# AI in Mid-Sized Companies: Starting on a Limited Budget

Author: Marin Niehues, Niehues Consulting · Published: 2026-05-11 · Updated: 2026-11-16 · AI Strategy & Business Case
Source: https://www.niehuesconsulting.com/blog/ai-in-mid-sized-companies-limited-budget/

A mid-sized company can start with AI on a budget in the mid five figures if it begins with one use case whose benefit can be measured in daily operations. The sequence of steps is decisive. A company that starts with a clearly defined process proves the benefit within six to twelve weeks and wins the support of the business units for the next step. This article describes a three-stage approach, sets out the cost of getting started and explains the criteria for choosing the first use case.

## How does starting with AI differ for a mid-sized company and a large corporation?

Mid-sized companies start with fewer specialists and shorter decision paths than large corporations, and that calls for a different approach. Below roughly 1,000 employees there is rarely a dedicated data team or platform department, and there is little room for initiatives that only show an effect after two years. In return, the executive team can decide within days where to begin, and the business units work close to the process.

Adoption figures show the gap. According to the official survey on the use of information and communication technology, just over a quarter of companies in Germany with at least ten employees used AI in 2025, 26 percent to be precise. Among large companies with 250 or more employees the share was 57 percent, among small companies with 10 to 49 employees 23 percent. Mid-sized companies have ground to make up, and an orderly start lets them close it.

From a delivery perspective, success hinges on one question: who in the company uses the result, and what changes in their working day? A budget that flows into licenses and tools before this question is answered produces spending without impact. This yields a principle for the executive team: every expense is tied to a process and to an accountable person in the business unit. Where either is missing, nothing is procured. This applies to licenses as much as to consulting services. The article [AI management: what executive leadership must steer](/blog/ai-management-what-executives-must-steer/) shows how the first step fits into the overall management of AI.

Some obligations apply from the very first use case. The European AI Act requires companies that deploy AI to ensure that their staff have sufficient AI literacy. Where the application processes personal data, the GDPR applies. Where a works council exists, it must be involved in technical systems capable of monitoring behavior or performance. The article on the [EU AI Act](/blog/eu-ai-act/) sets out the obligations. This article provides an overview and does not constitute legal advice.

## The three-stage entry model

The approach is divided into three consecutive stages. Each stage ends with a decision by the executive team.

- Stage 1, Understand: the process is mapped. Requirements are formulated and the business case is calculated. The result is a reasoned sequence of two to three use cases.
- Stage 2, Build: a prototype runs in the real process with real data and users from the business unit. It proves the benefit before further budget is released.
- Stage 3, Introduce: the prototype goes into operation. Processes and responsibilities are defined, and the business units are enabled to use the result permanently.

Each stage receives a decision criterion that is set in advance. After stage 1 it reads: the business case shows a benefit confirmed by the head of the business unit. After stage 2: users in the business unit work with the prototype day to day, and the metric moves in the expected direction. After stage 3: operations are assigned to a team, and the result is measured continuously.

**Each stage ends with a decision by the executive team based on evidence. Budget is released only once the benefit has been demonstrated in the process.**

## Selecting the first use case

The choice of the first use case determines the pace and acceptance of the entire program. Four criteria are enough to turn a list of ideas into a sequence. The detailed assessment method is described in the article [prioritizing AI use cases](/blog/ai-use-case-prioritisation/).

**Checklist for the first use case**
- Recurring process: the task occurs daily or weekly and follows recognizable rules.
- Available data: the required information is digital and assigned to an owner.
- Measurable result: processing time or error rate can be compared before and after deployment.
- Committed business unit: a head of department wants to use the result and provides time for testing.

Document processing, quotation, reporting and customer service meet these conditions. In quotation, the application drafts offers from the inquiry and the price list; in reporting, it turns monthly figures into commented reports; in document processing, it extracts data from incoming invoices and delivery notes; in customer service, it drafts answers to standard inquiries. Use cases with a direct effect on processing time are better suited to the start than those whose effect only appears in revenue, because the evidence is available sooner.

A use case that meets all four criteria becomes the pilot. The remaining candidates stay on an ordered list and follow as soon as the first case is in operation.

## What does it cost a mid-sized company to start with AI?

With external support, analysis and prototype for a first use case cost a mid-sized company around 20,000 to 85,000 euros, a sum in the mid five figures. Rollout and operation follow only once the benefit has been measured. The calculation is based on day rates in the German consulting market, which according to the fee study of the Federal Association of German Management Consultants typically lie at around 1,000 to 2,500 euros per consultant day.

**Worked example: what the start costs**
- Stage 1, analysis and business case: 5 to 10 consultant days, around 5,000 to 25,000 euros.
- Stage 2, prototype in the real process: 15 to 24 consultant days, around 15,000 to 60,000 euros, plus license and usage costs for the test phase in the low four figures.
- Total up to proof of benefit: around 20,000 to 85,000 euros.
- Stage 3, rollout and operation: planned after the proof on the basis of measured values. As a rule of thumb from delivery practice, rollout and operation cost at least as much as development over the life of the application.
- Internal time: for stages 1 and 2 the business unit additionally provides one to two days per week of a knowledgeable person.

The lower half of the range applies to a process with existing digital data and a standard application that is configured. The upper half applies as soon as interfaces to an ERP system must be built or data must first be prepared. A budget that covers only the prototype therefore funds half the initiative. An overview of all cost items is given in [planning an AI budget](/blog/ai-budget-planning/).

