Question 1

How does agentic automation differ from RPA?

Accepted Answer

RPA works well in stable, repeatable processes based on clearly defined steps. Agentic automation is the right choice where the process requires context analysis, working with data from multiple sources, exception handling, and handing decisions over to a human.

Question 2

Does UiPath Maestro replace UiPath Studio?

Accepted Answer

No. UiPath Studio still serves to build automations and bots. UiPath Maestro is the orchestration layer that lets you combine robots, AI agents, business systems and a human within a single process.

Question 3

Are AI agents safe for critical data?

Accepted Answer

They can be, provided they are implemented with the right security architecture: access control, restricted data sources, action logging, human-in-the-loop, monitoring, and a security review before production launch.

Question 4

How much does UiPath Maestro cost?

Accepted Answer

Cost depends on the scale of deployment, the number of processes, the scope of integrations, the use of AI agents, the existing UiPath environment, and the licensing model. SNOK helps analyse the options and match the scope of rollout to a real business case.

Question 5

How does Case Management differ from a classic workflow?

Accepted Answer

A classic workflow works well where the process follows a single, predictable path. Case Management is the right choice for cases whose flow depends on data, decisions, exceptions, risk and human involvement. It allows many variants of a case to be run while maintaining control, an action history, and reporting capability.

Question 6

What kinds of cases suit Case Management?

Accepted Answer

Most often these are complaints, applications, KYC and AML proceedings, service cases, disputes, requests requiring verification, and processes where timeliness, compliance with organisational policy, and documentation of how a case was resolved all matter.

Question 7

Does Case Management require a full UiPath platform rollout?

Accepted Answer

Scope depends on the existing environment, the number of case types, the integrations needed and the level of automation required. Case Management can use UiPath Maestro, Action Center, RPA robots, AI agents and integrations with enterprise systems. SNOK helps choose the technology and licensing scope to match a real business case.

Question 8

How does Case Management support compliance and audit?

Accepted Answer

Case Management brings order to the way cases are handled and enables key information to be recorded: who carried out an action, when, on what data, and with what outcome. This process history supports internal control, reporting, incident analysis, and preparing materials for audit.

Question 9

How does Enterprise AI differ from ChatGPT?

Accepted Answer

Enterprise AI works on the organisation’s own knowledge: contracts, procedures, policies, reports, decision history and business data. Unlike a general-purpose chatbot, it has access control, source citation, an audit trail, integrations with company systems, and mechanisms supporting compliance.

Question 10

Does organisational data go to OpenAI or Anthropic?

Accepted Answer

It depends on the architecture. Several deployment models are possible: Azure OpenAI or Anthropic Claude via controlled API, private cloud, the client’s own environment, or local language models such as Llama or Mistral for the most sensitive data. In each case we analyse the requirements for security, compliance, running cost and data control.

Question 11

How do you protect against prompt injection?

Accepted Answer

We apply an AI Security Review covering source validation, source-aware access control, prompt-injection testing, sandboxing, limiting the tools available to the model, usage logging, and escalation scenarios to a human. For critical data we recommend additional red-team testing and a review aligned with the OWASP LLM Top 10.

Question 12

How long does an Enterprise AI implementation take?

Accepted Answer

An MVP for one department, for example a legal or finance assistant, usually fits within a 4-8 week horizon. A full organisational rollout, covering multiple data sources, integrations, compliance and broader user groups, typically takes 4-6 months.

Question 13

How does BPA differ from a single RPA bot?

Accepted Answer

BPA covers the entire business process - from input data, through decisions, integrations and workflow, to the final outcome. A single RPA bot usually automates one task or a fragment of a process. In BPA we combine RPA, AI agents, system integrations and human involvement wherever the process requires control, a decision or an approval.

Question 14

How long does a typical BPA project take?

Accepted Answer

We usually deliver the first MVP for the selected process within 6-10 weeks. Scaling to further processes depends on their complexity, the number of exceptions, data quality and the organisation’s readiness for change. Full development of the automation platform is most often carried out in stages.

Question 15

How do we measure ROI from automation?

Accepted Answer

We measure ROI using specific indicators: process cycle time, the cost of handling a single transaction or document, the number of errors, the level of automation achieved, and the time recovered by the team. We establish the baseline before implementation, based on an analysis of the current process.

Question 16

Does BPA require UiPath?

Accepted Answer

Not always. UiPath is our preferred automation ecosystem, but the choice of technology depends on the client’s environment, existing infrastructure and project goal. We also deliver projects using other tools, API integrations, iPaaS and hybrid architecture.

Question 17

How does Document Understanding differ from classic OCR?

Accepted Answer

OCR recognises text on a document. Document Understanding also recognises the structure and meaning of the data: supplier, buyer, net amount, VAT rate, order number, invoice line items, dates, clauses or form fields. In addition, the solution can validate data against business rules and pass it on to the ERP, workflow or source system.

Question 18

What accuracy is realistic?

Accepted Answer

Accuracy depends on the document type, scan quality, form layout, number of variants, languages and validation rules. High extraction accuracy is achievable for typical documents, but it is always worth measuring per field rather than only at the whole-document level. This is why, before going to production, we define quality KPIs: accuracy for specific fields, automation level, exception rate and the scope of manual validation.

Question 19

Will Document Understanding handle low-quality scans?

Accepted Answer

Yes, but the output quality depends on the input quality. We can apply preprocessing: contrast improvement, deskewing, binarisation, noise removal and other image-enhancement techniques. For very poor scans we also recommend improving the scanning process at source, since even the best model cannot solve every problem caused by illegible documents.

Question 20

Will SNOK get us ready for KSeF?

