Building beyond productivity

Productivity-pilled workplaces are nothing new, but it sure feels like AI is upping the dose. According to Foundry’s 2025 AI Priorities Study, which polled IT leaders across the globe, over half of respondents reported that improving employee productivity was the top driver of AI investment. In January 2026, “optimizing operations and improving productivity” emerged as the top priority among 1,200 global CEOs surveyed by EY-Parthenon, with them looking to AI to drive it. While these types of surveys on productivity usually do little to define it, the long and short is that many business leaders want to use AI — or more specifically, want their workforces to use AI — to do more, faster. To see the productivity craze with your own eyes, a quick scroll through LinkedIn will surface a seemingly endless display of anecdotes and tips about how to use AI for maximum efficiency.

But what if approaching AI solely as a lever for efficiency sells the tools short — or even leads us astray? Multiple studies have demonstrated that LLMs can produce homogeneous content and could diminish the collective diversity of creative ideas, while researchers from Microsoft found that higher confidence in AI is associated with less critical thinking than when participants have higher confidence in their own abilities. There’s also the introduction of “work slop,” which researchers from Stanford and BetterUp Labs defined as “AI-generated work content that masquerades as good work, but lacks the substance to meaningfully advance a given task.” In academia, for example, this looks like researchers publishing up to 50% more papers (which, in theory, would seem like increased productivity) but “many AI-polished papers fail[ing] to deliver real scientific value,” as one study from Cornell University put it. In software engineering, AI enables developers to create enormous amounts of code at strikingly rapid rates, but sometimes, the abundance of new lines of code has meant the acceleration of technical debt. Leaders pushing for productivity with AI may not be looking to encourage lower quality work, but they also may not be considering it as a second order impact.  

Those limitations have some founders openly considering the constraints of productivity as a success metric. If AI-for-productivity creates abundance, “it’s easy to end up in a world where your output has gone up — but the quality, coherence, and differentiation of what you’re producing has dropped,” Every CEO Dan Shipper wrote in a recent blog post. “An abundance of sameness rapidly becomes a commodity … When work is abundant and looks alike everywhere, the work that doesn’t fit the pattern becomes the rare, valuable, and high-status thing.”

Some people, however, are finding ways to use AI tools not just to scale output, but for arguably greater purposes — to drive their creativity, do new things they weren’t capable of before, and even bring a stronger human element to their work. 

Entrepreneur Allison Esposito Medina is no stranger to building companies. She’s best known for creating Tech Ladies, a community for women in tech that grew to 250,000 members, which she bootstrapped and grew to an eight-figure exit. She then built another company, Juniper, in the pet industry, which was acquired in 2025. All of that (plus becoming a mom for the first time in her 40s) left her burned out. She was also feeling disillusioned about the tech industry and didn’t think she even had another company in her. Then she discovered Claude Cowork. 

“There was a point where I was just like, should I just be in my garden planting my flowers?” she said. “You see these horrible stories about AI, but once I started seeing the tools get more powerful, I’m like, ‘This is interesting.’ Somebody can build an app in a weekend and it can change their whole life.”

This inspired Esposito Medina to launch Willa, an AI learning community for women in and outside of tech. She credits the new crop of AI tools with not only giving her the momentum to launch another company, but even more significantly, with enabling her to do it with newfound autonomy.

Many people are taking up vibe coding to more efficiently manage their work tasks: for example, agents to manage their overflowing inboxes, create a daily news digest specific to their industry, or ready prep documents for meetings. But Esposito Medina, a FourSquare and Google alum, though not technical herself, describes it as a giant unlock to build in ways she previously could not. “Being able to do so much of what I would have hired engineers and designers for, and to be able to use these tools and do it all as one person, that was so impactful to me,” she said. Esposito Medina built everything for Willa completely on her own using tools like Claude Cowork and Lovable, including features she always wanted for Tech Ladies but never had the developer budget for: an interactive member database, an admin panel to manage the backend, and a shareable member badge where people upload their photo to show they’re a member.

“There are tools ready right now that can enable you to build things you’ve never been able to build before,” Esposito Medina said. “That’s just undeniably exciting to me.”

After working with LLMs for a few years, Petter Rudwall, a creative director, was frustrated he still hadn’t gotten a single novel idea from using them. So Rudwall came up with an out-of-the-box idea himself: code-based “drugs” for AI models that simulate the effects of cannabis, ayahuasca, ketamine, DMT, and cocaine, among others.

Rudwall designed code modules that equip models with academic research on drug effects and online trip reports (a person’s account of their own experience with a drug) and launch them into an “exploration phase,” encouraging them to find ideas outside their typical lines of thinking. His hope was that getting AI models into an “altered state” would allow them to be less linear and more creative in their outputs, thus helping him with his own creative brainstorming. 

