June 2026 Edition
AI Types, Sovereignty and Deployment Choices
Welcome to the June Edition,
After the first newsletter introduced basic AI terms and why AI matters for SMEs, this edition takes the next step: understanding what kind of AI you are looking at and where it should run.
For SMEs, the practical question is rarely “Should we use AI?” in general. It is more often: “Which type of AI fits this task, which data does it need, and what level of control do we require?”
Why This Matters for SMEs — Right Now
The AI market is growing quickly, and many tools use similar marketing language. Chatbots, forecasting models, document tools, image systems and workflow assistants may all be called AI, but they solve different problems and have different requirements.
A simple classification helps companies compare options more systematically. It also supports better decisions about data protection, cost, integration and dependence on external providers.
A Simple Map: Different Types of AI
AI can be classified in different ways. No single map is perfect, but a few distinctions are useful for SMEs: what the system does, what data it works with, how much autonomy it has, and where it runs.
Four Useful Ways to Classify AI
- Predictive or analytical AI - looks for patterns in data and supports forecasts, risk scores or recommendations, for example demand planning, predictive maintenance or quality control.
- Generative AI - creates new content such as text, images, code or draft emails. It is useful for communication, documentation, marketing and brainstorming.
- Perception AI - processes images, speech or sensor signals, for example detecting defects on a production line or turning spoken language into text.
- Process and decision-support AI - helps structure workflows, classify cases, route requests or suggest next steps in customer service, administration or operations.
- Agentic AI - an emerging form of AI that can plan and carry out multi-step tasks within defined limits. For now, SMEs should treat it as a topic to watch and test carefully.
European AI Sovereignty: Why It Enters the SME Discussion
AI sovereignty does not mean that every company must build its own AI model. For SMEs, it means having real choices: trusted providers, control over sensitive data, reliable access to services, and the ability to comply with European rules and customer expectations.
Europe is strengthening its AI ecosystem through public infrastructure, research and innovation initiatives. For companies, the business question is practical: which solutions give us enough performance and convenience without creating unnecessary dependency or risk?
Local AI, On-Premise AI or Cloud AI?
There are now more deployment options than a few years ago. A company can use cloud-based AI services, run smaller models directly on a laptop or workstation, or operate AI systems on its own servers or through a trusted European provider.
Cloud AI is often the easiest starting point. It offers strong models, fast updates and low initial effort. It can be attractive for general tasks such as drafting, translation, summarisation or customer communication.
Local AI means running AI directly on a device, for example on a modern laptop. It may be useful for experimentation, privacy-sensitive drafts, offline work or smaller assistance tasks. The trade-off is that local models are usually more limited and require some technical setup.
On-premise or private-cloud AI offers more control over data, integration and security. It can be relevant when sensitive customer data, production data or internal knowledge should not leave the organisation. The trade-off is higher setup and maintenance effort.
How to Choose the Right Setup
A practical starting point is often:
- Start with the business task, not the tool.
- Identify the data involved: public, internal, confidential or regulated.
- Decide how much control is needed over data, access and outputs.
- Compare cloud, local and on-premise options based on value, risk and cost.
- Test with a small pilot before committing to a larger rollout.
- Define who reviews results and who is responsible for decisions.
Small pilots are usually more valuable than large, unfocused experiments.
What This Means in Practice
A cloud chatbot may be perfect for drafting marketing texts, but unsuitable for confidential contracts if the data handling is unclear. A local model may be useful for quick internal assistance, but not powerful enough for complex analysis. An on-premise system may be justified for core processes, but unnecessary for low-risk everyday tasks.
The best choice is therefore not “cloud versus local” in general. It depends on the task, the data and the business impact.
Looking Forward
In future editions, we will continue to translate AI developments into practical questions for SMEs: how to assess use cases, how to prepare data, how to evaluate tools, and how to move from experimentation to reliable adoption.
Our goal remains simple:
Helping SMEs understand AI realistically, practically, and with confidence.
Content provided by the Machine Learning Group at RPTU for the Boost AI Monthly Background Newsletter series, June 2026 edition.
Steffen Reithermann <steffen.reithermann@cs.rptu.de> - ml.cs.rptu.de