Accepted Answer

Yes. We help design the KSeF integration architecture: receiving, issuing, validating, mapping data, and passing invoices to SAP or another finance and accounting system. We also advise on the resulting process changes in accounting: exception handling, user roles, approvals, status reporting and compliance control.

Question 21

How does Agentic Testing differ from Selenium or Playwright?

Accepted Answer

Selenium and Playwright execute scripts prepared by testers or developers. Agentic Testing extends classic automation with AI mechanisms that can support application analysis, scenario generation, change detection, regression prioritisation and results reporting. Classic tools still have a place, particularly for stable, repeatable regression.

Question 22

Does UiPath Test Cloud require UiPath as a foundation?

Accepted Answer

UiPath Test Cloud is a UiPath product and naturally works best within the UiPath ecosystem. It can, however, integrate with other testing tools, DevOps systems, Jira, Azure DevOps, and automation based on Selenium or Playwright.

Question 23

How does Agentic Testing help with S/4HANA conversion?

Accepted Answer

S/4HANA conversion changes processes, transactions, integrations, reports and data. Agentic Testing helps build regression based on critical processes, automate repeatable scenarios, and detect defects faster before go-live.

Question 24

Does SNOK carry out UAT?

Accepted Answer

UAT by design remains with the client, since it is the organisation that accepts the solution from a business perspective. SNOK can prepare UAT scenarios, run test sessions, document results, support defect management, and prepare a Test Completion Report.

Question 25

How does AI Security differ from a classic penetration test?

Accepted Answer

A classic pentest focuses on the application, API, front end, infrastructure and security configuration. AI Security additionally examines the model and AI orchestration layer: prompt injection, jailbreaking, retrieval poisoning, access to sources, tool use, MCP security and data-exfiltration risk. It does not replace a classic pentest - it complements it for systems built on LLMs and AI agents.

Question 26

Does the AI model provider protect the organisation against these risks?

Accepted Answer

Partly. Model providers apply their own safeguards and guardrails, but a large share of the risk arises at the application layer: in data access, retrieval, tools, integrations, permissions and process logic. It is this layer that the organisation needs to design, test and monitor.

Question 27

What is the lethal trifecta?

Accepted Answer

The lethal trifecta is a scenario in which an AI agent simultaneously has access to private data, contact with untrusted external content, and the ability to communicate or act externally. This combination increases the risk of data exfiltration through prompt injection. In practice it means the agent could read sensitive data, be manipulated by external content, and use an available tool to disclose information.

Question 28

Is an AI Security Review mandatory in SNOK projects?

Accepted Answer

For agentic automation projects and AI solutions with access to data, tools or systems, we recommend an AI Security Review as a standard element of the deployment. In production projects we treat a positive review result as a condition for the safe launch of a solution, or as the basis for a decision on additional remediation.

Question 29

Will a local model be as good as Claude or GPT?

Accepted Answer

For many enterprise use cases, local models can be sufficient, particularly when they are well chosen, run on suitable infrastructure, and connected to the organisation’s knowledge through RAG or fine-tuning. For the most demanding tasks, cloud models may still have an edge, but the quality gap keeps narrowing across many scenarios.

Question 30

How much GPU capacity do we need?

Accepted Answer

This depends on the model, expected throughput, latency and deployment mode. Llama 3.3 8B can run on a single A100 or L40 class GPU. Llama 3.3 70B in 4-bit quantisation may need 1x H100 80GB or 2x A100. A full 70B FP16 model may require a cluster of 4x H100. SNOK carries out sizing before any infrastructure purchase.

Question 31

Is on-premise cheaper than the cloud?

Accepted Answer

At low scale, LLM on-premise will usually be more expensive because of infrastructure CAPEX. At high scale, millions of tokens a day, and over a long usage horizon, it can be more cost-effective. An additional benefit is control over data and compliance with requirements that public AI services may not meet.

Question 32

Does SNOK also support closed models on-premise?

Accepted Answer

Some providers, including Anthropic and OpenAI, do not offer their models for classic on-premise deployment. Open-weight models such as Llama, Mistral, Qwen, DeepSeek and Phi are available instead. Scenarios with Azure OpenAI in a private Azure tenant are also possible, if the organisation’s security model allows it.

Question 33

When does the AI Act come into effect?

Accepted Answer

The AI Act entered into force in 2024, but its provisions apply in stages. Some obligations already apply, and further requirements come into effect at subsequent dates. In practice, the first step is not preparing full documentation for every system, but rather inventory, risk classification and a prioritised action plan.

Question 34

Does the AI Act apply to my organisation?

Accepted Answer

If the organisation supplies, deploys or uses AI systems, it most often does. The scope of obligations depends on the organisation’s role, the system’s risk category, the purpose of the solution, and the data the system works on. Most business use cases will require at least classification and basic documentation.

Question 35

What penalties apply for non-compliance?

Accepted Answer

The AI Act provides for substantial penalties for the most serious breaches, including the use of prohibited practices. The final amount depends on the type of breach, its scale, the severity of the case, the size of the organisation, and the role the organisation plays with respect to the given AI system.

Question 36

Does SNOK provide a binding legal interpretation?

Accepted Answer

No. SNOK provides classification, risk assessment, an AI register and technical-organisational documentation. Interpretations with binding legal effect require confirmation by a legal adviser. We work with the client’s legal team on this.

Less manual work, faster decisions, lower operating costs

Key areas of support

AI agents and agentic automation

UiPath Maestro

Case Management

Enterprise AI

Process and document automation

Business Process Automation

Document Understanding

Agentic Testing

AI trust and compliance

AI Security and the AI Trust Layer

On-premise LLM

AI Act compliance

An approach that works for us

A healthcare operator active in 21 countries

Get in touch