---
title: "AI in mid-sized companies: start pragmatically instead of planning big"
description: "How mid-sized companies can start with AI pragmatically: one bottleneck, one module, measurable benefit – plus criteria for GDPR, control and costs."
type: "wissen"
product: "webRichtung"
slug: "ki-im-mittelstand-einstieg"
source_language: "de"
target_languages: ["de", "en", "es", "pl", "tr"]
published: "2026-06-10"
status: "publish"
faq_json: [{"q":"What is the best way for a mid-sized company to get started with AI?","a":"Pragmatically: choose a concrete bottleneck – such as missed calls, document search or text creation –, introduce an AI solution there and measure the benefit. A well-defined start beats the big digitalization project in almost every case."}, {"q":"Do mid-sized companies need an AI strategy before starting?","a":"Not a hundred-page one. More important than the strategy paper is a first productive use case from which the company learns. The strategy emerges better from experience than from workshops."}, {"q":"What should SMEs look for in AI providers?","a":"Four criteria: GDPR and data location, control (does external impact run through your approval?), cost model (consumption-based instead of license stacks) and the question of whether the solution can work with your own company data – without context, AI remains generic."}, {"q":"Which AI applications deliver the fastest benefit in mid-sized companies?","a":"Proven use cases have a clear bottleneck: telephone availability (AI call answering), document filing with automatic indexing, texts and content in day-to-day business as well as task and deadline recognition from documents."}, {"q":"What does getting started with AI cost for an SME?","a":"That depends on the model. Consumption-based platforms significantly lower the entry barrier: With webRichtung, the account and users cost €0, and you only pay for actual usage via Credits (1 Credit = €1 net)."}]
language: "en"
source_id: "wissen/ki-im-mittelstand-einstieg"
source_hash: "946599ead9635b3fe7bd94c53c8eb91fa9c7b707bba3a20bd1a5c79c6aa4bf87"
---

The best way for mid-sized companies to get started with AI is unspectacular: choose a concrete bottleneck, put an AI solution into production there, measure the benefit – and only then expand. The big digitalization project with a strategy phase and a team of consultants regularly fails in SMEs due to time and day-to-day business; the well-defined start does not.

## Why "start small" is not just a platitude here

Mid-sized companies have a real AI advantage over corporations: short paths. One decision, one week of implementation, and the tool is in use – what would be a year-long project in a corporation. This advantage evaporates when mid-sized companies copy corporate methods: working groups, strategy papers, tool evaluations over months. The more productive order is the reverse: first a working use case, then the strategy derived from experience.

## Finding the right first use case

What works well combines three properties: a noticeable bottleneck, a measurable result, a limited blast radius. Proven candidates:

- **Telephone availability:** Missed calls are lost orders. AI phone assistants take calls, answer questions and summarize conversations – measurable by calls answered outside of busy times.
- **Document filing:** Searching costs time, filing gets postponed. AI-supported filing reads, classifies and makes documents searchable – measurable by search time.
- **Texts in day-to-day business:** Quotes, emails, descriptions – an AI assistant that works with the real company data drafts in advance and the team finalizes.
- **Deadlines and tasks from documents:** AI recognizes deadline signals in contracts and receipts and prepares tasks – for approval, not on its own.

Choose one of them – the one that hurts the most – and deliberately ignore the others for now.

## Four criteria for choosing a provider

1. **GDPR and data location:** Business data belongs in an environment with a clear legal basis. Ask specifically about the processing location and data processing agreement.
2. **Control as a principle:** Good systems let AI prepare and suggest – external impact such as emails or calls runs through your approval. Anything else is too risky for mid-sized companies.
3. **Cost model:** License stacks per user and tool become expensive as usage grows. Consumption-based models lower the entry barrier – with webRichtung, for example, the account and users cost €0, and you only pay for usage via Credits.
4. **Context capability:** The most important and most overlooked question: Can the solution work with your company data? An AI without access to customers, processes and documents remains a generic copywriter.

## Think about the second step, don't buy it along the way

The first use case should be expandable into something bigger without you having to buy the bigger thing right away. That is exactly what the [webRichtung platform](https://www.webrichtung.de/plattform/) is built for: You start with one module – such as call answering or document filing – and along the way build the shared data foundation from which every AI function added later benefits. One account, eleven modules, no data silos; since 2009 and with over 3,500 companies served, this is not an experiment but grown practice from Germany.

## The first 90 days

1. **Week 1–2:** Choose a bottleneck, set up the solution, test internally.
2. **Week 3–8:** Use it productively, track the metric (calls answered, search time, drafts created).
3. **Week 9–12:** Take stock, fine-tune – and only now choose the second use case.

Those who start this way will have something after a quarter that most AI strategies lack: a working example in their own company. The role the data foundation plays in this is explained in the article [What is an AI operating system for companies?](/en/wissen/ki-betriebssystem-fuer-firmen.html)