Skip to content

Getting started

What Do Companies Need for AI? The Key Requirements

To get started with AI, companies need neither their own data center nor data scientists. What matters are clear tasks, organized data, a suitable tool with sound contracts – and people who know what they are doing.

Lurus Editorial Team Published Reviewed 4 min read
Painted scene: two colleagues walk up a hill path at sunrise.

Key takeaways

  • Technically, a browser is usually enough to start: AI assistants run as cloud services.
  • More important than technology are tasks, data rules and a responsible person.
  • For personal data, you generally need a data processing agreement.
  • Employees need basic knowledge of capabilities, limits and rules.
Contents
  1. In short: less technology, more organization
  2. 1. Clear goals and tasks
  3. 2. Organized data and documents
  4. 3. A suitable tool
  5. 4. Contracts and data protection
  6. 5. Rules for everyday use
  7. 6. Skills in the team
  8. 7. A responsible person
  9. What you don’t need to get started

In short: less technology, more organization

When people think of AI, they often picture data centers, custom models and machine learning experts. For everyday work in small and mid-sized companies, that rarely applies. Modern AI assistants run in the browser, and the provider supplies the computing power. The real requirements are organizational: you need to know what you use AI for, which data may go into it and who is responsible.

The following seven points cover what you should settle before you start.

1. Clear goals and tasks

It all starts with the purpose. Which tasks should become faster, easier or better? Define two or three concrete use cases – such as “draft proposals from notes” or “write minutes from meeting notes”. Without a defined purpose, you can neither measure the benefit nor check whether the data processing is necessary. The German Data Protection Conference also names defining the purpose as the first step.

2. Organized data and documents

For many tasks, the AI only needs what you include in the request. But if it is to work with your company knowledge – handbooks, price lists, product data – the quality of those sources matters. Outdated or contradictory documents lead to outdated or contradictory answers.

So start by selecting a few reliable documents and decide who keeps them up to date. How AI works with your own documents is explained in RAG: using AI with company knowledge.

3. A suitable tool

The German Federal Office for Information Security (BSI) has published a criteria catalog for integrating external AI models. It is aimed at the federal administration, but many points are useful for companies too. When choosing a tool, pay particular attention to:

  • Data location and contract: Where are inputs processed and stored? Is there a data processing agreement?
  • Training: Are your inputs used to train the models? For company data, this should be excluded.
  • Access rights: Can you manage teams, roles and sharing?
  • Traceability: Are there logs of who used what, if you need them?
  • Model choice: Can you use different models for different tasks?

4. Contracts and data protection

As soon as personal data is processed – which happens quickly, for example with emails that contain names – the GDPR applies. You then generally need a data processing agreement under Article 28 GDPR, must identify a legal basis and may need to update your privacy notices. Involve your data protection officer early. Details are in Using AI in compliance with the GDPR.

5. Rules for everyday use

Employees need answers to simple questions: Which tool is approved? Which data may go in? When does a result need to be reviewed or labeled? An AI policy of one or two pages is enough. You’ll find a template in AI policy template.

6. Skills in the team

Article 4 of the EU AI Act requires companies to take measures to ensure the AI literacy of their staff. The European Commission makes clear that no specific format, test or certificate is required. Useful topics are the basics of how AI works, good prompts, typical errors and your own rules. Key terms are explained in the AI glossary.

7. A responsible person

Appoint someone to coordinate the rollout: answering questions, collecting experience, keeping the policy current and liaising with data protection and, where applicable, the works council. The BSI criteria catalog also recommends keeping an overview of all AI applications and who is responsible for each – in smaller companies, a simple spreadsheet will do.

What you don’t need to get started

  • Your own AI model: Training and running your own models only pays off in exceptional cases.
  • Perfect data: Start with tasks that work without company data and add sources step by step.
  • A big project: A pilot with a few people and clear tasks delivers insights faster than a comprehensive concept.

How to turn these requirements into a roadmap is shown in How to introduce AI in your company.

Frequently asked questions

Do we need our own servers for AI?

Usually not to get started. Most AI assistants are used as cloud services. What matters then is the server location, the data processing agreement and whether inputs are used for training.

Does our data have to be perfectly prepared?

No. Many tasks, such as drafting or summarizing text, need no company data at all. Only when the AI is meant to work with your documents is it worth selecting current, reliable sources.

Do we need an IT department?

Not necessarily. For a cloud service, one person who manages accounts and acts as a contact is often enough. External support can be useful for connecting other systems.

Sources

  1. Criteria catalog for integrating externally provided generative AI models (in German) German Federal Office for Information Security (BSI)
  2. Guidance on Artificial Intelligence and Data Protection, version 1.0, 6 May 2024 (in German) German Data Protection Conference (DSK)
  3. AI Literacy – Questions & Answers European Commission

Written by the Lurus Editorial Team according to our source standard. Editorial policy

Free initial consultation

Your AI start, planned together.

Show us your tasks. We’ll discuss data paths, rules and the right starting point for your team – no feature tour.

The initial consultation is free of charge. A phone number is not required.