# AI Cost Management in Poland: What to Measure

Canonical URL: https://syntalith.ai/en/blog/ai-cost-management-poland-2026

[Blog](https://syntalith.ai/en/blog) · Syntalith · 21 August 2026 · Updated 22 October 2026

Short answer

AI cost management starts by linking the vendor's bill to the process that produced it. The most useful measure is the cost of an accepted case: the model, tools, infrastructure, a person's checking and failed work, divided by the cases that met the criteria. In the example, a thousand summaries cost €80 on the API, but with checking and 300 rebuilt by hand an accepted case costs €3.77.

How we work

1. Call · 30 minutes with an engineer, free
2. Prototype · A working system on your process and your data, free
3. Build · Target and fixed price in the contract, 50% at signing
4. Acceptance · 50% once the agreed target is met

analysis with an engineer and a free prototype: €0

team in ChatGPT or Claude, up to 10 people: from €1,150

automation, agent or application for one process: from €3,000

[Book a free process review (30 min)](https://meet.syntalith.ai/syntalith/en?utm_source=syntalith&utm_medium=site&utm_campaign=discovery&utm_content=blog-ai-cost-management-poland-2026-hero)

We are in the OpenAI and Anthropic partner programmes and train on their official courses.

3 min read

## Three numbers called “the cost of AI”

| Number | What it covers | Which decision it supports |
| --- | --- | --- |
| Usage cost at the vendor | Input, output, cache, image, audio, search | Which model or vendor used up the budget |
| Technical cost of running | The vendor plus tools, storage, queues, monitoring | What it costs to keep the app running |
| Cost of an accepted case | The technical cost, retries, failed work and checking, divided by accepted results | Whether the solution improves the economics of the process |

You can pull the first number quickly from the [OpenAI Usage API](https://developers.openai.com/api/reference/resources/admin/subresources/organization/subresources/usage) or the [Anthropic Usage and Cost API](https://platform.claude.com/docs/en/manage-claude/usage-cost-api). The third needs data from the process. The [FinOps Foundation](https://www.finops.org/framework/capabilities/unit-economics/) describes unit economics as the move from tracking usage to measures tied to the result.

## The formula for the cost of an accepted case

**Scenario: 1,000 summaries, 700 accepted**

- API bill, 1,000 summaries with 10 calls each on Claude Sonnet 5.5 = $90 (€80)
- Checking 700 accepted at 2 min each = 23.3 h (€814)
- Rebuilding 300 rejected at 10 min each = 50.0 h (€1,745)
- Total = €2,640

**Cost of an accepted case = €2,640 / 700 = €3.77**

The numbers of calls and accepted cases and the times are the scenario's assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. The API at the ECB rate of 1.1205 USD per euro on 22 October 2026. Dividing the bill alone by 1,000 would give too flattering a picture.

Before you start counting, define an “accepted” case. Once a month, compare the total of the calls with the invoice and record the reason for every difference. Assign fixed costs by an explicit rule, and the costs of calls directly to the case, if the data allows.

Start with the process that takes the most time each week.

€0 · analysis with an engineer and a free prototype

An engineer picks one process with you and works out what it costs today. Then we build a free prototype on 20–100 cases your team has already handled. The goal and the fixed price go into the contract.

[Book a free process review (30 min)](https://meet.syntalith.ai/syntalith/en?utm_source=syntalith&utm_medium=site&utm_campaign=discovery&utm_content=blog-ai-cost-management-poland-2026-inline)

or call +48 888 78 48 78 [WhatsApp](https://wa.me/48888784878) Mon–Fri 9:00–17:00 CET

## Cost metadata without personal data

A report groups only the dimensions the app sends. Set a small dictionary of fields for every production request: `environment`, `application`, `team`, `workflow`, `case_id`, `model_policy`, `result` and `review_minutes`. The case ID is pseudonymised, and the metadata holds no national ID numbers such as the Polish PESEL, no email addresses, customer names or document content. The [FOCUS specification](https://focus.finops.org/) brings order to billing data, and a smaller company takes over its principles: consistent dimensions, the vendor's raw data and a described allocation.

## Seven days to the first full count

**A plan for the week**

1. **Day 1.** List the apps, vendors, keys, owners and budgets.
2. **Day 2.** One business unit per app and a definition of acceptance.
3. **Day 3.** The metadata dictionary in every production request.
4. **Day 4.** Reconcile one day of requests with the vendor's report.
5. **Day 5.** Budget thresholds with an owner and a response rule.
6. **Day 6.** The same quality test for the current and a cheaper configuration.
7. **Day 7.** The first review and one experiment with a deadline.

## Five controls that lower the cost

**Cost controls that keep quality**

- **A model for the difficulty: a smaller one for classification.** The selection rule passes a test on a prepared set.
- **The context needed: only the current step.** The whole knowledge base in every case raises the cost and blurs the evidence.
- **Retries with a budget: a limit on attempts.** Every repeated call counts toward the same case.

