Developing an AI Strategy in Your Company: From Quick Wins to Management

Which processes are best automated with AI in the company

In principle, AI can be meaningfully applied in almost every area of a company today. From quality control to production optimization, to creative processes and supportive as well as controlling measures, there are numerous use cases. Which ones make the most sense depends heavily on the specific company and its existing processes.

There is no one-size-fits-all answer, but there is a clear rule of thumb: Processes with recurring, well-documented workflows and a lot of manual effort are usually suitable as a first step. They provide quickly measurable results and create internal support for further steps.

Why AI automation also works with older infrastructure

A common misconception is that AI automation necessarily requires expensive, new infrastructure. This is not true. While a powerful server with a large GPU provides noticeably more speed and performance, existing older infrastructure is often sufficient for many use cases, especially when an answer does not need to be available within milliseconds.

Those working with an external provider are usually independent of their own server performance anyway. Even small, local instances starting from a few euros per month are often enough to get started with AI. Nowadays, there are even models that run directly on smartphones and handle text, image, and audio processing.

Typical Quick Wins in AI Implementation in Companies

Many companies are already using AI automatically without labeling it as a strategic project. Anyone who gathers information about a language model instead of researching for weeks has already achieved a classic quick win. In the corporate context, it looks quite similar: Instead of manually searching through forty pages of documents, an AI system delivers the relevant answers in seconds.

Furthermore, there are further transformation processes up to semi-automation of complex tasks, such as troubleshooting or classifying inquiries. Such tasks do not necessarily have to be performed by humans and can often be more effectively utilized elsewhere in the company.

Why AI implementation is also worthwhile for small and medium-sized enterprises

Especially for small and medium-sized enterprises, AI today offers enormous potential because it has become affordable. Many businesses are simultaneously struggling with a shortage of skilled workers and high international competition, while they cannot easily compete with cheaper locations as a high-price and high-tax country.

While large corporations can afford expensive, custom-developed infrastructure, modern LLM solutions like ChatGPT and similar systems provide a smart, supportive system for relatively little money that assists nearly everyone in the company.

Whether AI use already represents a competitive advantage for one's own industry or will only become a necessity in the near future depends heavily on the respective industry. However, it is clear: Those who hesitate today risk that a competitor will first benefit from the efficiency gains.

Make or Buy: How Companies Make the Right Decisions for AI Solutions

In classic software development, the question of Make or Buy has always arisen, and it is no different for AI solutions. In most cases, the decision today favors Buy, as companies rely on existing, pre-trained models. Training a proprietary model from scratch is not sensible or necessary for most businesses.

Where AI offers enormous potential for classic software development itself is in prototype development. Assumptions that previously had to be tested laboriously and expensively can now be quickly verified internally with a rapid, deliberately non-productive prototype before larger investments are approved.

This also applies to the classic make-or-buy decision regarding existing software solutions: Many companies rely on large, expensive licensing solutions that can do significantly more than is actually needed. With AI, a tailored, cost-effective alternative can often be built that pays off within one to three years.

What IT infrastructure and data are truly needed for AI implementation in the company

How much infrastructure and data are necessary for an AI transformation strongly depends on which processes are to be implemented. Modern LLM solutions require significantly less than many expect. Even small servers are often completely sufficient, especially when absolute top speed is not required.

Those working with an external provider usually do not have to worry much about this infrastructure question anyway. But even with internal infrastructure, a lot can be achieved as long as an answer in fractions of a second is not required.

Why IT management should be involved in AI projects from the start

A common scenario in companies: Management decides to outsource AI projects without involving their own IT management. Those in this position should first ask themselves whether they have been proactive enough on the topic. In many organizational structures, IT managers act rather cautiously. The danger here is that someone else in the company may take the lead and set the direction in a way that may not be sensible from an IT perspective.

It is therefore crucial that all parties communicate with each other. Management often provides information that is critical for a transformation process, and the same applies in reverse for IT. Those who do nothing leave the decision to others, usually without the diligence that would be desirable from an IT perspective.

How IT management or management convinces a skeptical management of AI

A fundamentally skeptical management is rarely convinced solely by a PowerPoint presentation. What matters is a well-thought-out concept with solid analyses, not a hype-based argument like: everyone is doing it, so we must too. Instead, concrete cases with actual, measurable added value are needed.

The biggest stumbling block often lies even before that: How do you even arrive at reliable cases? Here, exchanging ideas with people who already have experience with similar projects helps. A genuine analysis of current processes, including optimization potential, forms the basis for tailored solutions with realistic assessments.

Those who argue with concrete numbers and traceable results instead of pure enthusiasm have significantly better chances with a skeptical management. External moderation can also help: Internally, new ideas are quickly questioned, while an external perspective often creates more trust.

How AI implementation in family businesses succeeds without destroying what works

Especially during generational transitions, the question often arises of how to introduce AI and digitalization without jeopardizing structures that have proven themselves over decades. Many of these companies were once founded with an innovative idea and have since successfully operated for over thirty or forty years, often with targeted adjustments but without major structural renewal.

This is exactly where AI offers the opportunity to quickly and purposefully introduce something new. Many processes that have grown over the years can be questioned, even if they have worked well so far. It is important not to approach it with a sledgehammer and change everything at once, but to examine processes selectively for potential and to introduce changes slowly and tested where they can actually improve something. This significantly reduces the risk of damaging something that works.

In the AI workshop by LEAN-CODERS, we will jointly develop your prioritized use cases and a realistic roadmap. Also read Part 1 of this series, to avoid the most common mistakes from the start.