Dentro – AI Development & AI Consulting

Implement AI in a company

Implement AI in a company

Effective strategies for modern companies to successfully introduce artificial intelligence in the company.

The world of artificial intelligence (AI) can be intimidating, especially for decision-makers without a deep technical background. Many believe that complex model training and huge amounts of data are required to implement AI in a company, which is understandably daunting. However, the reality is far more accessible. This article looks at the practical side of AI implementation and describes a streamlined process that you can use to successfully use the possibilities of AI for your SME.

Implement AI in a company: common misconceptions

A common misconception is that AI projects require the creation of complex models from scratch. This can certainly be done, but it is completely overkill for most business applications. At Dentro, we therefore focus on a more practical approach that emphasises speed, efficiency and ROI (return on investment).

Options to implement AI

While some AI projects require customised model development, in many cases effective solutions can be achieved through alternative approaches. Different approaches have proven themselves in practice, which are briefly presented and explained below.

Programmatic solutions

Tasks that supposedly require “AI” can often be solved with simple programme code. We see time and again that we are approached with use cases for which it is simply not necessary to implement AI in the company. The focus should always be less on the use of attractive technologies and more on efficient problem solving. And sometimes good old, supposedly boring code simply does the trick.

Examples: Removing certain information from PDFs and other forms of comparably simple document processing.

Use of large language models (LLMs)

Large language models (LLMs) are pre-trained AI models with extensive knowledge and skills. For example, they can perform tasks such as text creation, translation or data analysis. OpenAI uses such an LLM in the well-known ChatGPT application. These language models can be used out of the box. This means that they are accessed via an application programming interface (API), with varying costs depending on the model. However, individual results can also be influenced when using ready-made AI models. For example, through specific prompts tailored to the use case and some other system parameters.

LLMs are often part of the first steps when we are asked how to successfully implement AI in a company.

Examples: Internal AI chatbots or more advanced document processing.

Retrieval augmented generation (RAG) and fine-tuning

In situations where large volumes of your own data need to be processed, LLMs quickly reach their limits without further intervention. This is where more advanced techniques such as Retrieval Augmented Generation (RAG) or fine tuning can help. Both methods mean that the AI solution can access both its existing general knowledge and the specific information from your data, which leads to more accurate and relevant results.

Example: Customer chatbot with internal knowledge about the company.

The Dentro process: a step-by-step guide to implement AI

Implementing AI in a company works particularly well at the beginning by identifying suitable use cases. The process for AI development projects described below is based on a wealth of experience and can be used again and again and individually optimised. It has proven itself in practice, especially for non-technical companies.

The Dentro process for successfully implementing AI in a company

Step 1: Identify a potential AI use case

The first step is to identify a specific business challenge, task or process where AI can potentially improve efficiency, accuracy or customer experience.

Step 2: Conduct a feasibility assessment

A quick assessment is then conducted to determine whether AI is the right solution for the identified use case. Factors such as data availability, project complexity and potential return on investment (ROI) are assessed.

Step 3: Development of a prototype

If the use case proves to be theoretically feasible, an initial prototype is developed to demonstrate the concept and obtain initial feedback. This enables early course correction and ensures that the solution meets the specific requirements. Ideally, the prototype can already be used productively and brings initial gains in efficiency and/or quality.

Step 4: Test and improve

The prototype is then thoroughly tested to identify and resolve any problems. Based on the feedback, the solution is refined and improved until it reaches the desired level of performance. This works best if target results are formulated in advance, the achievement of which can be measured and monitored accordingly.

Step 5: Provision for practical use

In the final phase, the AI solution is implemented in the existing workflows or systems. This ensures smooth integration and enables the benefits of AI to be utilised in day-to-day operations.

Helpful tips

  • Keep an eye on the foreseeable business benefits of the use case throughout the entire process. If the time and cost savings do not justify the effort, it usually makes no sense to pursue the use case further, regardless of its technical feasibility.
  • Concentrate on use cases that do not have to deliver 100% correct results, but for which it is sufficient if a large part of the work is done by the AI.
  • Prioritise use cases that are used internally. This is for the simple reason that the demand for quality, reliability and user-friendliness is in most cases lower for applications for internal use than for those that are to be used by end customers.

The benefits of a streamlined approach

Companies that take a streamlined approach to implement AI can benefit from several advantages:

Faster implementation

By using pre-trained models and proven tools, AI solutions can be deployed much faster than traditional methods. This allows you to realise the benefits of AI faster and gain a competitive advantage. This also increases internal acceptance and strengthens commitment across the organisation to further implement AI.

Lower costs

The need for extensive model training and data engineering is minimised, resulting in significant cost savings. This makes AI solutions accessible to a wider range of companies.

Focus on business value

The process outlined prioritises business value throughout the development cycle. It is always advisable to ensure that the solution directly addresses a specific need and delivers a measurable ROI.


It does not have to be an overly complex endeavour to implement AI in a company, especially in the beginning. Depending on the type of company, it may make sense to start with a single AI project or to implement AI on an ongoing basis. If the focus is on meaningful use cases, initial successes can usually be achieved with relatively little effort, which lays the foundation for further measures.