April 2026
Artificial intelligence is no longer a subject only for large companies. It increasingly appears in small shops, including bakeries and confectioneries. Cameras count customers. Programs help plan production. Systems analyse sales and suggest what may sell better.
W-MOSZCZYNSKI-PPIC-4-2026The European Union introduced a new law governing such solutions: the AI Act. It entered into force in 2024, but the most important obligations will become a practical reality in 2026. Bakery owners therefore still have time to prepare calmly. This is not about frightening anyone with penalties, but about order, data security and clear rules for using new technologies.
What is the AI Act and why does it concern bakeries?
The AI Act is a set of rules for using artificial intelligence, intended to make AI systems safe, transparent and fair and to protect customers and employees. Some uses will be prohibited and some solutions will require additional information for customers. Companies will have to know what AI systems operate within them. For many bakeries, the most important date will probably be the end of 2026, when full application and inspections begin.
Where AI already operates in a bakery
Many owners do not realize they use AI elements. Cameras at the entrance can count customers; sales software can analyse which products sell faster; an online shop can recommend products from earlier purchases. Facial recognition is sometimes used to control storeroom access or recognize regular customers. These solutions improve planning and service and reduce losses, but in 2026 businesses will need to know precisely how they work and whether they comply with the law.
Shadow AI—a silent threat
One of the largest problems is not expensive systems, but uncontrolled free tools. Shadow AI occurs when an employee uses a public AI chat or another free tool and enters company data: customer lists, sales data, internal reports or even a product recipe.
A marketing employee may paste quarterly sales data into a public chat and request a campaign; an accountant may send part of a flour-supplier agreement for summarization; a production manager may enter raw-material-loss data for an efficiency analysis. In every case, data leaves the company environment, often without knowledge of where it is processed or how long it is stored. In 2026, lack of supervision may breach AI Act obligations.
In recruitment, AI may score confectionery applicants and reject those who fail specified criteria. If employment decisions rely exclusively on the system, human oversight is required. Free versions may also use supplied data to train the system further, meaning a trade secret ceases to be secret. In the bakery industry, a recipe can be the most important capital. A clear rule is therefore needed: company data must not be entered into public AI tools.
Advanced AI systems
Some bakeries use facial recognition to count visitors, recognize regular customers or control employee access. Two years ago in Przegląd, I described how services such as Microsoft Azure can identify employee and customer faces. The tools are available, configuration takes several hours and results are immediate. But the new law introduces restrictions. AI may not analyse employees’ emotions at work—for example, whether someone is tired or upset—to assess performance.
Facial recognition processes biometric data, among the most sensitive personal-data categories. In 2026 such solutions may be classified as high risk. An inspection may require proof of legal basis, risk analysis, technical safeguards and purpose of processing.
Imagine a premium confectionery whose camera recognizes a customer regularly ordering birthday cakes. The salesperson learns that the person usually buys chocolate cakes and macarons and is prompted to offer a new collection of flavours. Commercially this is effective; legally, the question is whether the customer consented to biometric identification and knows their face is analysed. Customers must be clearly informed and personal-data rules observed.
A new supervisory authority
A special state authority is to be created in 2026 to supervise AI systems, conduct inspections and consider customer complaints. A customer who believes an algorithm treated them unfairly will be able to report it. Severe cases may bring financial penalties. Support programmes are also planned for SMEs to test solutions safely before full obligations apply.
Knowing the customer better
AI enables more detailed analysis of customer behaviour: purchase frequency, average spending and last visit. Better offers can follow—a reminder on the day a customer usually buys sweet rolls, or a new flavour offered to someone buying similar products. This increases sales, but customers must not be misled or manipulated; the law will examine fairness.
In Przegląd, I published articles on identifying customers, psychological characteristics and purchase propensity. These methods are effective, but if AI performs the analysis, documentation of algorithms and assessment of consumer impact may become mandatory in 2026.
Production planning and loss reduction
One of AI’s most practical applications is production planning. Traditionally, some bread remains unsold and returns may reach more than ten percent. Systems analysing historical data, weather and the holiday calendar can forecast real demand, reduce unsold goods, improve flour and sugar deliveries, lower storage costs and save labour. Less waste means savings and a more responsible, modern image.
A bakery system can forecast which shops need larger pre-weekend deliveries. An algorithm may label customers “price-sensitive”, “impulsive buyers” or “loyal” and adjust discounts automatically. If pricing decisions are automated, this may significantly influence consumer behaviour.
AI also appears in production, analysing oven temperature, humidity and baking time, identifying efficient batches and losses. It may suggest reducing employment on a shift because efficiency is below average. If personnel decisions rely on system recommendations, algorithmic transparency and verifiability are required.
How to prepare
Ignoring the subject is the worst option. Start calmly: identify every place AI is used—sales, monitoring, online shop or marketing tools. Train employees not to paste company data into public tools. Ask software suppliers whether their systems comply with the AI Act and what safeguards they apply. These actions are inexpensive and can substantially reduce future risk.
Summary
A bakery rests on ingredient quality and human skill; that will not change. But data and its protection are now equally important. AI can improve planning, reduce losses, improve service and build loyalty. The key is reason and control—not escaping technology, but using it consciously. The year 2026 need not be a threat; it can organize processes and strengthen market position. A bakery that ensures data security and clear rules will gain both legal compliance and customer trust.
Wojciech Moszczyński
Wojciech Moszczyński—a graduate of the Department of Econometrics and Statistics at Nicolaus Copernicus University in Toruń; a specialist in econometrics, finance, data science and management accounting. He specializes in optimizing production and logistics processes. He conducts research into the development and application of artificial intelligence and popularizes machine learning and data science in business.
Dodaj komentarz