Invoices, contracts, inbox, timesheets. We find one expensive, repetitive process in a company that runs on documents and mail, build a system that does it, run it on your hardware, and quote a fixed price. Below: what it actually looks like.
Left: the same check our engineers run on a first call. Right: a report from a running system; every line is a thing a person did not have to do that day.
We find the expensive repetitive work in a company and replace it with a system. Not everything at once: one process where the money leaks, proven, then the next. These four are where it leaks most in a company that runs on documents and mail.
Before · An accountant retypes 40 supplier invoices a day into the ledger
After · Fields extracted and checked against the PDF; one in twenty flagged for a person.
Before · People read a shared mailbox and forward requests by hand
After · Requests classified, filed to the case and routed; duplicates dropped.
Before · Consultants log hours from memory on Friday; some work is never billed
After · Work seen in mail and calendar is suggested to the person the same day.
Before · Reviewers read every listing, contract clause or expense claim
After · The model applies your written rules; only real edge cases reach people.
Most AI tools send your documents to someone else's cloud. For any company with client data - an accounting practice, a logistics operator, an insurer, an agency - that is often a non-starter. We build systems that run on your own servers, with open models, so the data never leaves the building. Cost is fixed: no per-token meter that grows with usage and surprises finance at the end of the month.
Honest note: when a client has no strict data requirements, a cloud API is often cheaper and faster to start with. In that case we say so and build on it.
The platform behind our cases handles 12B+ tokens and 4.5M+ requests a week at $0.084 per million tokens all-in, on our own fleet at 95% utilisation. Production numbers, not a forecast. Run the numbers for your own volume.
Moderation of millions of items a week without a room full of reviewers. Calls translated live on the client's own hardware. Years of documents searchable in seconds. Invoices read by a model instead of typed by hand. Each case comes with the problem, the system we built and the result.
All ten cases and the stack behind eachYou see the ROI math before you spend anything on development. These are reference points for a typical mid-size engagement, not a price list: a small pilot costs less, a heavy self-hosted deployment costs more. What never changes: the price is fixed in writing before we start, and there is no hourly meter.
| Step | Takes | Price | What we do and what you get |
|---|---|---|---|
| Audit | 2 weeks | €3,500 fixed | We sit with the people who do the work, map the process and measure where time and money go. You get a written picture: process map, top three automations, ROI per item. Fee credited if we build. |
| Pilot | 3-6 weeks | from €6,000 | One process, one success metric agreed in writing. A working system on your real data, checked against that metric before anyone decides on the next step. |
| Rollout | per scope | from €15,000 | The pilot becomes daily operations: integration with mail, documents and internal systems, access rules, training for the team, monitoring. |
| Care | ongoing | from €900 / month | Models change, documents change, people change. We keep the system running, update it, and stay responsible for how it behaves in production. |
If your system has an API, a database or a mailbox, we can read from it and write to it. If it has none, we work from the files and screens your people already use.
Microsoft 365, Exchange, SharePoint, Google Workspace, Jira, Confluence, Slack, Telegram, your document system, and anything else with an API.
Unrote is built by engineers who have spent years shipping AI systems in production. The founders stay hands-on in every audit and every build: no juniors, no hand-offs. Replies within one business day, in English, Russian, Serbian or Spanish.
Co-founder, delivery. Barcelona
Leads audits and delivery: from a mapped process to a working system with one agreed metric.
LinkedInCo-founder, AI and infrastructure. Belgrade
Builds and runs the GPU fleet behind our cases: 12B+ tokens a week, 99.9% uptime, self-hosted end to end.
LinkedInIt can stay entirely inside your perimeter: when needed, we deploy models on your own servers and nothing leaves. GDPR-compliant, EU hosting where required.
Chatbots answer; our systems do the work. One success metric is set up front, and you see it hit on real volume before you scale.
If two or more people do repetitive digital work, our target payback is months, not years.
Monitoring and alerts ship with every build; humans stay in the loop for edge cases; the care plan covers fixes, changes and updates.
The audit has a fixed price. Every build after it is quoted fixed too, see Four steps, one process at a time. No hourly billing, ever.
You upload them to an encrypted link under an NDA, or we run the test on your own server. They are deleted after the report, and the report is yours either way.
Invoices, contracts, forms, mail threads: whatever your team retypes or reads by hand. In five working days you get them back processed, with accuracy per field and a plain yes or no on automating it. Free, under your NDA, through an encrypted upload or on your own server.
Request a secure upload linkOn the call we point at 2-3 processes in your company worth automating and estimate the hours and euros you'd recover. You keep the numbers either way, and you talk to an engineer, not a sales rep.
An engineer's model answers in a minute. Honest, including no.