What if a ready-made tool is enough?

The review recommendation will identify it when it handles the process at lower cost. A custom build makes sense when the process or data is too specific for available solutions.

How do you keep the model from making things up?

We check the app on a fixed set of examples from your process, verify each answer against its source, and define the handoff to a person. Accuracy is measured before launch, and answers below the agreed threshold go to a reviewer.

What if the model gets it wrong in production?

For each answer, we record the question, sources, and decision, so a mistake can be traced and fixed at its source. Safety rules hold doubtful cases back from the user, while maintenance improves the rules and test examples after launch.

Where does our data live, and does it leave the company?

We deploy to an agreed cloud or your infrastructure, with minimum permissions and encryption. The app accesses the required data, while the code, system instructions, and test examples stay on your side. For personal data, we sign a GDPR data processing agreement and limit engineer access. The AI provider works under terms that exclude training on your data, or the system runs locally.

What about security and the EU AI Act?

We work on minimum permissions and with a decision log, and uncertain answers go through a human. We also provide technical input to the operator documentation the EU AI Act expects (accuracy, transparency, oversight), but we don't give legal advice or classify the system.

What if the app doesn't work in our process?

We set the quality threshold on your data before the build. If the review or test shows that accuracy is too low for your process, we tell you before paid development begins.

Who maintains the app after launch?

The same engineer who built it. After launch, maintenance covers monitoring, fixes, and updates to the safety rules and test examples as data and the process change. Scope and response time are set in the contract.

Is acceptance on my data included in the price?

Yes. Acceptance on your data is part of the fixed-price build. If the system does not reach the written accuracy threshold, we improve it within the same price, and you pay the second instalment only after acceptance.

When does a company knowledge base not need RAG?

For fewer than roughly 500 document pages, a simpler assistant over one file set is often enough. For larger collections, we combine semantic and keyword search and check permissions before releasing a document. The free process review identifies the suitable approach.

Custom AI applications. Answers from your documents, with sources.

We build an application for a specific process: searching documents, reading data or classifying cases. We check the quality on your examples, and the code and test kit are yours.

net per team app, fixed quote before signing: from €3,000
from contract to acceptance on your process and your data: 4–10 weeks

Book a free process review (30 min)

We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓

How we workDetails ↓

01Call30 minutes with an engineer, free
02PrototypeA working system on your process and data, free
03ImplementationTarget and fixed price in the contract; 50% at signing
04Acceptance50% once the agreed target is met

Why app

The answer is in the company's documents. Today, the team is searching for it manually.

The AI application is a tool for your team: a document search engine with a source, reading data from PDF or a panel for managing cases. It works on your data and only sees what a given employee can see.

answers from documents to which the questioner did not have access, in a 750-question test.

How a knowledge base with permissions works →

How it works

Your data. Your permissions. Your code.

Always answer with the source.
Each answer indicates a document and a fragment. When there is no basis in the documents, the application directs the question to the indicated person.
Permissions from your systems.
The employee will not receive a response from a file that he cannot open.
The code stays with you.
Upon receipt, you take over the code, documentation and test kit.

Verified on data

AI applications that already work.

All case studies →

Statistics: how much does it cost to search and submit data in 2026?

In the Syntalith test (2026), the application answered 750 questions without a single answer from a document to which the questioner did not have access. Half of the local model responses were ready faster than 521 ms.

There was not a single number in the Syntalith (2026) test reports without a source.

In the Syntalith test (2026), reading one request for quotation by AI cost €0.0034.

According to the Central Statistical Office, the average salary in the enterprise sector in September 2026 was PLN 9,266.48 gross.

With employer contributions and 168 hours a month, an hour of work costs the company PLN 66.45.

Sources: Central Statistical Office, Average employment and salary in the enterprise sector in September 2026. ↗ · Descriptions of Syntalith implementations

Data checked on October 25, 2026.

questions in the test, each with a different access card, with no answers outside the permissions: 750; questions in the test, each with a different access card, with no answers outside the permissions
comments with the number of unsourced comments retained before the publication of the report: 4/4; comments with the number of unsourced comments retained before the publication of the report

Scenario: cost of the same work in a month

Inquiries: 160 per month

Monthly closing in controlling

Respect. The number of questions, inquiries and hours are examples of assumptions from implementation descriptions. An hour of work costs PLN 66.45. We price the server cost for the local model separately.

How much does a custom AI app cost?

from €3,000 net

Full pricing →

An AI application for a team costs from €3,000 net, and implementation usually takes 4-10 weeks.

We determine the scope and price before the contract, we check the quality using your examples, and the code and test kit are yours. If a ready-made tool does it cheaper, we will recommend purchasing it.

What does the price depend on?

01The number of integrations and data sources
02Quality requirements and the scope of testing
03Security, consent, and action history

„Why should I trust you?”

Because the system is built by engineers with over 22 years of experience.

