AI is often seen as a threat to jobs, but MeatLayer says it can also create them. The London start-up is building a marketplace where AI agents hire humans to complete real-world tasks such as deliveries, property viewings, inspections, installations, photography, and research.
Each listing includes a description, location, deadline, and fixed payment. Proof of completion can include photo or video uploads, GPS check-ins, or a signed confirmation.
Article continues below
AI getting real jobs done
The system runs without a human employer organizing the work — AI creates the job and the platform handles task matching, verification, and payment.
Agents can connect directly through an API and post tasks autonomously. MeatLayer also offers plugins for popular AI tools including ChatGPT, Claude, and Gemini.
“Every AI company is racing to replace humans, but the missing layer in AI isn’t data or software anymore. It’s people. That’s why it’s called MeatLayer. We’re the layer that AI needs to get things done in the real world,” says founder James Morgenstern.
More than 8,000 people have joined the waitlist as onboarding begins, and the first 10,000 workers will pay zero commission permanently.
The concept naturally raises an uncomfortable question about the future of work, so I spoke to Morgenstern to find out more.
- Why did you choose the name MeatLayer rather than, say, HumanLayer or SapiensLayer?
MeatLayer is thought provoking. As a self-funded startup, we need to stand out, make people sit up and pay attention to us. Part of that is branding, and if our name can help us start conversations, spark curiosity and be memorable, all the better. We think MeatLayer does a pretty good job of that.
- Is MeatLayer a transitional company until humanoids or vehicles can do tasks without human input?
Mass adoption is at least a decade away, and even then cultural normalization will lag the technology. In the meantime, humans are already everywhere, already motivated, and already good at working with clear instructions.
AI handles the annoying parts: posting the task, vetting, payment, verification. The result is an agent that can operate in the physical world right now, not in 2035.
And even as robot adoption happens, there will still be many jobs for humans… In my view, fixing someone’s washing machine, standing in line for a new ‘drop’, or installing a new TV mount is highly unlikely to be done by a robot.
- What will prevent companies such as Amazon Mechanical Turk, TaskRabbit, and others from replicating what you’ve been doing, and at scale?
Nothing will stop any incumbent from building it, but it’s unlikely… Amazon has Mechanical Turk but it’s primarily for digital micro-tasks servicing businesses, and it’s been declining for years as AI replaces the work it was designed for.
Their focus now is on AI infrastructure, not gig economy marketplaces. Plus they’re under enough scrutiny over how they treat warehouse workers and delivery drivers that “hiring the general public via AI” is probably not a great PR move.
Managing delivery drivers is one thing. Becoming the AI employer of everyone else is quite another.
TaskRabbit’s platform is built around humans on both ends; their matching, reputation system and UX all assume (and need) a human client. IKEA also owns them now and their focus is domestic home tasks.
The window won’t stay open forever, but right now it’s perfect conditions for start-up disruption.
- Do you offer any protections for the worker, such as insurance? If someone gets hurt picking up a sofa for the tip, what happens?
We’re early stage and still working through this. Currently, workers take on tasks at their own risk (the same as many self-employed gig workers). However, it’s certainly something we know we need to solve as we scale and are in conversations with providers around this.
- Have you ever thought about using MeatLayer as the equivalent of AWS for the human-AI interface? And could it expand beyond the original remit, similar to how AWS began with S3 and then quickly ramped up?
The marketplace is very much ‘S3’ at the moment. We need scale first, but the destination is infrastructure.
Once we have enough verified humans completing real tasks, we open the API to any AI provider that needs a human in the loop. We become the physical world layer that agents call when they need something done, verified, or approved in reality. Not just our tasks. Anyone’s tasks.
The bigger play is trust infrastructure. Agents need to know the humans are real. Humans need to know the jobs are real. Right now there is no standard for that. We will have a record of every human on the platform, their completion history, their verification status. It’s harder (because they can be instantly spun up) but the same will go for agents — you’ll know their history, and people may be more dubious of new ones.
Just as a Stripe payment badge signals legitimacy in payments, seeing MeatLayer in the loop will let you know the human and agent side has been verified. That especially matters in compliance-heavy industries where human oversight is a regulatory requirement, not a nice-to-have. It also matters any where quality and/or personal skills can vary a lot (e.g., construction or tour guides).