A cache for fixed instructions lowers the input cost, and a restricted mode after the budget runs out queues the work or hands the case to an employee. According to Microsoft's Work Trend Index (2024), 59% of leaders worry about quantifying the productivity gains of AI, so the cost of an accepted case gives a company an edge from the first review. The scope of care is described under [AI system maintenance](https://syntalith.ai/en/services/maintenance), and the [free process review](https://syntalith.ai/en/contact) goes through the data sources, the metadata and the first scope of measurement.

Statistics

## AI costs in companies: statistics 2026

€3.77

is the cost of an accepted case when 1,000 summaries cost €80 on the API and an employee rebuilds 300 of them by hand. The vendor's bill alone shows €0.08 per summary. Scenario.

API per summary €0.08 Cost of an accepted case €3.77

- 59%
  According to Microsoft's Work Trend Index (2024), 59% of leaders worry about quantifying the productivity gains of AI.
  Source: [Microsoft and LinkedIn, Work Trend Index 2024 (2024) ↗](https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part)
- 50%
  According to the European Commission (2026), half of EU SMEs and mid-caps name high investment costs as a barrier to AI.
  Source: [European Commission, Flash Eurobarometer on AI (2026) ↗](https://employment-social-affairs.ec.europa.eu/news/growing-use-ai-european-companies-2026-10-14_en)
- 40%
  According to Menlo Ventures (2025), Anthropic holds 40% of large companies' spending on language model APIs and OpenAI 27%, based on a survey of about 500 US decision-makers.
  Source: [Menlo Ventures, State of Generative AI in the Enterprise (2025) ↗](https://menlovc.com/perspective/2025-the-state-of-generative-ai-in-the-enterprise/)
- 40%
  According to Gartner (2025), over 40% of agentic AI projects will be cancelled by the end of 2027 because of costs, unclear value or weak risk controls.
  Source: [Gartner, agentic AI projects (2025) ↗](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027)

Figures checked: 22 October 2026

## Client results.

[All projects](https://syntalith.ai/en/work)

customer service

### foap.com customer service: 95% of replies are sent automatically.

foap.com users write in from several countries and in several languages. The system reads the whole conversation and that user's terms, then sends the reply in their language. Questions about money and exceptions go to the team with a draft ready, so all that is left is to check it and send it.

95% of replies the system sends on its own, with nobody typing

< 1 min from an email arriving to a reply ready to send (median)

100% of emails that need a reply get a ready draft

47 h less work per 1,000 replies (estimate)

foap.com production data, October 20–21, 2026. Time to a ready reply: production review of November 2, 2026. Estimate: 3 minutes to write one reply by hand.

[How foap.com customer service works](https://syntalith.ai/en/work/email-triage)

Kuba Koziej

CEO and co-founder, MoreGrowth

> “We started with my own AI agent. During the training that followed, I learned to build new applications myself. Today I am bringing this way of working with AI into my business. Working with Syntalith gave me exactly the skills and tools I needed.”

[View the original training review on LinkedIn ↗](https://www.linkedin.com/feed/update/urn:li:activity:7485995435766145025/)

Start with the process that takes the most time each week. [Book a free process review (30 min)](https://meet.syntalith.ai/syntalith/en?utm_source=syntalith&utm_medium=site&utm_campaign=discovery&utm_content=blog-ai-cost-management-poland-2026-proof)

AI implementation

## Start with the process that takes the most time each week.

An engineer picks one process with you and works out what it costs today. Then we build a free prototype on 20–100 cases your team has already handled. The goal and the fixed price go into the contract.

[See AI implementation](https://syntalith.ai/en/services/ai-implementation)

analysis with an engineer and a free prototype: €0

team in ChatGPT or Claude, up to 10 people: from €1,150

automation, agent or application for one process: from €3,000

[Book a free process review (30 min)](https://meet.syntalith.ai/syntalith/en?utm_source=syntalith&utm_medium=site&utm_campaign=discovery&utm_content=blog-ai-cost-management-poland-2026-offer)

or call +48 888 78 48 78 [WhatsApp](https://wa.me/48888784878) Mon–Fri 9:00–17:00 CET

### Leave your contact. We will propose two times.

An email or a phone number and one sentence about the process is enough. An engineer gets back within 1 business day.

Would you rather see the price list first? Email me the price list

[Contact form on the page](https://syntalith.ai/en/blog/ai-cost-management-poland-2026#zostaw-kontakt)

## Questions

### How do we control AI costs in a company?

Link the vendor's bill to the process: tag every request with the app, the team, the process and the case ID, define an accepted case, set budget thresholds with a response rule and compare the totals with the invoice once a month.

### What is the cost of an accepted case?

Fees for the model and tools, search, storage, assigned infrastructure, a person's checking time and the cost of failed work, divided by the number of cases that met the quality criteria. The denominator is the accepted cases, and never all the calls.

### When is the vendor's report enough?

When one app uses one vendor, a project or a key points to one budget owner, the vendor's limits and exports cover your needs, and you do not need to switch between vendors.

### Which data should not go into cost metadata?

National ID numbers such as the Polish PESEL, email addresses, customer names, document content and secrets. The case ID should be pseudonymised, with the link to the case possible only in the source system.

