Is multitasking back?
Insight from Kevin DePopas — Demand Curve Chief Growth Officer
If you were job hunting between 1997 and 2001, you probably listed "multitasking" on your resume. It was the era of always-on work culture and the first BlackBerrys. Being "busy" was a status symbol.
Then research started coming out that killed the party. A widely-cited 2001 study from the University of Michigan found that task switching creates measurable "switch costs." Every time you shift between tasks, your brain stumbles and you lose time and mental energy getting back into what you were doing.
Another blow came from Stanford in 2009. A team of researchers studied heavy multitaskers and found they were actually worse at everything. Worse at filtering irrelevant information. Worse at task switching (despite all that practice). Their memorable conclusion was that heavy multitaskers are "suckers for irrelevancy."
By the early 2010s, the conventional wisdom had flipped. Multitasking became shorthand for distracted, ineffective work. And since then, most of us have adapted. Single-task. Deep work. Batch similar activities. Block out distractions.
For 15 years, that advice mostly held up.
When Multitasking Became Productive Again
Last year, OpenAI released ChatGPT o1 with extended "thinking" time (they've since rebranded it as ChatGPT 5 thinking, but same concept). Claude followed with their own extended thinking mode. These models often take 10 to 60 seconds (sometimes several minutes) to process complex requests.
Most people frame this as a limitation. "ChatGPT 5 thinking is slower." "Claude extended thinking takes forever."
In my actual workflow though, I've found a way to make use of the downtime. When I send a prompt to ChatGPT thinking or Claude's extended thinking mode, I'm not the one holding the cognitive load anymore. The AI is doing the heavy lifting on that task. My working memory is free.
So instead of staring at a loading bar, I'll switch to something else. Reply to an email. Edit a doc. Review a piece of copy. Then when the AI finishes, I come back and review the output.
Unlike traditional multitasking, I'm not trying to hold multiple complex problems in my head simultaneously. I'm acting as a router. I have two AI systems handling complex, independent tasks, and I'm structuring what I want them to accomplish. It turns anyone into a manager, orchestrating the work rather than doing it. When I return to review each AI's output, I'm evaluating and refining rather than creating from scratch.
Don't get me wrong, this workflow is susceptible to producing work and analysis for the sake of it. You still have to gut-check yourself, just as you would if you were managing an employee.
Is it worth having your AI work on this task, or are you just creating busy work?
The Parallel AI Workflow
Here's my current pattern when I have multiple AI-heavy growth/marketing tasks:
- Fire off the first complex prompt to ChatGPT thinking (ad research, competitor analysis, whatever)
- While it thinks, open Claude, Gemini, or another LLM reasoning model and start a second task (landing page feedback, email draft, etc.)
- When ChatGPT finishes, I review and refine its output (this takes maybe two minutes)
- Send the follow-up to ChatGPT, then return to your second AI tool
- Repeat with staggered timing across both tools
I'm basically orchestrating parallel AI workflows. The "wait time" between sending a prompt and getting results becomes productive time for other work.
I tried to find examples of other people writing about this pattern. Simon Willison and Gergely Orosz have both written about running multiple coding agents in parallel, but most of the discussion I found focuses on the agents themselves.
What I'm describing feels slightly different. The human is genuinely multitasking during the AI processing time. And because each AI holds the cognitive burden of its own complex task, the context-switch penalty seems way lower than traditional human-only multitasking.
For Discussion
I'm curious whether you're experiencing this too.
When ChatGPT or Claude is processing a complex prompt, do you wait? Or have you started working on something else? Have you noticed the context switch feeling different than it used to?
Hit reply and let me know. Am I genuinely unlocking efficiency through multitasking or am I just another "sucker for irrelevancy?"
Kevin DePopas
Demand Curve Chief Growth Officer
P.S. Any Stanford researchers reading this? Might be time for a follow-up study.
No comments:
Post a Comment