The New Anatomy of Customer Experience - Part 2

A Quick Recap:

In the first part of this series, I talked about how building a trusted relationship is the core tenet of customer success and how over the years, technology has played a central part in execution of that core principle.

We looked at some of the most pertinent challenges that modern software, primarily automation and chatbots, face in terms of widespread adoption:

1. Humans still have a better understanding of needs and emotions
2. Humans can solve complex issues better
3. Humans can provide more options that might work
4. Customers still, simply don't have enough trust in the outputs of a chatbot or AI

Given that GenAI is uniquely poised to address the above, I started outlining the first part of my playbook, focusing on the infrastructure required to set up a GenAI powered customer experience ecosystem.

In this article, I am going to expand on the playbook and look at the operational and governance aspects which would be essential in scaling and maintaining any GenAI driven customer success workflow.

The Playbook Part 2: Strengthening the foundation

The following sections are what I would classify as non-negotiables when it comes to establishing GenAI powered workflows. The technical foundations described in the previous post can be achieved through significantly different combinations of different AI models and data and compute infrastructure, depending on the use case. However, without proper governance and guardrails, any technology of this capability used to serve the workflows, will eventually be risk-prone at best and at worst, irreparably accelerate the erosion of customer trust.

Non negotiable #1: Establish clear AI governance from the start

Trust and Safety are paramount when using powerful probabilistic models like LLMs, especially in real customer interactions. Rogue AI behavior can, and in fact has, resulted in significant financial harm for the provider, and in my opinion more importantly, caused emotional distress for the customers. So how would an organization go about effectively controlling the output of its AI chatbots? The below might help:

• Creating AI councils or task forces to oversee deployments, define usage policies, and monitor outcomes. Adhering to frameworks like the EU AI act and NIST are of course the bare minimum, but organizations that are actually serious about scaling up their AI workflows also create contextual principles and guidelines for consistent usage of AI.
• Guardrails like configuring (primarily through robust system prompts) AI to flag high-risk or sensitive topics and involve humans in the loop are known to yield an overall better experience for the customer
• Confidence Thresholds and Regular Audits:  72% of consumers think AI can spread misinformation. In today’s geopolitical climate, that is simply not a risk that anyone can afford to take. Regular audits of AI transcripts help establish appropriate confidence thresholds, beyond which AI can engage human assistance. This also ensures transparency and maintenance of customer trust in the long run, by controlling and curtailing rogue AI behavior at its inception

Non Negotiable #2: Don’t stop investing in your most valuable assets: People

Its no secret that for most, if not all businesses, the ability to do much more with less is one of the most alluring aspects of Generative AI. Unfortunately, what even the most seasoned organizations fail to see often is that this tool, however powerful it may be, is only as good as the people that wield it. A modern support organization looking to scale sustainably while maintaining high levels of customer trust will empower its staff to scale along with AI:

• The AI Conductor: Reskill and upskill your agents to be AI conductors. These are the humans who have had the right training to orchestrate the AI driven workflows and step in when the AI is at risk of providing erroneous responses or taking incorrect decisions
• Up-level the agents role: With the right training and by hiring from the right backgrounds, the customer support agent can double up as the data analyst or a seasoned Sales rep. This is the real acceleration that AI will eventually provide. In fact, 75% of CX leaders already frame AI as augmenting their workforce, with AI tools supercharging their human agent’s workflows

The ‘So What’ of it All

Over the course of these 2 posts, I have tried to establish a robust implementation framework for the modern Customer Experience organization, supercharged with AI. The 4 key pillars were:

1. Hyper-personalize journeys at scale
2. Turn unstructured feedback into Predictive Intelligence
3. Robust Governance and Controls
4. Continuous Investment in your People

Needless to say, all of the above go several layers deep as you start putting the playbook to practice. However, that becomes somewhat simpler (not easier), when the end goal is clear - better customer experiences.

While metrics like cost reduction is important (and indispensable), the true effectiveness of this playbook will be reflected in the measuring customer centric metrics like CSAT (Customer Satisfaction), NPS (Net Promoter Score) and traditional support KPIs (response time/resolution rate/cost per contact). Cases like the success of United Airline’s ‘Every flight has a Story’ initiative - that resulted in a 6% increase in customer satisfaction - underscores the effectiveness of a well executed AI driven customer experience program.

All things considered, customer support and experience, like almost every other facet of modern life, is set to see some major disruptions due to AI. Following well intentioned and diligently thought out playbooks like the one I outlined above, would be key for serious players in this space to uplevel their game and set the foundation for a new era of Customer success in the age of AI