AI tools make individual tasks faster but often make workflows more absurd. Shopify CEO Tobi Lütke warned about this trend, saying, "The failure case of lazy work is not lack of output. It's actually over-output." Many workplaces now see AI writing content that only other AI tools read and summarize.
Summary
AI can make individual tasks faster but the whole workflow more absurd, and that calls for a rethink.
You have a nice report to submit to a client. I would probably just say, "Hello... here it is!" But that's me. It's clearly better to write a professional covering note with it, perhaps even highlighting the top findings. That's best handled by your AI, so you hand it over with a short prompt.
The AI gets to work and, in seconds, comes back with three neat paragraphs of a cleanly written summary, everything neatly tied up with a bow. The client receives it. And promptly hands it to their AI to summarize the summary in two sentences. A response is soon generated with an elaborate thank-you note. You receive the thank-you note. Is there anything you should be acting on here? You ask your AI to look and let you know. There isn't. You can send the bill. The AI will make that. But you'll have to check it out.
And so continue the hours at thousands of workplaces, where AI writes, only for AI to read.
The CEO of Shopify, Tobi Lütke, was asked on a podcast whether AI had made anything worse at his company. "The failure case of lazy work is not lack of output. It's actually over-output," he said. He gave two examples. In the first, an employee asks River, Shopify's AI, to make a change to some code. River generates a pull request, the employee barely glances at it and passes it on to colleagues to review and figure out what's wrong. The employee has saved time by passing the work of judgement to someone else.
In another example, Lütke receives a long email and, while reading it, realises there's a simple point buried under all the verbiage. So he puts it into an LLM and asks it to compress it again. Why, he wonders, have we invented all this decompression and recompression? If you're already using an LLM, use it to synthesize the point simply rather than blowing it up into a large missive that wastes someone else's time.
Slop grenades
At Shopify, they have a term for AI output lobbed over to someone else to deal with: a "slop grenade". But the grenade needn't even be slop. It might be accurate, thoughtful and beautifully written. It just didn't need to exist.
In this common scenario, every decision actually makes sense. The sender saves time by having AI write the covering note. The recipient saves time by having AI summarize it. The original sender saves time by interpreting the reply. It's just that the whole chain makes no sense at all.
The problem isn't that AI is a useless time-waster. We can blame the technology for many things, especially for escaping its confines and wreaking havoc wherever it likes. But in the nonsensical chain, the issue is that we haven't really redesigned work around AI; we've inserted AI into existing habits and conventions, including those that may no longer serve any purpose once AI is involved.
Lütke's "decompression and recompression" complaint isn't really an objection to AI. He's famously pushing Shopify to use AI everywhere. It's an objection to using it in the wrong place in the chain.
Rethinking work
People are beginning to figure out that what's really needed is to redesign the workflow as a whole rather than insert AI into each separate step where it makes tasks faster. Of course, companies aren't sitting around waiting to change the way they work just to accommodate new technology. But when the technology is as disruptive and significant as AI, it makes sense to deeply review the workflow and the parts that will leverage AI effectively.
Maybe an email wasn't needed. Maybe a covering note wasn't needed if the report began with highlights. Maybe so many hand-offs are not necessary for a task. One won't know unless one looks. Force-fitting AI can, in fact, lead to fatigue, as decisions have to be made at additional points and work already done has to be reviewed carefully at every change.
Some new research from MIT shows why looking at the whole chain matters. AI is more useful when the things it does well happen together, so that it can get on with several steps and a human checks the result at the end. Scatter those same steps through a workflow and suddenly humans have to keep stepping in, checking and handing things back.
The tasks haven't changed and neither has the AI. Just changing where they occur in the chain can change how useful the AI turns out to be for a company.
At the very least, it can break the wasteful cycle of having AI output stuff only for AI to process, confusing humans in the process.
The New Normal: The world is at an inflexion point. Artificial intelligence (AI) is set to be as massive a revolution as the Internet has been. The option to just stay away from AI will not be available to most people, as all the tech we use takes the AI route. This column series introduces AI to the non-techie in an easy and relatable way, aiming to demystify and help a user to actually put the technology to good use in everyday life.
Mala Bhargava is most often described as a 'veteran' writer who has contributed to several publications in India since 1995. Her domain is personal tech, and she writes to simplify and demystify technology for a non-techie audience.