The Mood on AI: Let me speak to a human

One Bluesky user had been stuck for a long time with an insurance billing problem. By the end, she had one request: less AI in customer service, and an experienced person who could actually fix it.

The Mood on AI is a weekly read of how people are talking about artificial intelligence, put together by Fíor from a review of public posts on X, LinkedIn and Bluesky. It is a snapshot of the conversation, not a poll.

This week's mood was conditional. Very few people rejected AI outright, and very few welcomed it without terms. What divided the posts was not the technology but where it had been put. People were warm towards AI that helped them do something. They were sharp about AI that stood between them and a person.

Where a person used to be

Customer service drew the most heat. Several people described the same experience: a delayed delivery, a billing error, an AI assistant that could not solve it, and no easy way to reach someone who could. One argued that AI may well improve service one day, but that companies are replacing staff now, even where it makes things worse. Another asked, with some sarcasm, whether anyone had checked that the AI agents would actually be better.

There was a counterpoint. One post described an AI assistant handling customer requests well across several systems. The difference was the kind of problem. Simple and well defined went fine. Unusual and tangled did not.

Assist or replace

Jobs came up again and again, and the line people drew was the same one. AI that takes the repetitive parts of a job so people can focus on the rest was welcomed. AI that takes the job was not.

Microsoft's offer of voluntary buyouts while it spends heavily on AI drew sarcastic replies about promises that AI would not replace anyone. Barclays' job cuts prompted a worry about young people: if AI takes the entry-level roles, where do they get their first experience? A widely followed AI educator argued that companies are leaning on AI in hiring too heavily, and replacing people before the tools are ready.

Who answers for it?

Accountability ran through the more measured posts. People working in banking discussed AI that speeds up lending decisions, and welcomed the speed while asking who is responsible when an automated decision is wrong. Others in security described AI agents that can act on their own, and said the more they can do, the clearer their limits need to be. One person raised AI being used in asylum casework, where an error could change someone's life, and asked why it was not getting more attention.

Even the optimists set conditions: oversight, clear permissions and a named person answerable for mistakes.

Checking the work

A familiar complaint got sharper. One writer described AI making confident wrong assumptions and inventing details, so that guiding and correcting it became extra work. Another argued that professionals should not bill clients for hours AI saved them. Both point to the same thing: the time AI saves is only real if checking its work does not eat it back.

The money question

The week's market falls showed up in the mood too. People asked when AI spending will earn back the cost of the chips and data centres, worried about the debt behind it and the strain on power grids, and questioned chipmakers buying shares in their own customers. Others doubted whether the leading AI companies, including the newly listing Anthropic, can turn fast growth into lasting profit. Most of this was not opposition to AI itself. It was doubt about the money around it.

Harder to see

A few posts worried about AI that people cannot easily spot. One shared figures on AI-generated political adverts in the US midterms, many without any label. Another reacted to chatbots designed to build emotional closeness, especially with children. An artist said people now sometimes treat artists like image generators, expecting instant output on request.

These worries are the closest to what Fíor is for. When people cannot tell what was made by a person and what was made by a machine, trust suffers on both sides.

Where the optimism was

The hope was specific again. People were excited about AI helping correct errors in quantum computers, estimating building repair costs, and spotting security threats faster. The more precisely someone could name the problem AI solved, the more positive they were.

The reading

Last week the question was who is watching the agents. This week it was more personal: who do I talk to when it goes wrong? People are not asking AI to disappear. They are asking for a person to still be there behind it.

Try the browser extension at Fíor.