"Simplicity is the ultimate sophistication." - Leonardo da Vinci
The future of work is a portal away from screens, away from doing a lot of BS stuff. And it's not some distant sci-fi dream, it's happening right now.
Here's what I think it looks like: We'll talk into our computers, riff on what we want, and AI agents will execute it (connected to all our applications and context). The monotony of learning new UIs and repetitive manual tasks disappears.
And what's going to matter is clarity: knowing what you want, understanding the fundamentals of your job function, and how to get great work done. The AI is going to help clarify your thinking, and AI agents are going to do the work for you.
Why now
Being part of the Origami team has given me a front-row seat. Everyone’s a first adopter. We're building an AI-native product, so we adopt new tech constantly. And Eric and I have carried a massive workload.
When I joined Origami, I was in back-to-back calls from 7am to 7pm for three straight weeks with minimal support. We had to rely on new tech to automate work and not get drowned in it.
As a result, I've deeply intertwined LLMs into my work: hundreds of thousands of words into Wispr Flow, constant use of Claude, ChatGPT, and Gemini. I'm a sucker for doing only what I'm best at and eliminating the BS.
And now, with the announcement of Claude Cowork and seeing very talented friends show crazy AI demos, it's undoubtedly clear to me where the future is pointing.
The future of work is away from app interfaces, straight to getting shit done.
Here's the thing: all software applications are database wrappers. Before, you'd log into these apps, and manually patch things together across platforms. You were the one doing the work. Now, AI agents can do this patchwork for you. They understand how to use these apps, they can use your computer, and they execute reliably. You just tell them what you want done.
People are starting to realize:
"Holy sh*t, I don't have to hire someone or shell out 5 caffeinated hours on b*tch work. I can just prompt, prompt, prompt. Boom! A custom workflow embedded in my specific applications with deep understanding of my context and what I'm trying to do. And it does that work for me.
And if it's repeatable, I just store it: build once, run whenever. It's better than any other software because it's deeply embedded in my workflows and custom knowledge.”
The best software developer for you is you. The problem was, people couldn't compose and manipulate software easily. Now you can. You get software's flexibility, tailored to you.
What this looks like in practice
Alright, so what will this all actually look like?
You'll have a chat interface that plugs into all your applications: HubSpot, LinkedIn, Linear, Gmail, Slack, whatever. It'll have context on all your data—Slack messages, customer info, product iterations, emails—plus external data: best practices, recent news, and real-time APIs to access extra goodies.
You basically have the world's greatest intern: knows every application, accesses any knowledge (internal or external), acts on your behalf at a massive (but customized!) scale. You'll need to train it (give context, explain what you want) but it executes.
- A salesperson will be able to say: "Hey, go and look at all of my past closed-lost accounts in the past three months that closed due to budget. Check if they got a new round of funding. If yes, then go and find my main point of contact. Look through all our previous emails. Look through all of our call transcripts. And then draft a very custom email checking in with them… and oh by the way add a call script + task for me in HubSpot. Show me everything beforehand so I can edit.” Wait 2-5 min for the rough draft. Edit slightly. And then boom, sent.
- A product person can say: "Hey, I want you to look at this specific Slack thread of us jamming on this product. Go and create a Linear ticket that describes this new feature. First, go and check all of our previous docs and all of our product information and our code so that you can get all the relevant context. And then fill out a very detailed specific Linear ticket outlining what the product feature is, why it's useful, where within the codebase it plugs in. Then, create a GitHub repo and actually do the rough draft of the feature for me. Link that rough draft in the Linear ticket so my engineer can double-check it."
- Ops folks will say: "Hey, go look at all of our current SaaS subscriptions in our accounts, and then find me all the users on each of those accounts. Check our Stripe accounts to see how much those contracts cost. Give me a very detailed report of how much we spend on each specific piece of tech stack, how many seats there are, and who uses then. Then go ahead and suggest me the three ways that I can cut these things or consolidate these tools. Give me that report, and then based on that report, I'll tell you what I want you to do. Likely Slack message these people to see how much they use it. Then, we’ll go and cut costs where necessary."
