Articles grouped by business problem, industry, and system type.
Claude Duet is a new Anthropic app in which Claude works as a proactive agent. It sends a morning brief, watches your email and calendar, runs tasks on its own cloud computer and asks for approval before it sends a message or makes a payment. Since 21 October 2026 it has been a waitlisted research preview for Pro and Max subscribers in the US only.
Syntalith fits small and medium-sized Polish companies best when they want their first AI process in production at a fixed price. For large programmes in corporations, deepsense.ai and Addepto are the better choice. Vstorm leads in building agents, 10Clouds in banking and insurance, VM.PL in manufacturing, Protopia on Microsoft Azure and allclouds.pl in the public sector.
A2A, or Agent2Agent, is an open protocol through which AI agents from different companies find each other, describe their skills and hand each other tasks. Google announced it in April 2025, the Linux Foundation took it over in June 2025, and the stable version 1.0 came out on 12 March 2026. MCP connects an agent to tools. A2A connects an agent to another agent.
You prepare a store for AI agents when an agent finds in your data everything it needs to compare: the variant, the price with delivery, availability and returns. Add access for ChatGPT's bots to the site and measurement of traffic from chats. The automation costs from €3,000 net at Syntalith and, with 400 product cards a month at 6 minutes each, pays back in 3.1 months.
A model that names an account and an amount has prepared a proposed entry. An entry in the books needs a check of the source document, the agreed rules and control of the whole workflow, which Articles 20, 21 and 24 of Poland's Accounting Act describe. The accountant approves the proposal. At Syntalith the automation costs from €3,000 net, and with 1,000 cost invoices a month at 3 minutes each it pays back in 2.3 months.
AI in HR reads working time information from managers' messages, flags gaps such as a date or an employee ID, and prepares a proposed correction of the time records. HR approves the correction, and the history keeps the original entry. At Syntalith the automation costs from €3,000 net, and with 600 time record corrections a month at 6 minutes each it pays back in 2.1 months.
Before you buy booking automation, walk through a change to a saved appointment: the new slot may be taken, the old reminder may already be on its way, and 02:00 occurs twice when the clocks change. In our test of 11 October 2026, 42 tests in 4 suites checked these cases with no AI model and no external calendar.
MCP connects an agent to tools and data, A2A to other agents, UCP and ACP describe shopping in a store, and AP2 secures a payment an agent makes on your behalf. All are open standards at different stages. We build an MCP server with your catalogue and availability from €3,000 net, and a readiness audit costs from €600 net.
AI in industry starts safely with data that only informs people, for example a downtime report built from a copy of an export. A system that writes a machine parameter works on the OT side and needs a different scope, responsibility and safeguards. At Syntalith the automation costs from €3,000 net, and with 40 downtime reports a month at 120 minutes each it pays back in 1.5 months.
A person starting a conversation with a chatbot has to know they are dealing with AI, at the latest at the first interaction. The transparency obligations of Article 50 of the AI Act apply and are enforced from 2 August 2026. The notice has to be visible everywhere the chatbot opens, and a handover of the conversation to an employee has to be clear.
Automated reading of a CMR consignment note copies the fields. It does not establish whether the delivery was made without reservations. An illegible note in the reservations box should be marked for checking, and the system must never record it as „no reservations”. „File read” must never mean „delivery confirmed”. At Syntalith the automation costs from €3,000 net, and with 1,000 documents a month it pays back in 2.0 months.
AI can collect a counterparty's data from agreed sources and prepare a card for review, and checking the data does not by itself fulfil AML duties. Those apply to obliged entities, and a person makes the risk assessment and the decision. At Syntalith the automation costs from €3,000 net, and with 300 counterparty checks a month at 15 minutes each it pays back in 1.8 months.
AI in customer service first identifies the type of case: withdrawal from the contract, a complaint, or a cancellation on the terms of the offer. The word „refund” is not enough. The deadlines and the reply depend on the type, and the first test runs with no automatic refusals. At Syntalith the automation costs from €3,000 net, and with 800 requests a month at 6 minutes each it pays back in 1.6 months.
AI helps accountants prepare a draft list of questions to a client about documents: a missing purchase description, an ambiguous case number or an attachment mentioned in correspondence. The accountant sets the rules for the check, and the model orders the material. It decides neither the accounting treatment nor the tax treatment.
AI on a farm is best started with office work: reading delivery notes from photos, matching them to the supplier and the delivery, and flagging missing attachments. Choosing plant protection products and the conditions of subsidies are different tasks. At Syntalith the automation costs from €3,000 net, and with 400 delivery documents a month at 5 minutes each it pays back in 3.7 months.
AI in a car workshop orders a client's request before it reaches the service adviser: the description of the symptom, the conditions the client gives and the information still to collect. It does not diagnose the fault or judge whether the vehicle is safe to drive. The workshop sets the rules for requests that concern safety.
AI in construction prepares a working list of changes in the documents: what changed, which document and revision the item concerns, and who should check it. The designer decides what a change means, and drawings are best compared in the design tool. At Syntalith team training costs from €1,150 net a day. With 10 people and 1 hour saved a week each, it pays back in 0.9 months after licence costs.
Start AI in a hotel with the enquiries that do not fit a template: groups, meeting rooms, packages, several dates at once. Templates in your current sales channel often handle the arrival instructions. At Syntalith, automating reply drafts costs from €3,000 net. At 120 enquiries a month of 25 minutes each it pays back in 2.0 months, and reception still confirms prices and availability.
AI in logistics helps you prepare the handover of open cases to the next shift. Events from the system, messages and the operator's notes go into a short description: what is confirmed, what we are waiting for and who takes the next step. Start with one kind of exception, for example deliveries waiting for a confirmed time.
AI in HR prepares a list of gaps and mismatches in the data before payroll is run. Calculating pay and approving the data stay in the HR and payroll software and with people. At Syntalith the system costs from €3,000 net. At 1,000 documents a month, handling them by hand costs €2,908, and with the system €442, so the payback takes 1.2 months.
AI agents
The agent moves a case through its steps, and actions with consequences wait for approval from the person you name. Before you sign, we write down the goal and a test set built from your cases.
Times are shown in your own time zone. We work with clients across time zones.
See AI agentsOpenAI 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 NetworkParticipation in programs does not mean that OpenAI or Anthropic endorses our services.