The brief is a headline, not a problem.
“Add AI to our business” isn't a scope. Turning it into one — which documents, what accuracy bar, what happens on failure — is real work. On gig boards you do it free, in the proposal.
For LLM & data engineers
Founding members wanted
Extraction · classification · RAG over business documents · reporting copilots
LLM engineers: clients who've already paid to know what they need.
You put AI where it's better and keep humans where it's worse. Our clients already bought that idea. Each one paid $1,500 for an assessment and got a written report — ProcessScout calls it a Trail Map — signed by a human analyst. By the time you meet them, the argument is over.
Free to join · Every application read by a human · Answer in 3–5 days
The problem
LLM work is bought on hype and judged on demos.
On the usual marketplaces, careful engineers lose. Four ways:
Demos beat production.
A slick prototype wins the bid, then falls over on real invoices. The buyer can't tell the difference until month two.
Honesty loses bids.
“A human should stay in this loop” is the right answer. Next to someone promising full autonomy by Friday, it sounds weak.
No baseline, no proof.
Without a measured before, your after is an anecdote. Accuracy against what? Hours saved compared to what?
The work
What work we'd send you.
Our clients are small businesses — 10 to 49 people, heavy on documents, email and reporting. Example: the report says four people spend 22 hours a week typing in supplier invoices. Budget $14–20k. That job comes to you.
Swipe to see more →
| What the report found | The job you'd get | What you'd know before quoting |
|---|---|---|
| Invoices and supplier documents typed in by hand | Build document extraction with a human review queue | Document volumes and formats listed, the typing hours counted |
| Month-end reports built by copy-paste | Automate the reporting — every figure traceable to its source | The report chain mapped, the wasted hours costed in dollars |
| A shared inbox where orders and complaints pile up | Build email triage that routes work to the right person | Message volumes and handling times, measured |
| Company knowledge buried in documents nobody searches | Build search (RAG) over their documents, scoped to real questions | The documents named, the questions the team actually asks |
Examples, not live jobs. What we send you depends on what client reports find.
AI where it's better. Humans where it's worse.
That's how the client's own assessment worked: an AI assistant gathered the evidence, and a human analyst made the judgement calls. So when your design puts a review queue behind the extraction step, you're not apologising for a limitation. You're matching the method they already paid $1,500 for.
Before you meet the client
You get a measured “before” — so your results mean something.
Every client paid $1,500 for an assessment first. The signed report is the “before” your system gets judged against — in the open, on agreed numbers.
- You know the data quality first. Every claim is marked high, medium or low confidence — before you promise an accuracy number against it.
- The cost is in dollars. The value of your build is clear from the first call.
- There's a written skip list — including the places the analyst said an LLM is the wrong tool. You inherit that honesty instead of fighting for it.
- There's a 90-day plan with an owner. Your system ships into a plan, not a vacuum.
- Results get checked at 30 and 90 days. Measured results become proof you can show the next client — no NDA problem.
Said plainly
No promised amount of work. We introduce you when a client matches your skills — and we only earn when you do.
Pricing
What it costs.
Four numbers. That's the whole price list.
$0
To join. No listing fees, no bid credits, no subscription. A human reads your application in 3–5 days.
$0/mo
For the software. ProcessScout — normally $390/month. Free while you're a member using it with clients.
10%
On clients we send you. Drops to 5% after $25,000 with the same client.
0%
On clients you bring. Always. Even on the free software.
Example: an $18,000 project from us → you keep $16,200 · plus $300 for each business you refer to the $1,500 assessment · payment protection on some jobs Not on every job yet
Launch pricing. Your exact terms are fixed in writing before your first project.
Questions LLM & data engineers ask
Frequently asked
Are these clients ready for LLM work, or chasing hype?
Readier than most. They paid $1,500 for a diagnosis, not a demo — and their report includes a list of things not to build yet. A client who has accepted that is a very different buyer from one who wants a chatbot by Friday.
Do I need a particular model or stack?
No. The report names the problem, the documents and the measured baseline. Architecture, models and tools are your call, agreed with the client. We review how you handled testing, failure and human review in past work — not your vendor logos.
What do you charge?
10% on clients we send you, dropping to 5% after $25,000 with the same client. Clients you find yourself: 0%, always. Joining and the software are free. We don't promise a volume of work.
Founding members
Bring the engineering. The argument is already won.
Join free. Free software for your own clients. LLM and data work with a measured baseline. A human reads your application and answers in 3–5 days.