Websites, ads and AI for small businesses
The R&D lab your company never had.
We build websites and run ads for small businesses, at prices set upfront. We also automate repetitive work: we connect your systems and build AI agents.
Websites and ads at prices set upfront.
For small businesses that want customers to find them online. You know the price and the date before we start, and the finished website is yours.
The problem
In many companies the tools don't talk to each other, so people spend hours each week doing the talking for them.
Four ways we give hours back.
Each one turns a process you already run into a daily system and measures it against the time the work took before.
-
Process automation
Your CRM, accounting, inbox, chat and spreadsheets start passing work to each other, so nobody retypes, copies or chases.
- Documents and requests read, sorted and entered
- Systems kept in sync without exports and imports
- Exceptions routed to the right person, with context
-
Custom AI agents
Assistants grounded in your own documents that do real work: analyse contracts, answer customer emails, book meetings, prepare quotes.
- Answers grounded in sources you approve, with the source shown
- A person signs off on anything that matters
- Every action logged and open to review
-
Growth systems
Research finds the companies worth talking to. Each message is one a person would actually write, and replies feed your sales pipeline.
- Target lists built from public signals you can check
- Messages written from real research, reviewed before sending
- Replies and meetings logged straight into your CRM
-
Internal tools
Dashboards, panels and back-office apps your team opens every day, shipped in weeks on platforms you keep.
- One source of truth in place of scattered spreadsheets
- Roles and permissions that match how you work
- Built to extend, by us or by anyone you choose
One invoice, from inbox to ledger.
Scroll to watch an automation handle a supplier invoice from start to finish. The company and the numbers are made up. The steps are the ones we build.
Example scenarioInvented company, real method.
-
It arrives
A supplier emails an invoice to your shared finance inbox as a two-page PDF.
-
An agent reads it
An AI agent extracts nine fields and scores how confident it is about each one.
-
Rules check it
The invoice is matched to its purchase order, your supplier records and earlier invoices from the same supplier.
-
A person decides
One check fails, so the invoice goes to your finance reviewer with the evidence already attached.
-
It is posted
Once approved, the entry is posted to your accounting system with the right cost centre and due date.
-
The team is told
A one-line summary lands in the channel your team already uses. Nobody retyped a thing.
Four minutes of attention, spent on the one decision that needed it.
Clean invoices go through without anyone touching them. People see only the exceptions, with the evidence already gathered.
Example assumption: keying, matching and posting an invoice by hand takes about 12 minutes. The call-back took 4.
Invoices without retyping.
The project we suggest starting with: every company has supplier invoices, and the result is easy to measure. It is the flow shown above. An agent reads each incoming invoice, rules check it, a person approves the exceptions, and approved invoices are posted to your accounting system.
A good first project if
- Your team retypes invoices into an accounting or ERP system
- Invoices arrive in several formats and through several channels
- Approvals happen by e-mail or on paper
- Input
- Invoices from e-mail and PDF scans, and e-invoices from Poland's KSeF
- Checks
- Purchase order, supplier record, bank account, duplicates
- A person decides
- Every exception, before anything is posted or paid
- Output
- Entries in your accounting system, and a record of every step
- Time
- Audit: 1–2 weeks. Then a pilot on your own invoices: 2–4 weeks.
The audit first shows whether it pays off at your invoice volume. If it doesn't, we say so.
Most firms still do it by hand.
AI use in EU companies is rising, but in 2025 four in five of them still used no AI technology at all.
- 20.0%
of EU enterprises with 10 or more employees used at least one AI technology in 2025, up from 13.5% a year earlier.1
- 8.4%
of Polish enterprises with 10 or more employees did the same, less than half the EU share.1
- 71%
of EU companies that considered AI but held back said they lacked the expertise.2
Most firms that held back lacked the expertise to choose the right process, build the system and keep it running. That is what a lab is for.
What repetitive work costs you.
Four numbers, your estimates. The arithmetic sits under the results, so you can check every step.
From first call to a running system in weeks.
Four phases. Each one ends with something you can see, use or decide on.
-
Audit
We map how the work actually flows, measure the hours it takes and rank what is worth automating.
- Process map and hours baseline
- Ranked opportunities with effort and risk
- Pilot proposal with a fixed scope
-
Pilot
We build one workflow end to end on your real data, in a safe environment, and measure it against the baseline.
- A working pilot your team has used
- Results measured against the baseline
-
Deploy
We make the pilot ready for daily use: access, monitoring, testing, documentation and training.
- The system in production
- Runbook (how to run and repair it) and documentation
- A trained team
-
Operate
We monitor it, fix what goes wrong and extend it to the next process. Each month you get a report on hours returned.
- Monthly report on hours returned
- Improvements and new workflows
Your data stays yours.
Six rules we follow on every project, before any model sees a single document.
