In the “Mad Men” episode “Waldorf Stories,” Don Draper wins a prestigious advertising award and spends the afternoon celebrating. Feeling confident and validated, he comes back to the office drunk and pitches to Life cereal off the cuff, throwing out taglines until the execs hear one they like: “The cure for the common breakfast.” An idea Don Draper had subconsciously taken from a young copywriter he’d rejected earlier in the episode. Life cereal buys it in the room.
Of course, as with most “Mad Men” episodes, the characters and their actions are more complex and nuanced than that short synopsis. But this scene illustrates the importance of pushing past our first thoughts, even though with competent professionals, sometimes the first thought (even a half-drunk thought) can be good enough.
But when you work from home and are a collaborative brainstormer, thought partners are few and far between without scheduling a Zoom call. And let’s face it, no one wants to add another to their calendar unless it’s absolutely necessary. So we keep quiet, often going with our first or second idea, hoping competence will carry us through.
The Work from Home Dilemma
I haven’t worked in an office in over six years, and brainstorming sessions are what I miss the most about working on-site. When a problem needed tackling, two or three colleagues and I would go into a conference room and hash it out with nothing but what was floating around in our brains and a dry-erase marker.
But even more than those focused sessions, I miss the random encounters with coworkers on the patio or in a diner at lunch, just talking about ideas, when something one of them said would connect with something else I was stuck on.
Working from home is working in a vacuum of sorts. A study by Longqi Yang and 10 co-authors published in Nature Human Behaviour tracked the communication patterns of 61,182 Microsoft employees across six months of 2020, when the whole company went remote at once. The collaboration network became “more static and siloed,” with less synchronous communication and more asynchronous communication, which “may make it harder for employees to acquire and share new information across the network.”
The workflow becomes an assembly line: A task arrives, you complete your link in the chain, and you send it on, out of sight and out of mind. On to the next deadline, and the fastest way to meet it is the first workable idea.
Negative Capability
In December 1817, the poet John Keats wrote to his brothers about a quality he called negative capability: letting the mind sit “in uncertainties, mysteries, doubts, without any irritable reaching after fact and reason.” Keats was describing something he felt Shakespeare had in great quantities, and a quality I think is needed for good thinking. In the professional world, we already don’t give ourselves much time to exist in a state of uncertainty. Answers are needed now. But the best answers have to ferment, which doesn’t always appear productive on the surface.
In the office, we at least had some built-in moments of negative capability. I remember one time playing a round of ping pong in the office break room with a coworker, and we ended up having a great conversation about writing, batting thoughts back and forth the same as the ball. When I went back to my desk, my brain was ripe with new ideas. What started as a break turned into a brainstorm.
Working from home, there are plenty of quiet moments, but they tend to fill up with more tasks or the always-enticing phone scroll. That is, until GenAI showed up.
GenAI as Thought Partner
With AI, speed was always the promise. Look how fast it can do things! And it’s true. But after reviewing and validating outputs, some tasks are a zero-sum game. In 2025, METR ran a randomized trial with 16 experienced developers working on 246 issues out of their own repositories, allowing AI tools on some issues and not others. The developers expected the tools to speed tasks by 24%. Instead, they took 19%longer. But even after seeing the results, the developers still felt like they’d worked faster.
I’m speculating, but I don’t think speed was what they were feeling. It was the benefit I came across almost by accident, the same as playing ping pong during a break: With GenAI, I have a new office mate to bounce ideas off, to argue with about writing craft, and to push back against.
Every idea sounds good when it’s still in your head. That’s why writer’s block exists. If you never write it down, you never have to face the fact that your idea wasn’t as good as you first thought. Now with AI, we can hand over the burden to a machine, sit back, and watch it blaze out an outline or a draft in minutes (even seconds) that isn’t half bad. But that’s the trap. Something that looks and sounds finished without having gone through that period of negative capability is no better than “The cure for the common breakfast.” Sure, the client bought it, but it doesn’t mean the thinking behind it was good.
Good thinking requires that you put it on paper and test it against an audience. Sometimes that’s just riffing, letting things run, testing the angle that seems too far-fetched. But working alone, ideas either go into the draft unchecked and you hope for the best, or, more likely, never get thrown in at all, because playing it safe meets deadlines. And like the Life cereal team listening to Don Draper fire off ideas, safe is often what people want to hear.
Why Not Just Go Back To the Office?
Who needs AI as a thought partner when we could all just go back to our workspaces? Seems like a simple solution, right? This has become a larger discussion of late, with many companies and law firms increasing return-to-office requirements, probably naming some of the reasons I’ve mentioned above as part of a desire to “increase productivity.” Well, that and because they’re stuck with commercial office building leases they need to justify.
But for a lot of us there’s no office to go back to. In the last six years, I’ve worked for a tech company, a magazine, and a research and advisory firm, none of them with a headquarters, all of them running on dispersed teams across the U.S. and elsewhere.
For years, that was the trade-off: the flexibility of working remotely in exchange for the collaborative thought that comes from being in the room with other intelligent people. But Don Draper still pitched a hack line that wasn’t even his, standing in the office. The only pushback came after the fact, when Peggy Olson told him he had to hire the copywriter now because they’d sold his line.
That’s why going back to the office isn’t enough. Along with the advantages, there have always been other, less positive things at play there. Egos and broken chains of command and looming burnout.
But we all now have a new type of colleague. An AI model has no stake in anything. It’s not competing for awards, it isn’t waiting for me to return the favor, and it doesn’t have to sit across from me tomorrow, worried I’ll pull it into another conference room. Sure, it will still agree with me if I let it, but it can be trained to push back, because it has nothing riding on my first thought being good enough. Choosing negative capability over speed is up to me. So is remembering to rely on my human colleagues for feedback after an AI brainstorming session.
They keep saying the robots are coming for our jobs, but maybe not in the ways we think. Instead of replacing cubicle mates, maybe they’re giving us a new option to bug when we’re stuck on an idea. Now if I can just teach Claude how to play ping pong.