Fundwork: AI marketplace where projects find specialists

Working idea: Funded AI Talent Marketplace / Marketplace Where Projects Find Developers.

In the era of the new AI boom, we can transform the concept of traditional freelancing and create something new for ourselves.

It would be ridiculous to spend a year implementing an idea that isn't going to take off. That’s why I want to present the idea to the public first and get some feedback.

My primary goal is to find 100 potential clients and 100 experienced freelancers who are willing to say: “Yes, I’d be willing to give this model a go”.

Main idea:

There is also a link to a landing page FundWork. It’s more concise and visual over there. There, you will be able to submit an application to participate in the project.

What i want to build:

No 100500+ applications and endless cover letters.
No Connects, paid applications, subscription plans, or paid visibility.
0% platform commission for freelancers. The client pays for the service.
At the initial stage, the marketplace will work with freelancers from a limited list of countries/regions. We intentionally want to limit supply while testing the model, without mass price competition and dumping. Geography can be expanded after the mechanics are proven.
No price dumping: freelancers don't compete over who can do the job cheaper.
The client proves that the project is serious with money. Before the project reaches freelancers, the client deposits a minimum protected amount or fully funds the agreed first milestone/project --- for example, based on an AI estimate.
Verified completed work should not become worth $0 simply because the client stopped the project. Payment protection and dispute resolution will be a separate mechanism in the full version.
If the project description is weak, AI asks questions and turns the idea into a structured specification that the client approves.
If a serious estimate requires hours of technical work, that estimation can become a separate paid task.
AI matches candidates to a specific project based on experience, similar projects, evidence of experience, specialization, availability, and budget/timeline fit.
The system does not search for the "#1 freelancer." It searches for specialists who fit a specific project.
One active full-time project through the platform, but future projects can be reserved in advance.
In the full version, future reservations will have clear rules and a financial commitment from the client, so a specialist's calendar cannot be blocked for free.
AI monitors stages, requirements, and deadlines. If it detects a risk, a manager steps in.
AI can perform an initial review of the result against the specification, but serious disputes and decisions about people are not made automatically by AI.
No single JSS-like magic rating.
Freelancers are not punished for being new to the platform.
Freelancers are not rewarded simply for having been on the platform for a long time.
Being new to the platform is not the same as being new to the profession. 19.1. A freelancer cannot take more than one active project at a time. They must finish the current one first.
Every candidate shown to a client must independently meet the project's quality threshold.
If a freelancer repeatedly has verified problems, the platform may stop working with them.
Initially: vetted Middle/Senior+ professionals only.
Initially: IT/software only - SaaS, backend, frontend, mobile, design, DevOps, AI integrations, and related areas.
Automatic CV/portfolio import from freelance platforms and unique AI-generated profile pages.
No time trackers or screenshot monitoring.
Fixed-price/milestones at the start, no hourly rates.
Minimize unpaid work: no multi-hour free estimates, pointless calls, or constantly monitoring a job feed.
No fake projects, bots, or artificial activity.
No talking to a wall: if one side disappears, the system/manager reacts.
Eventually: a forum/community where users can influence the development of the platform.

What i do now

I not building a huge platform yet.

I am looking for 100 potential clients who say: "Yes, I would try hiring this way."
I am looking for 100 experienced freelancers who say: "Yes, I want to receive projects this way."
After that, build a small MVP/private alpha.
The first real success metric is not registrations or upvotes. It's real projects that go through this new workflow and actually reach delivery.

If the model doesn't work, i shut down the experiment.

I'm looking for ideas and criticism.

What i want to build:

No 100500+ applications and endless cover letters.

No Connects, paid applications, subscription plans, or paid visibility.

0% platform commission for freelancers. The client pays for the service.

At the initial stage, the marketplace will work with freelancers from a limited list of countries/regions. We intentionally want to limit supply while testing the model, without mass price competition and dumping. Geography can be expanded after the mechanics are proven.

No price dumping: freelancers don't compete over who can do the job cheaper.

The client proves that the project is serious with money. Before the project reaches freelancers, the client deposits a minimum protected amount or fully funds the agreed first milestone/project --- for example, based on an AI estimate.

Verified completed work should not become worth $0 simply because the client stopped the project. Payment protection and dispute resolution will be a separate mechanism in the full version.

If the project description is weak, AI asks questions and turns the idea into a structured specification that the client approves.

If a serious estimate requires hours of technical work, that estimation can become a separate paid task.

AI matches candidates to a specific project based on experience, similar projects, evidence of experience, specialization, availability, and budget/timeline fit.

The system does not search for the "#1 freelancer." It searches for specialists who fit a specific project.

One active full-time project through the platform, but future projects can be reserved in advance.

In the full version, future reservations will have clear rules and a financial commitment from the client, so a specialist's calendar cannot be blocked for free.

AI monitors stages, requirements, and deadlines. If it detects a risk, a manager steps in.

AI can perform an initial review of the result against the specification, but serious disputes and decisions about people are not made automatically by AI.

No single JSS-like magic rating.

Freelancers are not punished for being new to the platform.

Freelancers are not rewarded simply for having been on the platform for a long time.

Being new to the platform is not the same as being new to the profession. 19.1. A freelancer cannot take more than one active project at a time. They must finish the current one first.

Every candidate shown to a client must independently meet the project's quality threshold.

If a freelancer repeatedly has verified problems, the platform may stop working with them.

Initially: vetted Middle/Senior+ professionals only.

Initially: IT/software only - SaaS, backend, frontend, mobile, design, DevOps, AI integrations, and related areas.

Automatic CV/portfolio import from freelance platforms and unique AI-generated profile pages.

No time trackers or screenshot monitoring.

Fixed-price/milestones at the start, no hourly rates.

Minimize unpaid work: no multi-hour free estimates, pointless calls, or constantly monitoring a job feed.

No fake projects, bots, or artificial activity.

No talking to a wall: if one side disappears, the system/manager reacts.

Eventually: a forum/community where users can influence the development of the platform.

What i do now

I not building a huge platform yet.

I am looking for 100 potential clients who say: "Yes, I would try hiring this way."

I am looking for 100 experienced freelancers who say: "Yes, I want to receive projects this way."

After that, build a small MVP/private alpha.

The first real success metric is not registrations or upvotes. It's real projects that go through this new workflow and actually reach delivery.

If the model doesn't work, i shut down the experiment.

I'm looking for ideas and criticism.