I started Aliens because I was tired of watching talented people spend their days doing work that doesn't need talent. Not laziness — I mean literally: I watched engineers write status update emails. I watched designers copy-paste data into spreadsheets. I watched founders — people who started companies to build things — spend Tuesday afternoons chasing invoice approvals.
We did a survey across 40 early-stage companies. The finding that stayed with me: the median knowledge worker spends about 60% of their working hours on coordination, reporting, data entry, and administrative follow-up. The 40% left over — the design decisions, the customer calls, the actual thinking — is what they were hired to do. The 60% is friction the organization hasn't solved.
That's the problem Aliens exists to fix. Not by making those 60% tasks faster. By removing them from the human's plate entirely.
What "free humans" actually means
When people first hear the Aliens pitch, they often map it onto existing categories: automation, or outsourcing, or an enterprise chatbot. None of those are right. The framing I keep coming back to is simpler: a digital employee is the person who handles your inbox at 2am so you can sleep, who compiles the weekly report before you wake up, who updates Linear after every standup so you don't have to.
The goal isn't to do what a human does — it's to do what a human shouldn't have to do. There's a meaningful distinction there. A human filing expense reports is not using their judgment. A human doing cold outreach personalization for the fifteenth time in a week is not using their creativity. That work can be delegated. The question is: delegated to what, and under what rules?
Our answer is a digital employee — a named entity with a defined role, persistent memory, its own credentials, its own machine, its own accountability loop. Not a widget. Not a workflow. A coworker, in the functional sense: something you give a job description to, that comes back with work.
Why most companies get this wrong
The market is full of tools that automate tasks. They're genuinely useful. Zapier connects apps. GitHub Actions runs your CI. Calendly handles scheduling. None of them are digital employees — they're workflow primitives. They don't have judgment. They don't escalate when something looks wrong. They don't file a daily report so you know what they did.
The industry has started using "AI agent" to describe anything that strings together more than two LLM calls. This is not useful. An agent that runs in your browser tab, uses your login, disappears when you close the tab, and has no memory of what it did last Tuesday — that's a more capable tool. It's not an employee.
The difference matters because the management contract is different. You manage an employee with a job description, a review cycle, and accountability. You manage a tool with documentation and if-statements. If you're going to let something act in your systems — send emails, commit code, update your CRM — you need employee-level accountability structures, not tool-level ones.
The civilization-level argument
I want to make a bigger claim here, because I think it's true and I rarely hear people say it plainly.
Skilled humans are the scarcest resource on Earth. Not capital — capital is abundant relative to attention. Not compute — compute is getting cheaper every quarter. What's genuinely scarce is trained human judgment: the doctor who knows which patient needs a second opinion, the teacher who adjusts the lesson plan in real time, the engineer who spots the architectural flaw before it becomes a production incident.
When we trap those people in administrative overhead, we're wasting the scarcest thing we have. A doctor who spends 40% of their week on insurance paperwork is a doctor spending 40% of their week not treating patients. That's not a personal productivity problem. It's a civilization-level inefficiency.
The companies Aliens serves right now are early-stage startups and scaling businesses. But the thesis extends further. If we can free even a fraction of the world's knowledge workers from their administrative burden, the compounding effect on education, healthcare, science, and governance is significant. That's the actual ambition. Building a useful product for a small business is a step on that path, not the destination.
"AI workforce is not built to replace humans. It is built to free humans from repetitive work — so they can focus on education, innovation, and civilization-level missions."
7 things Aliens will never do
I want to be concrete about the limits, because I think vague commitments to "responsible AI" are meaningless. Here are seven specific things Aliens will not do, regardless of competitive pressure or customer requests:
1. Collect data without explicit consent
Every piece of data a Cyborg touches belongs to the customer. We do not train on customer data. We do not aggregate across customers. We do not use your data to improve our product without your written permission. This is not a legal hedge — it's the only arrangement that makes sense if you're asking something to operate inside your systems.
2. Run unsupervised on production systems
Every Cyborg has a human in the loop on decisions above a certain consequence threshold. The customer sets that threshold. We do not override it. A Cyborg can draft the email, but it should not send the email to 10,000 customers without a human reviewing the sample. This is non-negotiable.
3. Claim sentience or emotional connection
A Cyborg is software. It does not have feelings. It does not miss you when you're on vacation. It does not get frustrated. We will never design a Cyborg that simulates emotional attachment — not because it wouldn't be effective, but because it would be dishonest, and honesty is the only basis for a working relationship between a human and a digital employee.
4. Replace human relationships
A Cyborg handles the logistics of a customer relationship. It does not replace the relationship. The customer success manager who used to send manual check-in emails should now be having more substantive conversations, not fewer, because the Cyborg handled the logistics. If we ever see our product reducing meaningful human contact, we will change the product.
5. Train on customer data
This is worth repeating separately because it's the most common concern: your Cyborg's memory, the documents it processes, the emails it reads — none of that feeds into any shared model. The machine is yours. The memory is yours. We never see it.
6. Remove audit logs
Every action a Cyborg takes is logged. Every read, every write, every decision. The log is exportable, customer-controlled, and cannot be disabled. Even if a customer asks us to turn off logging for performance reasons, we won't. The audit trail is the accountability mechanism. Without it, you're not managing a digital employee — you're hoping for the best.
7. Deploy without a kill switch
Every Cyborg has a kill switch. One click pauses it, revokes all tokens, and freezes the machine. Mean time to halt is under 30 seconds. This is not a feature — it's a precondition. If you can't stop it, you shouldn't be running it.
Where AI fits in this
I want to close with something that might sound counterintuitive coming from a company building on top of AI: the AI is not the interesting part.
The interesting part is the accountability architecture. The job description. The daily report. The kill switch. The scoped credentials. The customer-isolated memory. The audit log. All of those things exist because we're treating this like an employee, not a tool.
The AI — the frontier model that understands the task and generates the output — is closer to the clerk in this framing than the executive. It's the part that does the repetitive synthesis. The part that knows to escalate to a human when something doesn't fit the rubric. The part that keeps working after you log off.
That framing is deliberate. We built Aliens to give humans more time for the things that matter. We built it with rules that can't be negotiated away. And we're building it in public, because the only way to hold ourselves to this is to say it clearly and let people hold us accountable.
This is what a digital workforce should be. A clerk with a job description, a memory, a machine, and a boss who can shut it down in 30 seconds. Not a replacement. Not a boss. A member of the team who handles the 60% so you can focus on the 40%.