, higher deflection, and better Customer Satisfaction (CSAT) and Net Promoter Score (NPS). As we move into 2018, it emerged as an approach that must be explored!