Public funding lowers the entry barrier in Germany. The Mittelstand-Digital centers funded by the Federal Ministry for Economic Affairs offer small and medium-sized enterprises free information events and implementation projects on AI. In addition there are digitalization programs of the federal states, whose conditions and funding rates change regularly and should be checked with the responsible state development bank before applying. Funding does not replace the business case. A use case that is economical only with a grant will not pay its way in ongoing operation.

## How long does AI implementation take in a mid-sized company?

The first use case is in regular operation after four to six months if the three stages follow one another without interruption. Stage 1 takes two to four weeks, the prototype in stage 2 proves the benefit within six to twelve weeks, and the rollout in stage 3 needs a further two to three months, above all for adapted processes and user training. Each further use case moves faster, because data access and operational accountability are already settled.

Two metrics per use case are enough for ongoing management. The benefit metric measures the effect, for example processing time saved per transaction. The usage metric measures adoption, for example the share of transactions that actually run through the solution. Together they show the executive team whether a use case should be expanded or adjusted. How a proven prototype reaches regular operation is described in [from pilot to scale](/blog/from-pilot-to-scale-ai-operations/).

A further control point concerns accountability. If several business units buy their own licenses, uncontrolled growth with duplicate costs follows. The executive team therefore appoints one function to coordinate the selection and sequence of applications.

## Niehues Consulting as a partner for getting started

Mid-sized companies rarely have the capacity to handle analysis and implementation alongside daily business. [Niehues Consulting](/) supports all three stages from a delivery perspective. In AI management, requirements and the business case are developed, and the sequence of applications follows from them. The prototypes prove the benefit in the real process. In operationalisation and rollout, the solution goes into operation with internal or external teams, with clear responsibilities and enabled business units.

The approach follows the Niehues Delivery Model with its phases Understand, Build and Introduce. Three of its review questions matter most for a start on a limited budget. Which business result should change measurably? Does the prototype prove the benefit in the real process with real data? Who is accountable for operations from day one? Marin Niehues works on projects from VC-funded scale-ups to DAX 40 corporations and has been deployed in 20+ countries on six continents. The entry point is the free 30-minute initial consultation, in which the first use case is identified: [book an initial consultation](/#kontakt).

**Routes into AI**
Criterion | Tool first | Internal solo effort | With Niehues Consulting
Starting point | License and features | Spare capacity in the business unit | Process and business case
Proof of benefit | After rollout | After several months | Prototype in the real process before further budget release
Adoption in the business unit | Not planned | Depends on individuals | Part of the approach through to rollout
Transfer into operation | Open | Open until capacity frees up | Planned with internal or external teams
Delivery accountability | With the vendor for the product | With the business unit alongside daily work | With a delivery manager accountable for the result

**Recommendation for the executive team**
A company starting on a limited budget chooses a use case with measurable benefit, releases funds in stages and plans the rollout from day one. Up to proof in the real process, a sum in the mid five figures is enough, and that proof is the basis for any expansion.

**Sources**
- Federal Statistical Office of Germany (Destatis): Use of information and communication technologies in enterprises, 2025
- Federal Association of German Management Consultants (BDU): Fees in management consulting, fee study 2022

## Frequently asked questions

### How much budget does a mid-sized company need to start with AI?

For analysis and prototype of a first use case, a sum in the mid five figures is enough, around 20,000 to 85,000 euros with external support. Rollout and operation are planned only once the benefit has been measured. Over the life of the application they cost at least as much as development.

### How many SMEs already use AI?

According to the official German survey, 26 percent of companies with ten or more employees used AI in 2025. Among small companies with 10 to 49 employees the figure was 23 percent, among large companies with 250 or more employees 57 percent. Smaller firms therefore lag clearly behind large corporations.

### Where should a mid-sized company start with AI?

With a recurring process whose result can be measured and whose business unit is committed. Good candidates are extracting data from incoming invoices or preparing quotations. The choice follows the four criteria of the checklist.

### How long does it take to implement AI in an SME?

The first use case is in regular operation after four to six months. Analysis and business case take two to four weeks, the prototype six to twelve weeks and the rollout a further two to three months. Subsequent use cases move faster, because data access and responsibilities are already settled.

### What funding is available for AI in German SMEs?

The Mittelstand-Digital centers of the Federal Ministry for Economic Affairs support small and medium-sized enterprises free of charge with information and implementation projects on AI. In addition there are digitalization programs of the federal states, whose conditions should be checked with the responsible state development bank before applying. The use case must pay off without a grant as well.

### What legal requirements apply when an SME starts using AI?

The AI Act requires that staff who use AI have sufficient AI literacy. The GDPR applies to personal data, and an existing works council must be involved in systems capable of monitoring behavior or performance. Assessing the individual case is a matter for legal advice.

### Does a mid-sized company need its own AI team?

For the start, a named accountable person in the business unit and a function that manages delivery are enough. Internal or external teams build the prototype and transfer it into operation. A dedicated AI team pays off once several use cases are running in operation at the same time.

### How does the executive team know the start is working?

By user behavior and by the metric. If employees use the solution day to day and the benefit metric defined in advance moves in the expected direction, the evidence is in place. The next stage can then begin.
