Support lab home

Exploring the Roles of LLMs for Support

Matt Lewis works inside an LLM company and runs its support. He separates the AI hype from the tools that already work — and those about to change the industry.

A smiling man with a neatly trimmed beard and dark hair, wearing a light shirt, against a dark background representing a.

with Matt Lewis, Head of User Operations at AI 21 Labs

A smiling man with a beard, in black and white, against a vibrant pink and purple abstract background reflecting Birdie's.

play the video

36:13

Introduction

Few people sit at the intersection Matt Lewis occupies: he runs user operations at AI21 Labs, an LLM company — which means he supports customers with the very technology everyone is speculating about.

Recorded as the first wave of the LLM boom hit support, this conversation separates signal from noise: the mature tools worth adopting immediately, how AI21 automated over 90% of its B2C volume while keeping satisfaction high, and how to interrogate vendors so you buy real language-model capability instead of a marketing sticker.

What you will learn in this episode

  • The two LLM tool categories that were already mature: paraphrasing and summarizing

  • How AI21 Labs pushed B2C automation past 90% without tanking CSAT

  • Questions that expose whether a vendor really uses modern LLMs

  • Why context and knowledge bases determine what AI can actually resolve

  • What the AI wave means for support careers — and how to stay ahead of it

All episodes