Since ChatGPT launched in 2022, most of us have accidentally fallen into a trap where we ask AI to do the parts of our work that are arguably the most human. We ask artificial intelligence to write our emails for us, to brainstorm new ideas, and to turn scattered ideas into clear plans and strategies.
We humans have outsourced the fun, challenging stuff that requires we use our brains, but are still spending enormous amounts of time on the mundane. Slack found that the average employee spends about a third of the workweek on tasks they personally consider “low value.” For someone in finance, this might mean reconciling spreadsheets or hunting down missing information; for someone in sales, it could mean reviewing contracts or updating the CRM.
These are necessary tasks to keep a company running, but in a world with humanoid robots, self-driving cars, and a path to superintelligence, why are they still left to humans? A 2026 Boston Consulting Group survey found that 42% of regular frontline AI users save at least a full workday each week with AI, yet more than half say they aren’t redirecting that time toward more strategic work.
These “boring” tasks are precisely the tasks that are actually better suited to current AI progress, because they share something important: There’s usually a right answer. The best work to hand AI right now is work where you can clearly tell whether it got the answer right. This is something I’ve long believed, but had to experience firsthand. As the CEO of Tabs, even after we’d raised several rounds of funding, I still managed our own billing and collections for years, because I didn’t want to lose sight of what we were able to automate and where AI wasn’t ready to take the lead.
Most companies have this division of labor backwards, but the reason is unsurprising. The explosion of generative AI made creativity the most obvious entry point for everyday employees experimenting with AI. A blank ChatGPT box feels like an open invitation to write copy, brainstorm things, or generate ideas. It’s no surprise that workers turned to AI for these tasks first. OpenAI’s own workplace data shows writing remains one of the most common ways people use ChatGPT on the job.
Some of that writing is useful. I use AI to brainstorm all the time, including to help generate additional examples for this article. Some of that writing was okay, but a lot of it was quite bad—at least in my eyes. The problem is that creative quality is inherently subjective. There is no objectively correct marketing tagline or perfect strategy memo that won’t spark debate. Two smart people can read the same AI-generated paragraph and disagree about whether it’s insightful or “AI slop.” Creative work is a poor place to ask AI to be authoritative.
Companies evaluating AI use cases tend to start by asking, “Can AI do this?” but a better question is, “If AI does this, how will we know whether it did it correctly?” If there is no clear answer, you may still have a useful copilot, but probably not something you should build critical business processes around. If the output can be validated, you have a much stronger candidate for a part of the business that can truly be transformed by AI.
Executives should therefore be skeptical of the idea that AI’s primary purpose is to replicate the highest-level thinking their employees do. I don’t believe that currently exists, nor will it for some time. When a decision is strategic, contextual or dependent on taste, we humans have the upper hand. The good news is, these are also the parts of work many people actually want to spend more time doing.
The companies that get the most value from AI over the next few years will be the ones that redesign mundane workflows around verifiability: letting AI take on the work we can check, so people have more time for the work only they can do.
