An AI productivity secret hiding in plain sight
Summary
Most executives are reaching for ways to become an AI-first organisation, but many of them are missing the productivity power up that’s hidden in plain sight – asynchronous collaboration. Not only does asynchronous collaboration free up unproductive time for workers, it helps create the digital exhaust of knowledge that LLMs can feed on.
A few weeks back, I was in an office. I was one of 20 people in a meeting, sharing intelligent-sounding thoughts about twice a day and doing my best to keep up with the back-and-forth. Most conversations dissipated into thin air. If anything, someone recorded a bullet or two into a random slide of a PowerPoint deck. By the end of the week, it didn’t feel like the team or I were any wiser or that we’d used our time productively. Yet, everyone nodded their heads in agreement.
Conversations don’t always lead to clarity
I don’t want to use the example of a poorly run in-person experience to diss all in-person experiences. Still, that week in an office made me reflect on how misaligned most in-person experiences are to an AI-first way of working.
Imagine the same set of meetings on Zoom, but now with an AI-first mindset.
Not everyone travels to an office, so not everyone needs to attend every meeting. Each meeting can have only a small number of contributing attendees and a narrow focus.
Collaborative whiteboard tools like Mural can help everyone contribute on an even footing. Not just the person in charge of the projector. AI features built into such tools can speed up synthesis, decisions and agreement.
AI-powered transcripts and summaries, and digital whiteboards can help people who don’t attend a meeting catch up on what happened and avoid FOMO.
After all meetings conclude, AI can help us synthesise the documentation trail we’ve created all week. We’d still nod in agreement, but this time it’ll also be a nod to our shared reality.
Clarity emerges from a shared reality
Blending remote, synchronous and asynchronous work practices helps us achieve better outcomes than simply rocking up to an office and hoping for magic.
As you know, I’m AI-bipolar. But if I wear my AI optimist hat, it feels like we’re living through a strange time in tech leadership. On one hand, executives are issuing "return to office" (RTO) mandates, arguing that innovation can only happen through serendipitous hallway conversations. On the other hand, the same leaders are pushing an "AI-first" agenda, desperate to unlock the efficiency that LLMs promise.
Here’s the uncomfortable truth: you cannot have both.
The "office-first" hypothesis is at fundamental odds with the "AI-first" hypothesis. If you are betting your company's future on AI while simultaneously forcing your team back into the physical office to "collaborate," you are driving with the parking brake on.
AI can’t learn from the water cooler
The primary argument for RTO usually revolves around the "magic" of in-person interaction. The idea is that people make critical decisions and let creative sparks fly when they bump into each other in a physical space. But in contrast, think about how an LLM works. It can’t read your mind or listen to the watercooler conversation.
It needs data.
It needs context.
It needs a digital trail.
Office-first workstyles prioritise oral communication over written. Nothing stops people from documenting their interactions in an office, but the fact is that when we’re in person, we don’t create as many written artefacts. For example:
We make decisions and never record them.
We share context over a coffee and never transcribe it.
Ways of working are implicit and take hours of “shadowing” to teach.
If your company operates on a proximity principle, it never builds the knowledge infrastructure for AI to succeed. You can throw as much AI as you like, but without a solid corpus of information, even a basic chatbot will spew generic nonsense. Until you bring people into a room and throw information up onto a projector, you’ll only feel your way around and make guesses like the characters in the parable of blind people and the elephant.
In contrast, effective remote-first teams are also async-first teams. By necessity, they write everything down. They create the "digital exhaust" that fuels intelligent systems. Location doesn’t matter. Clarity can emerge anywhere.
The meeting tax vs. deep work
AI without deep work is a recipe for AI work slop. I’ve written earlier as to why we all must cultivate taste. It’s almost impossible to cultivate taste if your work only involves several AI button presses punctuated by scores of meetings and a tsunami of email and instant messaging.
While writing my book, I surveyed 1800 technologists to learn about their remote work patterns. Back then, the average respondent was spending a third of their time in meetings, and a third of those meetings were ineffective. That’s a colossal waste of productivity. Add instant messaging and other interruptions and and you end up with an environment that’s downright hostile towards deep work. And in case you think that collaboration practices have gotten better, think again.
Meetings and interruptions have a colossal impact on productivity
If AI indeed helps us produce more at lightning speed, then we’re dealing with the challenge of entropy. Challenges and bottlenecks emerge in days and weeks rather than months and years.
The problem with the office-first and meeting-first hypothesis is that presence in meetings and hallways becomes a proxy for productivity. Shallow work becomes the norm. Shallow work won’t help you deal with AI-powered entropy. It’s as simple as that.
The secret hiding in plain sight
My conclusion is simple. Most knowledge work companies will use AI as part of their toolset, just as we use other technologies. And to be AI-first in that regard, will also mean embracing an async-first mindset.
Async-first isn't just about working remotely. It is about shifting your operating system from "being present" to "being effective." Here’s how I’ve advocated for the shift in a previous article.
Smaller, more capable teams: Instead of large, siloed groups, favour smaller teams (perhaps 3-5 people) where members possess a wider range of skills, augmented by AI. Think "T-shaped" or even "comb-shaped" individuals who have deep expertise in one or more areas but can contribute across adjacent domains.
End-to-end responsibility: Allow these smaller teams to own problems or features from conception to deployment. Expect higher ownership and fewer handoffs. Use team APIs to agree on cross-team collaboration patterns.
Increased individual focus: If AI handles routine tasks and potentially complex implementation details, individuals can spend more time on higher-order strategy, user needs, and focused deep work. Yes, with AI, individuals can operate at the level of intent, while machines handle the implementation. Describing the purpose, evaluating, and fine-tuning AI outputs are no trivial pursuits.
Reduced synchronous communication: When teams are smaller and individuals have broader capabilities, you don’t need constant alignment meetings. You can use async-biased collaboration models, such as Shape-up, to plan, prioritise and execute work. Reserve synchronous communication for high-stakes interactions, such as making important decisions or resolving conflicts, or to address time-sensitive issues.
Most importantly, avoid the allure of tracking badge swipes. Instead, encourage transparent work practices that generate knowledge and information for AI to feed on.