- Marketing pro’s will say: "Hey, go and look at all of the recent tech product launch posts on X that got over 1M impression. Compile the list with links. Then, go into all of our most recent Linear pushes. Tell me what new features we could do our own launch for. Give me 3-4 announcement ideas for each feature."
How it works:
Then the chat will create a bunch of sub-agents to do work. It’ll do that work in the background. Then come back to you with an example. You’ll iterate. Prompt, prompt, prompt ‘til it’s good. Then do it across the entire set of work. Maybe even save the workflow for later use.
Everyone becomes a manager… and an individual contributor
The fundamentals stay the same:
- Product still needs to identify user problems, design solutions, and ship features that drive adoption.
- Engineering still needs to architect systems, solve technical problems, and ship features that scale
- Sales still needs to build pipeline, close pipeline, and turn revenue into higher contracts.
- Etc.
But the execution is done by AI agents (except for cold calling!).
The best workers will be both great managers and great ICs.
Managers: can clearly communicate where they are at, where they want to go, and what needs to happen (generally) to get there.
ICs: can get shit done, individually accountable, and can get into the details of granular work.
What actually matters now
The bottleneck is understanding how your systems pipe together. It’s your ability to articulate what is in your head and what your specific knowledge is. The bottleneck is your willingness to take action and get sh*t done. The bottleneck is your intuitive understanding of how to build a business, and what needs to happen now.
So, the three skills will become even more important:
- Clear thinking.
- Translating thoughts to even clearer writing.
- Ability to divide and keep multiple work streams running in parallel.
It's already happening
Look on LinkedIn or X you'll see a bunch of people showing off their Claude Code and Cowork workflows. I say you go pay $100 for Claude Cowork and try it yourself! It’s frame-breaking. Getting 10 hours of work done in 30 min is realistic (for a lot of tasks).
This “frame-braking” of AI is a trend you'll see across mostly every industry.
It’s already happening in education with Alpha School.
Students learn 2x as fast with only 2 hours of school per day!! It takes them 30 to 40 hours total to complete an entire year's subject and be great at it. Read those stats again. What??!?! It doesn't even compute in my brain.
The reason it doesn't compute is we're comparing outcomes from an old + broken system. A system where every step isn’t custom to the student w/ infinite knowledge on the subject. These students talk to AI and create custom lessons exactly tailored to what they need + like.
In the same way, workers are going to directly talk to AI and create custom outputs (whether it’s a CSV, website, email, new product, whatever).
Throw user interfaces out the window. People are going to create their own user interfaces. What matters is functionality.
It's beautiful (and a little dangerous)
Here's what people miss when they panic about AI doing all the work: This is what humans are meant to do.
We're meant to riff on ideas. Be creative. Be intuitive. Not spend five caffeinated hours on grunt work, trudging through different platforms and coordination loops.
BS work is out the door. What's left is stuff that energizes you.
You see it with those Alpha School students: they love learning!! I see it with my friends (and myself) doing this modern work style. We love it because we’re unlocked to do what we’re actually good at; things that we were meant to do.
You are not going to be replaced by AI (for now?). You're going to be freed by it.
Freed to think clearly. To iterate faster. To be more creative. To handle far more than you could handle before (and do a better job at it!).
The monotony disappears. Manual tasks disappear. What remains is you, your ideas, and agents executing them.
It's beautiful. And it's also worth pausing on what we're actually building here. We're becoming more abstracted from the actual work, in the same way farmers are abstracted from farming. Farmers today use GPS-guided tractors, automated irrigation systems, etc; most don't actually know how to plow a field by hand.
The same thing is happening with knowledge work. We'll be able to get more work done, but we'll be farther from how things actually happen. It’s complicated, but the building is getting taller, but the foundation is getting narrower. Society is getting stronger and more fragile.
You draw this out a hundred years, and people might be 1000x more productive, but the power goes out and shit hits the fan even more. That's something very worth discussing and thinking about, because you don't want to build a great society that can crumble if a few things disappear.
To be quite frank, I don't really know what work will look like in five years. But, I do know people who hate BS work and love getting stuff done will have a glorious next few years.
The portal is open. Time to walk through it.
- L
Read more pieces here.
Read my newsletter here.