You talk to people who then write code. They have worked in large companies and built systems that operate under high traffic.

You won't spend money on a prototype that won't do anything.

First a free conversation, then a free prototype in your trial. If we don't see a return, we don't build and write it straight. You only pay for a system that makes sense.

Polish-Ukrainian team of engineers.

Artem, Przemek and Piotr studied computer science together at the Lublin University of Technology in 2017⁠–⁠2021. Artem and Serhii founded Syntalith together. They both come from Kiev and have been living in Poland for years. We build systems for Polish companies, in Polish and according to Polish law.

What our OpenAI and Anthropic partnerships mean for you.

OpenAI makes ChatGPT and Anthropic makes Claude. We belong to the partnership programs of both companies, so we work with their tools every day and learn from the creators themselves.

Claude Partner Network

01We will select a tool for the task: We implement both ChatGPT and Claude. We check both on your documents, and you choose the one that makes fewer corrections and saves more time.
02Knowledge straight from the creators: We learn from official OpenAI and Anthropic courses and materials. Your team learns about the features and settings that work today.
03Licenses without our margin: When we roll out ChatGPT, Claude or Copilot, you buy the licences directly from OpenAI, Anthropic or Microsoft and they invoice you. We add nothing to the licence.
04Certified engineers: On the team, we have an AWS certification in machine learning and an Anthropic certification in systems architecture on Claude.

Participation in programs does not mean that OpenAI or Anthropic endorses our services.

How we build automation and agents

From a conversation to a working system in four steps.

01 · 30 min
Free process review
€0
You're talking to an engineer. You receive the written plan within 2 business days.
02 · 5 days
Free prototype
€0
The system works on 20-100 cases that your team has already handled. You see every result.
03 · 2–16 weeks
Build
50%
Fixed price and goal written in the contract. First installment upon signing.
04 · after reaching the goal
Reception
50%
You pay the second installment when the system reaches its target on your cases.

How a free prototype is created

Mon.

We set a goal

What result do you consider good?

Tue–Wed

The system is working

It goes through your affairs, one by one.

Thu

We compare

We check every difference with the team's work.

Fri

Result and price

List of errors and fixed implementation price.

You choose the matters. We anonymize them upon request, they stay in the EU and we delete them after the test.

On the market, such a test usually costs from PLN 3,000 to PLN 8,000.

Book a free process review (30 min)

We establish a prototype during a free analysis.

Data checked on October 25, 2026.

Let's talk

Just write, call or choose a date.

We work fast because we hate corporate cotton wool. We don't have salespeople or managers to keep an eye on managers. You talk immediately to engineers who then build your system. We respect each other's time and yours too.

We will accept 3 more companies by the end of the year.

Write: contact@syntalith.ai
Call: +48 888 78 48 78
WhatsApp: write on WhatsApp

Mon–Fri 9:00–17:00 CET

Common questions

What if a ready-made tool is enough?
The review recommendation will identify it when it handles the process at lower cost. A custom build makes sense when the process or data is too specific for available solutions.
How do you keep the model from making things up?
We check the app on a fixed set of examples from your process, verify each answer against its source, and define the handoff to a person. Accuracy is measured before launch, and answers below the agreed threshold go to a reviewer.
What if the model gets it wrong in production?
For each answer, we record the question, sources, and decision, so a mistake can be traced and fixed at its source. Safety rules hold doubtful cases back from the user, while maintenance improves the rules and test examples after launch.
Where does our data live, and does it leave the company?
We deploy to an agreed cloud or your infrastructure, with minimum permissions and encryption. The app accesses the required data, while the code, system instructions, and test examples stay on your side. For personal data, we sign a GDPR data processing agreement and limit engineer access. The AI provider works under terms that exclude training on your data, or the system runs locally.
What about security and the EU AI Act?
We work on minimum permissions and with a decision log, and uncertain answers go through a human. We also provide technical input to the operator documentation the EU AI Act expects (accuracy, transparency, oversight), but we don't give legal advice or classify the system.
What if the app doesn't work in our process?
We set the quality threshold on your data before the build. If the review or test shows that accuracy is too low for your process, we tell you before paid development begins.
Who maintains the app after launch?
The same engineer who built it. After launch, maintenance covers monitoring, fixes, and updates to the safety rules and test examples as data and the process change. Scope and response time are set in the contract.
How much does a custom AI app cost?
Implementations start at €3,000 net. Price follows the integrations, data sources, and quality requirements. The proposal gives you a fixed quote before signing.
Is acceptance on my data included in the price?
Yes. Acceptance on your data is part of the fixed-price build. If the system does not reach the written accuracy threshold, we improve it within the same price, and you pay the second instalment only after acceptance.
When does a company knowledge base not need RAG?
For fewer than roughly 500 document pages, a simpler assistant over one file set is often enough. For larger collections, we combine semantic and keyword search and check permissions before releasing a document. The free process review identifies the suitable approach.