Camille E. Acey

Someone on Mastodon posted that Shopify was downsizing their customer support team in favor of leveraging an AI tool instead and I spent several posts sharing my Expert Opinion on the matter and figured I ought to drop it here for posterity.

I have been a customer support leader in the tech industry for almost a decade, and I can confidently say that the challenge is for companies to build products that mostly have routine tasks as solutions to problems. I have worked at tiny startups AND large companies, and the vast majority of the issues (outside of billing questions) required research, testing, and then a nuanced (non-template) response.

I am not saying there is no place for AI. Companies like Zendesk, Intercom, and Solvvy have been building tools to deflect/augment support tickets and answer customer questions via chat for years. Leveraging LLMs could help improve those exchanges. One application that comes to mind is scraping the agent responses in the ticket queue rather than the agent having to periodically document it or ask the documentation team to.

The challenge, however, is separating the wheat from the chaff in terms of agent responses. That is likely still human work to keep telling the machines:

• “Yes, this is still what we consider the right/best answer to this problem at this current moment in time.” OR
• “Wait, Computer, a new *better* answer has emerged” OR
• “From our perspective, problem X is solved. If someone reports it, please flag to a human.”

Which, of course, jibes with your point that a lot of the job may boil down to training AI. I just don’t think the AI training is liable to be the lion’s share of the work. What *will*, however, probably happen is the end of customer support “tiering”. Routine task customer support with tier 1 “grunts” is a wrap. I was never a fan of tiering. It creates the bad “clueless support person” situation everyone always complains about.

If people need to be experts across the “stack” in order to get/keep their job, it will (hopefully!) result not only in better/less-frustating employee opportunities and experiences but also better/less-frustrating customer experiences.

The question of course, then, maybe becomes about barriers to entry in a company. People often think of CS as a “stepping stone”. It makes me crazy. I never hire support people who think that way.

All this to say, that my fingers are crossed that I hope the further adoption of LLM tools *might* support an elevation of the field of customer care such that people know that when they are dealing with a human, it is because their problem warrants that level of skill and expertise.

(This post was written by a boring old person)