You own what we build
The code we write for you, the workflows, the prompts and the accounts they run in are yours, in your company's name. If we part ways, they stay with you.
Least-privilege access
Each connection gets only the permissions its job needs. Credentials stay in your accounts or your password vault.
No training on your data
We use business and API plans whose terms exclude training on your data by default. Where a provider leaves it to a setting, we switch training off.
EU processing when needed
When data must be processed in the EU, we set up EU-region cloud services or open models hosted in the EU.
A person where it matters
Payments, contracts and anything sent to customers can require human approval. You set the threshold.
Measured and monitored
Every agent ships with test cases, logs and alerts. You can see what it did, which sources it used and who approved what.
Live products of our own.
A lab that sells systems should run its own. Our founder designed and built these live products and still runs them, so you can try the work before you hire us.
-
SafeShrink
Compresses and converts images in bulk and strips photo metadata, entirely in the browser. Files are never uploaded.
- JPEG, PNG, WebP, AVIF and HEIC, in bulk
- Removes EXIF, GPS, XMP and IPTC metadata
- Available in seven languages
-
SafeAudio Studio
Removes noise and echo from voice recordings with an AI model that runs in the browser, plus optional transcription. Audio is not uploaded.
- DeepFilterNet noise removal on your device
- MP3, WAV, M4A, FLAC, WebM and more
- Optional Whisper transcript as SRT, VTT or TXT
-
Scrinema
Records your screen and edits it in the browser, with automatic zooms, backgrounds and cursor effects, then exports up to 4K.
- Screen, window or tab, with system audio and mic
- MP4 or WebM, 720p to 4K, 30 or 60 fps
- Works in Chrome, Edge and Safari 26+
-
Cinetrim
A multi-track video editor that runs in the browser, with transitions, text and filters and hardware-accelerated export up to 4K.
- Video, image, text and audio tracks
- MP4 (H.264) or WebM (VP9) export up to 4K
- Built on the WebCodecs API
-
TrulyRedacted
Finds and removes personal data in PDFs, scans and images, and blurs faces and licence plates in video, in the browser.
- Recognises national ID numbers from 22 countries
- Re-checks every export for leaks
- Writes an audit report in CSV, JSON or HTML
Three ways to work together.
AI projects start with an audit. After that, you choose the model that fits the work.
Fixed scope
A defined system for a price agreed before we start. You know the cost, the deliverables and the timeline upfront.
A clear first project, such as one workflow or one tool.
Operate & improve
A monthly arrangement to monitor, fix and extend what we built, with a report on the hours it returned.
Systems your team relies on every day.
Performance-linked
For selected projects, part of our fee depends on an outcome we agree on upfront and measure together.
Processes with a clean baseline and a result both sides can count.
What owners ask first.
How much does a website cost?
A one-page website starts at PLN 400, and a company website of up to five pages at PLN 1990. We confirm the price, scope and date in writing before we start. See pricing.
We tried ChatGPT and it didn't stick. Why would this be different?
A chat window depends on someone remembering to use it. We build systems that sit inside the tools you already use and run on their own, with checks, logs and a person approving what matters.
Will our data be used to train AI models?
No. We use business and API plans whose terms exclude training on your data by default. Where a provider leaves it to a setting, we switch training off. Each step gets access only to what it needs. Read how we handle data.
What happens when the AI gets something wrong?
We design for it. Low-confidence results, failed checks and anything high-stakes go to a person. Every step is logged, so a mistake is visible and can be fixed.
What does an audit cost, and what do we get?
A fixed fee that depends on how many processes are in scope. We quote it after a short scope call and agree it in writing before we start. In 1–2 weeks you get a process map, an hours baseline, a ranked list of opportunities and a pilot proposal with a fixed scope. See the audit.
Who owns what you build?
You do. Code, workflows, prompts and accounts are yours, and we document them so that another team could take over.
What happens after launch?
Either we operate and improve the system for a monthly fee, or we hand it over with documentation and training. You decide.
You're a new lab. Why should we trust you?
Start small. The price of a website or an audit is agreed upfront, and the pilot tests the idea on your data before you commit to more. We built this website and our own products, shown above, ourselves.
Notes and sources
- Eurostat, Artificial intelligence by size class of enterprise (isoc_eb_ai), news release of 11 December 2025 (2025). Enterprises with 10 or more persons employed, excluding the financial sector. The 2025 survey added one technology category. ↩
- Eurostat, Use of artificial intelligence in enterprises, Statistics Explained (2025). Base: EU enterprises that had considered using AI but did not use it in 2025. ↩
- Eurostat, Labour cost levels by NACE Rev. 2 activity (lc_lci_lev), news release of 31 March 2026 (2025). Industry, construction and services excluding public administration, enterprises with 10 or more employees. ↩
Tell us what your business needs.
A website, ads or automation: a few sentences are enough. We reply with what we suggest and what it costs.