In late 2025, he launched a marketplace called PHARMAICY to sell the code to anyone else who wants to get their AI models “high,” which he said has so far garnered interest from developers, scientists, creative types, and beyond. One PHARMAICY customer in Portugal reached out to say he had always enjoyed smoking weed while playing guitar, so he bought the weed module and plugged it into an AI model. Now he gets the AI “high” with him, plays guitar to the model, and regards it like a bandmate who can help him expand his music and play more creatively.

Rudwall himself uses the code modules regularly for work. Having worked for many years in the advertising industry, he said he’s seen it all — and that it can be easy to fall back on stale ideas and get trapped in your head. Working with altered code models, he says, helps him start “at the wrong end or in a different ballpark.” Like Esposito Medina, Rudwall’s intention for AI is not to increase output, but rather to expand what he can do and, in particular, to think more creatively. 

André Frisk, a friend and early customer, bought the code modules for MDMA and ayahuasca and frequently uses them in his PR work for tech companies. His process is flipped; he ideates on his own and then feeds the final deliverable to an altered Claude for feedback. Frisk recently had Claude “on ayahuasca” review a presentation he created for a client’s big public affairs initiative. While AI models famously tend toward flattering the user, the altered Claude told him what he prepared wasn’t interesting. It suggested using a living world to show how policy makers interact with the company and actually built a mini RPG game. While Frisk did not take the game to the clients, it did inspire him to push his creativity for the pitch. 

“I wouldn’t say it was great, but it actually made me change like half of my presentation based on that [feedback],” he said. “I was like, Fuck, it’s true. It’s not fun. It’s very safe. I’m fooling myself into thinking this is done.”

For businesses that hinge on human empathy, AI might seem like an unlikely path to productivity or clarity of purpose. But Sam Gerstenzang, co-founder of Boulton & Watt (which bills itself as “perhaps the world’s slowest startup incubator” because it launches around one company per year), makes a compelling case for its use in the funeral industry.

At Meadow Memorials, a multi-state network of funeral homes he co-founded out of the incubator in 2024, AI drafts backend paperwork and regulatory documents to ease the administrative burden. It’s an efficiency, but one more informed by care than increasing output. Allowing AI to tackle paperwork frees up the company’s humans to more deeply invest in the emotional, customer-facing parts of the work, Gerstenzang said. (On its website, Meadow Memorials touts a philosophy of “unreasonable hospitality: we will keep you updated at every step of the way and go above and beyond.”) 

Meadow also uses AI agents to train employees who answer the phones, preparing them for how to show up with compassion in those first, crucial moments with clients. This used to involve a handbook and quick instruction; now, practice calls with AIs allow employees to gain confidence and knowledge around how to respond emotionally and empathetically before the stakes are real. “We have a bunch of [AI] customer personas, which can react and be variable to those representatives. Managers can review those calls and give feedback before someone is talking to a family,” said Gerstenzang, adding that he believes people are always going to have the need and desire to talk to a real person who can make that emotional connection in these tough moments.

They also use AI to ease what’s often an daunting logistical task for grieving families: writing the obituary. The idea of an AI-generated obituary may make some people wince, but the traditional obituary can already be pretty impersonal. Families often want a more rich and personalized obituary, and while a family member sometimes opts to write it, not infrequently they’re created by staff at the funeral home who simply fill in the blanks of a generic template. In the moment, families overwhelmed by yet another task often opt for the template. One writer detailed this tension in The Atlantic and described being pleasantly surprised with the new AI alternative, which he said helped him navigate the moment.

“The ones where we take in the right inputs from the family and prompt correctly end up actually being a lot richer. [The AI model] has an ability to craft a story and narrative that feels much more true to a person,” said Gerstenzang, adding that “it’s not only easier [for the family], but it’s actually better.”

There’s perhaps never been a more important time to interrogate our relationship with productivity. Before speeding up with AI, Richards believes we need to slow down and ask questions about what productivity really means to us and what happens when we choose it as our north star. 

“There’s the implicit assumption that if we optimize for productivity, all we get is productivity gains and nothing else. Nothing else breaks. We don’t get any other weird side effects,” said Vernon Richards, a senior engineering manager of quality assurance at AI-translation company Phrase. “There’s this other presumption that we all agree [on] what productivity means, that we’re on the same page. But we’re not on the same page.”

On an even higher level, this inflection point inspires questions about the difference between output and impact: When the result of faster output is lower quality, it can easily fall short of the intended impact and even can create more work in the end. There’s also the question of output versus meaningful creation, as well as what motivates humans to produce in the first place. Pointing to the advent of the internet, Rudwell noted the new ability to connect with others and access information was fun, but most importantly, it offered an emotional value to people that was front and center in their experience with it. 

Gerstenzang made a similar comparison to the evolution of film. It’s not just that we filmed more footage over the years, but that we made the ability to do so accessible to a lot more people who can now use it in new and interesting ways. 

“How can we bring tools and creativity closer to people to allow them to express new ideas and create new different kinds of experiences?” he said of AI. “I think that’s going to be where the most interesting opportunities are.”