AI use cases in companies across 10 industries – what works and how to choose your first project

There are more examples of AI use than one company can implement in a year. So before you turn to the description of your industry, prepare a simple selection method. It protects you from a project that looks good in a presentation but does not change day-to-day work.

The steps below usually take one or two meetings. For the cost–benefit calculation you can use the formula from our article on the cost of implementing AI.

To begin with, look at the four universal uses in the list below. In many companies these are the best first project, because they concern work that most of the office does.

Some of them do not require building a separate solution. An AI assistant in an office suite, such as Microsoft Copilot, can summarise emails and meetings and prepare draft documents. It only shows data that the employee has access to.

Law firms work with long texts and templates, so AI quickly finds a use here. Legal information systems now also have AI assistants, such as Wolters Kluwer’s LEX Expert AI, which answers based on the content of the LEX database and cites its sources.

The biggest constraint is professional secrecy. Client documents may only go into tools covered by a contract and clear data processing rules. Our case study of a law firm that implemented an AI assistant for contract analysis shows how such an assistant works in practice.

KSeF, Poland’s National e-Invoicing System, has changed the starting point. Invoices from domestic suppliers now mostly arrive as structured data, so reading them from PDFs is no longer the main use of AI in accounting. Document reading is still useful for foreign invoices, receipts and documents outside KSeF.

AI has more work to do, however, with what happens to an invoice afterwards: assigning the cost, matching it to the order and explaining discrepancies.

In business-to-business sales, the biggest time drains are quotes and the administrative work around customer conversations. AI takes the load off salespeople where they repeat the same work for every enquiry.

For this to work, the data must be in a customer relationship management (CRM) system. If the customer’s history sits in the salesperson’s emails, AI will not see it. Our case study of a stonemasonry company shows an example of cutting the time needed to prepare a quote.

In logistics, a lot of the work is answering status questions and handling transport documents. Some uses, such as route planning, require specialist software, not just an AI assistant. Such systems are usually chosen together with the supplier of the transport management software.

Marketing was the quickest to adopt AI for text and images. The challenge today is not using it but setting the rules: which customer data may be pasted into a tool, and who checks content before publication.

From 2 August 2026, companies that publish realistic images, recordings or statements generated by AI must disclose this. In an online shop, many AI features, such as product recommendations, are built into the e-commerce platform. First check what your subscription already includes.

In HR, AI works best for answering employees’ recurring questions and preparing documents. Caution is needed when it comes to assessing people.

The AI Act classes systems for selecting candidates and assessing employees as high-risk. The obligations for them apply from 2 December 2027 and include human oversight and informing employee representatives before the system goes live. The ban on recognising employees’ emotions in the workplace has applied since February 2025.

Customer service is one of the oldest areas of AI use. The difference between a simple chatbot of a few years ago and today’s assistant is that the assistant uses company data: order status, customer history, the current terms and conditions.

From 2 August 2026, customers must know that they are talking to AI. Also plan an easy route to a human – a customer who gets stuck talking to a machine is more dissatisfied than one who waited for an agent.

In manufacturing, AI works on two fronts. The first is the machines: quality control with cameras and predicting failures from sensor data. The second is the office and the shop floor: technical documentation, procedures and orders.

Machine projects require sensor data and usually take longer. Document projects deliver results faster. Our case study of a knowledge assistant for production engineers shows what this looks like in practice.

Wholesalers and distributors receive orders and documents in many forms: by email, in files, by phone. Retyping them into the system is a typical place for a first AI project.

Our case study of a plant importer shows an example: the system reads supplier documents in different formats and transfers the data into the company’s system. A similar problem of scattered data is solved in our case study of a made-to-order furniture manufacturer that replaced several spreadsheets with a single system.

In healthcare, AI reaches registration and administration fastest and diagnostics slowest. Health data enjoys special protection under the GDPR, and tools that support diagnosis may count as medical devices.

AI in medical devices that require assessment by an independent notified body is subject to the obligations for high-risk systems from 2 August 2028. These obligations come on top of the requirements of medical device legislation. Our case study of a network of medical facilities shows what online registration linked to doctors’ schedules can look like.