Javier Silverio has spent eight years in the Help Scout support queue, and now he's building internal AI tooling. He and Mat get into what it means to hand your data to a company when you can't check what happens to it, why people have been surprisingly quick to accept chatbots that aren't 100% infallible, and how trust is built when something breaks.

Subscribe to The Supportive in your podcast player (or via Spotify and Apple Podcasts), or watch the full episode below. Every episode this season will also be on video, so please subscribe to the Help Scout YouTube channel for new episode notifications.

Some key moments

  • (3:27) What DJ requests and feature requests have in common

  • (6:41) "Should I build it just because I can?"

  • (7:47) The Monzo co-founder who records every conversation

  • (12:16) Trust gets built when something goes wrong

  • (21:21) Some people will wrestle with a bot, some just want the answer

  • (23:34) A year spent teaching an AI tool to say "I don't know"

Too long, didn't watch? Here’s a concise version

  • Understand the job before you AI the job. Don't automate it away until you know the true extent of the work.

  • People are adaptable. Some customers will happily go back and forth with a bot until it gets there, and many are OK with trading off mostly right for very fast.

  • Be responsible. Not just in doing the right thing yourself, but in standing behind automated answers and decisions, and making mistakes right.

Links from this episode

Like what you see? Share with a friend.
Mathew Patterson
Mathew Patterson

A veteran support leader, Mat spent years teaching better online support at Help Scout before founding More Human Content. Subscribe to The Supportive or find him on LinkedIn.