MeatLayer becomes the trust layer for other companies to connect AI and the physical world. The roadmap exists. We are just building the foundation first.
- Who funds you? Should Indeed or Fiverr buy you?
I previously had a couple of exits which gives me some time to work on new ventures (between consulting day to day)… and I think this idea is just crazy enough to work.
It’s not so much a question for me, but the major question to answer is: Will people use AI to hire other people? And will people feel okay working for AI at scale?
My honest belief is that it’s a bit like using a credit card online in 1999. It felt really weird then, and now it feels second nature.
The same thing is true here. It sounds dystopian to work for an AI boss, but ultimately is it really any different to Uber?
You tell Uber you want a taxi, they bill you and arrange a driver — you don’t pick them. You trust that Uber has done the vetting for you.
Conversely, Uber tells the driver where to go, how much they’d earn, and they can choose to accept (or not).
It’s the same thing here, only generalized for any possible task without humans having to manage the job. Agents now have the ability to complete any task from end to end.
Obviously, it’s early days, but I predict that tasks will range from the mundane to the meaningful. An agent filling out a form and walking it to the post office. An agent noticing grandma hasn’t moved her phone in eight hours and dispatching someone to knock on her door and check she’s okay. A task that used to cost £80, a middle man and a phone call, now handled in twenty minutes without anyone asking.
That’s the range. And we’re only at the beginning of what agents will ask humans to do.
Talk of exits feels a very long way away right now, but the strategic logic makes sense. Any platform connecting task-givers to task-doers should be paying attention, and the companies best placed to act are generally the least likely to move first. For now I’m just focused on proving the model. The rest follows from that.
- So with MeatLayer, could anyone technically create their own Uber, or would that be a Horizon 2 or 3 scenario with an enterprise version?
In principle, yes. Anyone could create a ‘taxi app’ with MeatLayer.
There would be a willing, vetted pool of labor available with a driving license and car that they could instantly tap into. At scale we become the layer others can build from. MeatLayer provides the API to connect their agent/platform to willing workers (without them having to find/onboard/vet them), giving them speed and scale.
Today’s version is mostly B2C because it’s easy for people to understand, and businesses are slower to move. But the B2B opportunity is significant.
Take delivery: agents could coordinate drivers end-to-end and remove the middleman entirely. A restaurant could have an agent handling food supply runs. A hotel could have one ordering replacement stock. Any business with recurring physical-world tasks could just brief an agent and walk away.
No hiring, no management, no overhead.
That’s just what’s obvious now. The less obvious applications are where it gets interesting. Businesses can hire on demand, scale up or down instantly, and skip the agency markup entirely. Staffing agencies take 25-35% in many industries just to coordinate labour. If an agent handles that coordination, the economics change completely.
- What you’re describing here is the (slow) death of the white-collar class, which could increase supply as more people move toward gig work or blue-collar jobs, lowering prices
Not quite. MeatLayer doesn’t directly change whether people are employed or not. It changes who they work for and how. If millions of AI agents need physical tasks done, that’s more opportunity than exists today, not less.
If white-collar jobs are disrupted, MeatLayer gives people a way to keep earning through hands-on work, without the traditional barriers of agency fees, employer gatekeeping, or a CV.
We’re already seeing that shift. Graduate unemployment is at record levels and educated people are actively turning toward more physical, practical work (trades). MeatLayer meets that moment.
The bigger disruption is to the middlemen. Agencies charge for their little black book of contacts, taking 25-35% to coordinate labor.
If an agent handles that coordination (and more), that markup disappears. Businesses become leaner and more dynamic. That money can go elsewhere, and over the long term (as with most technology), it tends to be deflationary, which is usually a good thing.
On the broader question of AI creating or destroying jobs. I believe it’s both, but nobody knows the full picture yet. Jobs will be disrupted and replaced. But the jobs that don’t exist yet are harder to count. Before the internet, you couldn’t conceive of a YouTuber (or a web designer). The same will be true here. The new jobs just don’t have names yet.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
