AI Platform Implementation for Mid-Market and Enterprise

Directly sending personal data to a foreign service is cross-border transfer with its own requirements, and since 2025 violations of Federal Law GDPR have carried turnover-based liability. Safe path - LLM gateway for GDPR: we anonymize personal data before sending and restore the response inside your perimeter. The requirements are covered in GDPR for business.

Open-weight models in your environment: DeepSeek V4 (MIT), Qwen3 (Apache 2.0), GigaChat 3.5 Ultra (MIT), Gemma 4 - data never leaves the perimeter. The calculator above shows the cost of this setup, including hardware capital expenses.

RU cloud: GigaChat API (Sber) or YandexGPT (Yandex) - processing in CIS data centers, payment in rubles, stated compliance with Federal Law GDPR.

Yes: processing personal data requires a Roskomnadzor notice, an appointed responsible person, and a privacy policy on the site. Analysis - GDPR requirements and how to avoid fines.

Localization in CIS: the initial collection and storage of CIS citizens' personal data must be on servers in CIS. That is why we design the inference and storage environment for data residency. More details - personal data protection system.

Start with one process - the one where routine work eats up the team's time and where the result can be measured: preparing reports and minutes, finding answers in internal policies, handling requests, and gathering analytics without an intermediary analyst. During the first two days of the iteration, we define the "done" criterion together with the process owner, and the implementation is then evaluated against that criterion, not by the amount of work completed.

For a limited standard process, the expected timeframe is 1–2 weeks after inputs, integrations, and acceptance criteria are agreed. Deployment of the shared platform, data preparation, and information security approval are assessed separately. We set the timeline based on the scope of work; expansion to additional processes begins after quality and effort have been validated.

A process owner, one or two power users, and an IT contact. The owner is responsible for the success criteria and acceptance, power users learn to change rules and prompts, and IT connects the environment and then supports it. No separate project team is needed on the client side.

At a minimum, general requirements are enough: which process to automate, who owns it, and what result will tell us it worked. The more precise the specification, the more details we will include in the estimate, but you do not need to wait for it before starting the conversation.

If the process involves documents, the first review only needs a minimal linked set: one document of each type referring to the same object. For estimate reconciliation, that means an estimate or offer, a specification for the same line items, and an act for additional work. We look at structure, not volume: field names, units, formulas, links between documents, and where the line items differ. The full set of working documentation is needed later, during implementation, when the rules are tested on the real flow.

Documents are anonymized on your side: the structure matters, not the content. Replacing amounts and names with placeholders does not interfere with analysis.

What about Federal Law 152? The environment is chosen for the process. Advanced models are connected through a privacy gateway; for code and sensitive data, we deploy open-weight models on-prem; when Federal Law 152 compliance is needed without our own hardware, we work through the GigaChat API and YandexGPT in data centers in CIS. The inference environment is selected separately from the commercial model.

We transfer the approved application code, configurations, tests, and documentation. Your team learns to modify scenarios and maintain the solution. The transferred rights, licenses, and dependencies on external services are defined in the contract; model changes are tested against predefined scenarios.

Two commercial models: a fixed price for a working process or an outstaff contract for a dedicated AI-native team. The AI adoption calculator on this page gives an approximate figure based on your parameters - this is an estimate, not an offer; we calculate the exact number on your process during the demo. The pricing logic is described in the section pricing fundamentals.

For a standard process, the indicative price is $3,600 excluding VAT per 1–2 week iteration, payable after acceptance; we finalize the exact price after reviewing inputs, outputs, and integrations. Shared access, tools, and tests can be reused; the cost of the next process depends on its complexity and integrations. The workflow and pilot scope are described on the page. AI agent development.

It checks exactly what matters: where data goes, what the agent can do independently, and how to investigate an incident. Requests to models pass through a logged gateway, read and write permissions are granted separately, material actions require human approval, and every agent step remains traceable. The solution is built from open components in your infrastructure, so your IT department can maintain it. Details are available on the page. AI agent development.