I recently came across the story of British Museum object no. 1953,0411.71, also known more famously as the tablet of Ea-Nasir. This clay tablet, dating back to 1750 BC, is likely the first documented piece of a customer complaint. This discovery, besides being obviously hilarious, also made me think about the one constant in human trade that has not changed in the 4,000 years since: the importance of the customer experience. What has changed though, has been our ability to pre-empt, address, and improve that experience for a customer (with intentionality). We are at the precipice (or perhaps already in the deep end) of what may be the greatest technical revolution in modern history with Generative AI, and I wanted to outline my thoughts and propose a playbook in this two-part series on how this technology can transform the relationship between the modern commerce engine and the customer.
A bit of Historical Context
Before diving into a playbook of the future state, I want to set the stage by outlining why I think LLM-powered GenAI is especially poised to disrupt the customer support workflows. Customer support, at its core, is about solving an issue or providing meaningful guidance to the consumer of your product or service at any point in the customer’s lifecycle. It is essentially establishing a trusted human connection within a finite amount of time. And although modern technology (1-800 numbers, emails, CRM, chatbots, IVRs) has enabled enterprises to make it exponentially easier for customers to initiate contact and get a rapid response, the customer still overwhelmingly relies on a human to really address or solve the problem they are facing. In 2018, a PwC study found 82% of US consumers stated they wanted more human interaction. In 2025, this number still remains at a high 75%, especially for complex and emotionally sensitive issues.
Recent studies provide more insights into the above, with respondents stating that the key drivers for driving skepticism towards non-human interactions are:
- Humans still have a better understanding of needs and emotions
- Humans can solve complex issues better
- Humans can provide more options that might work
- Customers still, simply don't have enough trust in the outputs of a chatbot or AI
As it turns out, the above issues are exactly what Generative AI is uniquely poised to address. With newer models, specialized agents, and more data than ever in human history to power it, LLM-powered GenAI is surely the next frontier in elevating the trusted connection a customer craves to have when reaching out for support and aid.
The Playbook Part 1: Creating the baseline infrastructure
Augment (not replace), Human Agents with AI
- Use GenAI and Agentic AI to handle repetitive, but time-consuming analyses and tasks, freeing humans to focus on high-empathy interactions
- In practice, this means deploying AI for basic to intermediate inquiries, automated case summarization, and AI-assisted reply suggestions, while escalating nuanced issues to human support
- This is possible through personalization that is less about flashy algorithms and more about pervasive “next best action” and durable memory. Service flows should adapt to intent, value, and risk; This is possible with the latest AI (agentic) frameworks being able to persist conversational history and context across multiple channels and threads
Case(s) in Point
- This hybrid approach helped Verizon to not only derisk customer interactions, but also resulted in a net growth in their bottom line
- Ferrari’s Amazon Bedrock-powered configurator reduced configuration time by ~20% while improving lead quality - a good proxy for how AI can make high-choice journeys feel effortless
Addressing Complexity and Speed: Turn unstructured feedback into predictive intelligence
Build a Strong Data and Knowledge Foundation:
- Needless to say, but the ‘smartness’ of any GenAI tool model, is almost exclusively a function of the quality of data it’s powered by. To ensure your AI support tools give accurate, helpful answers, invest in curating a robust knowledge base and integrating your key systems
- This essentially means being able to effectively mine the unstructured data that makes up around 80-90% of an enterprise’s customer data
- This enables GenAI models and agents embedded in the Customer Support workflow to supercharge:
- Theme detection and root cause analyses
- Generate Risk and Propensity Signals
- Auto-knowledge creation
Case(s) in Point:
- United Airline’s successful creation of a data hub helped drive customer success at scale
- Overall, companies report up to 70% of customer requests can be handled by comprehensive AI knowledge systems when data is well organized
Continuing the Playbook in Part 2
Just setting up the data and technical infrastructure won't be enough to create a successful GenAI-powered Customer Support ecosystem. In part 2, I will continue expanding on the more operational aspects of implementing this playbook, covering topics such as AI governance, Effective training and investment in people, and focusing on CX Specific metrics for success.
As I stated at the start, the goal of technology in customer support and success is to continuously improve the trust and relationship that exists between the provider and the consumer. From a clay tablet in ancient Sumeria to an LLM-powered bot in modern Silicon Valley, humans will always find a pathway to express their intent. It is up to us to use the advancements for the overall good, helping elevate human interactions to the next level.