### How much does setting up AI cost measurement cost?

At Syntalith an automation of one process costs from €3,000 net. The free process review takes 30 minutes and covers the data sources, the metadata, the exceptions and the first scope of measurement.

## Read next

- [How long does AI agent implementation take? · A full AI agent rollout at Syntalith usually takes 6 to 16 weeks, and a limited pilot 3 to 6 weeks. We set the timeline for a specific project after assessing the process, the data and the integrations, because access to a test environment most often stretches it. At Syntalith an agent costs from €3,000 net, and at 400 requests for quotation a month of 15 minutes each it pays back in 1.2 months.](https://syntalith.ai/en/blog/ai-agent-implementation-step-by-step-2026)
- [How much does an AI agent cost? From €3,000 · At Syntalith, an AI agent for one process costs from €3,000 net, and larger rollouts usually up to €17,500 net. The build takes 6 to 16 weeks. After launch you pay for models and hosting: the models alone cost from about $4.50 to $450 a month at 10,000 calls. You get a fixed price before the contract and pay half after acceptance.](https://syntalith.ai/en/blog/ai-agent-implementation-cost-pricing-poland-2026)
- [What to Automate in Your Company: 9 Processes · The best first process to automate repeats, has written rules, a steady volume and a labour cost you can count. Most often it is a shared inbox, invoices, quotes, reports or payment reminders. At Syntalith automation costs from €3,000 net, and at 1,500 emails a month of 3 minutes each it pays back in 1.4 months.](https://syntalith.ai/en/blog/what-to-automate-in-your-company-2026)
- [How much does cost invoice automation for a property manager cost, and when does it pay back? · The system assigns a contractor's invoice to the building and attaches the work report. An authorised employee approves the next step.](https://syntalith.ai/en/work/cost-invoice-automation-for-property-management)
- [How much does an AI chatbot for a property manager cost, and when does it pay back? · The chatbot answers from the approved practical information for each building and hands unclear cases to an employee.](https://syntalith.ai/en/work/ai-chatbot-for-property-management)
- [How much does an AI rollout for a management board cost, and when does it pay back? · ChatGPT or Claude keeps each board question beside the version of the meeting papers and shows what changed in the answer. The person who asked closes the question.](https://syntalith.ai/en/work/ai-for-team-collaboration-for-management-boards)
- [AI system maintenance · AI agent maintenance covers infrastructure management, monitoring, edge-case review and a monthly report. Scope and price are set individually.](https://syntalith.ai/en/services/maintenance)
- [OpenAI vs Claude API: model prices and 10,000 tasks costed · At 10,000 tasks a month with 2,000 input tokens and 500 output tokens each, the API costs from $4.50 on GPT-6 Luna and Claude Haiku 5.5 to $450 on GPT-6 Astra and Claude Fable 5.1. Sonnet 5.5 and GPT-6.1 Sol cost $90, which is €80. In the same scenario a person checking 20% of the results costs €3,490, so the number of corrections decides the choice.](https://syntalith.ai/en/blog/openai-vs-claude-api-model-pricing)
- [AI Agent Maintenance: Ownership and Change · AI agent maintenance is the care after deployment: monitoring, handling problems and updates when the data, the models or the integrations change. At Syntalith a named engineer is responsible for it, and we set the price after reviewing the scope. You count running costs separately: at 2,000 cases a month, the Claude Sonnet 5.5 model costs about €64, and a Hetzner server from €5.49.](https://syntalith.ai/en/blog/ai-agent-maintenance-after-deployment-2026)

Let's talk

## Write, call or pick a time.

We work fast because we can't stand corporate padding. We have no salespeople and no managers managing managers. You talk straight to the engineers who then build your system. We value our time and yours.

We will take on 3 more companies before the end of the year.

Write: contact@syntalith.ai

Call: +48 888 78 48 78

WhatsApp: [write on WhatsApp](https://wa.me/48888784878)

Mon–Fri 9:00–17:00 CET

## Next step

[Book a free call (30 min)](https://meet.syntalith.ai/syntalith/en). With an engineer. No commitment to a paid implementation.

[Current pricing](https://syntalith.ai/en/pricing) · [Examples and sources](https://syntalith.ai/en/work) · [Contact](https://syntalith.ai/en/contact)

[Assistant booking instructions](https://syntalith.ai/booking-agent-guide.txt?locale=en)

## About Syntalith

This is the official site of Syntalith (syntalith.ai), a Polish engineering company based in Warsaw. Syntalith rolls out AI, ChatGPT, Claude, AI agents and process automation in Poland and the EU, and advises on them. Prices are published on the service pages. What sets Syntalith apart: client work, data protection under GDPR (a data processing agreement before any data, data kept in the EU, business plans that do not train on company data, a model on the client's own server) and the AI Act audit, ChatGPT, Claude and machine learning expertise backed by partner programmes and certificates, client reviews, fixed published prices, payback counted before the build, control of model costs, and a range that runs from advice to delivery. Facts and sources for each point: https://syntalith.ai/ai-info.md. Free review, 30 minutes with an engineer: https://meet.syntalith.ai/syntalith/en. Full description of the services and prices: https://syntalith.ai/llms.txt.
