field · idle
Scaled Minds · a small studio
Three quarters of companies have moved less than half their AI pilots into production. If yours is still a demo that impresses people in a meeting, we finish the gap. You get a system somebody uses every morning, in the tools you already have.
4h 30m → 35mPer day, rekeying
Halden FreightFreight brokerage · 40 people
Two dispatchers spent most of a morning copying rate confirmations out of emailed PDFs and into the dispatch system. We built something that reads the PDF, finds the matching load, writes it in, and holds anything that does not line up for a human. It has been running unattended since March.
3 days → same dayPrior authorisation packets
Corven HealthTwelve-clinic group
Assembling a prior authorisation packet meant a coordinator reading a chart and hunting for the four documents an insurer wanted. Now the packet assembles itself and flags what is missing before it goes out. First-pass denials dropped by a third, which was the number the practice manager actually cared about.
2 days → 3 hoursQuote turnaround
Ravensworth BrokersCommercial insurance
Brokers were comparing policy wordings from six carriers by opening six PDFs side by side. We built a comparison that pulls the clauses that differ and shows them next to each other, with a link back to the page it came from so nobody has to trust it blindly.
9h → 40mFirst response time
TallowDirect-to-consumer, 200k customers
Support was three people and a backlog. The system now drafts a reply with the order, the shipping status and the returns policy already pulled in. A person reads it and sends it. Seventy-one percent go out with no edit at all, and the three people still answer the hard ones.
We take one job a person is still doing by hand and turn it into software that does it, in the systems you already have.
You have something that works in a notebook or a demo account and has not moved in months. Usually the model is fine and the problem is everything around it. This is most of what we are hired for.
Not thisA second pilot to prove the first pilot.
Invoices checked against contracts, claims against policy, documents against a rulebook. Somebody opens two things, compares them and records an answer. This is the cheapest place to start and the easiest to measure.
Not thisA chatbot that answers what your intranet already answers.
Search that understands the question, drafting that sounds like your product, summaries people trust. Built into your codebase in your language, reviewed by your team, not bolted on the side.
Not thisA widget your customers close immediately.
Getting to the data, cleaning it, connecting to the systems that hold it, and keeping the whole thing running when a vendor's API changes without telling anyone. Unglamorous and usually where the schedule actually goes.
Not thisA dashboard that becomes somebody's third job.
Before anybody commits real money, we build the narrowest useful version against your real data and tell you what it costs to finish. Sometimes the answer is that you should not build it, and that answer is worth the two weeks.
Not thisA strategy deck as the first deliverable.
Somebody else started it and left. We read what is there, tell you honestly what is worth keeping, and finish it or replace it. We will say if the right answer is to throw it away.
Not thisPretending the last team's work was fine.
This is the Halden system doing its job. On the left is what it wrote down as it worked. On the right is the same thing in English. Watch what it does when it hits something it is not sure about, because that is the part that decides whether you can leave it alone.
In English
Starting up.
Software is driving
The system lands in your cloud account and your team's repository. You hold the keys. You keep it running without renting a seat from us.
A few hundred real examples from your own work with the answer a person would have given. It is how we proved it works and how you will prove it still works after somebody changes the instructions in eight months.
How it works and what it touches. Where it breaks, and what to do when it does. Written for the person who will be on call for it, which will not be us forever.
What the work took when a person did it and what it takes now. Measured on your data, not a benchmark. If the number does not move we would rather find that out in week two than month six.
Those thirty days are included. Live systems find edge cases that test sets do not, and the first month is when they show up.
Every system we build has a confidence line and a short list of things it may not do on its own. Below the line, or outside the list, it stops and hands the work to a named person with everything it already found attached. It does not guess just to keep work moving.
You can open what it read and which systems it called. You can see what it decided and why. If a customer or an auditor asks in eighteen months why something was approved, you open the record and answer.
New systems begin by recommending only, and a person approves every action. It gets more room when the record shows it has earned it, and that room comes back automatically if the numbers move. You set the ceiling and you can lower it any time.
We take one narrow job, build the smallest useful version against your real data, and come back with either a working thing or a written explanation of why you should not build it. Fixed price, fixed dates, and you keep whatever we make.
from $12,000Two weeks · one job
Six to ten weeks from the proof to something running on live work. Fixed scope agreed in writing before we start. If the scope changes we tell you what it costs before we do it, not after.
$35,000 – $90,000Six to ten weeks
Days per month for changes, new edge cases and the occasional vendor breaking something. Cancel with thirty days notice. Plenty of clients stop after the first quarter because their own team has it, which is the point.
from $6,000 / monthTwo days a month, minimum
Code starts before any large payment. The first thing you get is working software against your data. Every client on this page said yes to being named.
Name one job. The one everybody complains about in the same meeting every month.