A “lean AI”?
The use of Artificial Intelligence (AI) by companies is already an unquestionable and inevitable reality all over the world. It is not even news anymore.
In 2022, for example, IBM commissioned and released a survey that, among other things, revealed that, already at that time, 41% of Brazilian companies were actively implementing AI in their business. And that 73% of IT professionals in these organizations were accelerating investments in this technological area.
More recently, in March of last year, another study, this time commissioned by Microsoft, showed that 74% of the micro, small and medium-sized companies surveyed, from a variety of sectors, were already using AI. And 90% of them said they were thinking of adopting it even more intensively in the coming years.
The most important question at this moment, then, is no longer “whether”, but “how” to use AI, let's say, intelligently, if you will pardon the redundancy. And, above all, without repeating what happened in the early days of Information Technology (IT), when it became clear that the vast majority of IT projects simply did not deliver the promised impacts and results.
In this context, we at Lean Institute Brasil (LIB) are developing and exercising practices aimed at supporting the implementation of AI, but from a lean perspective, ensuring that it is used to reduce waste, enhance managerial capability, improve working conditions, support problem solving and deliver more value to customers. Without such guidelines, the chances of incurring unnecessary (or ineffective) investments increase greatly.;
The starting point should always be simplification. The greater goal of AI is to make things easier, more direct and more fluid for employees and customers. By simplifying steps and processes, we increase operational efficiency and improve the quality of services and products, contributing more effectively to the organization's strategic objectives.
Considering simplification as the fundamental target for the use of AI, we list below lean concepts and practices that can inspire us when facing the challenge of using the right technology, in the right place, at the right time. A “lean AI”.
Management innovation starts now. Don't miss the chance to be part of it – register for Lean Summit 2025!
Value for the customer
To be truly lean, every AI project, even before any production or digital implementation, should begin with a clear and broad understanding of how the AI being designed can concretely create value for customers.
This lean fundamental also needs to fully underpin each and every investment in AI. Using technology without transparently understanding what the purpose is and what impact one wants to produce can lead to unnecessary spending, that is, more, larger and new forms of waste, as well as frustration with results not achieved. We need to be able to keep the “end customer” perspective as the deciding factor.
On the other hand, if, in addition to the above, we manage to use artificial intelligence to capture consumption behaviors, “click patterns”, that is, the user experience, this could be converted into relevant feedback for leaders and their teams to, for example, make necessary adjustments to products and services, as well as generate more relevant innovations.
Elimination of waste
The use of AI can have a tremendous impact on the redesign of processes and on how human work needs to be organized. This means directing efforts so that each and every unnecessary step at the process level is eliminated or minimized, directing human capacity toward the activities that really matter.
A classic example, which we come across every day inside companies, is redundancy. Steps and activities are repeated in various ways, by several people, in different departments, without anyone even noticing. This is an excellent opportunity to map and use AI to automate work, aiming to eliminate overlaps that only rob us of time and energy.;
In these cases, one can make use of so-called “Robotic Process Automation (RPA)”, which is increasingly being used to automate cyclical activities. However, once again, it is necessary to apply the lean view to automation challenges. Before any intervention involving technology, value streams must be carefully mapped so that critical points are made evident.
Identifying errors, failures and defects
Perhaps one of the most promising contributions of AI is its ability to support us in identifying abnormalities. One of the most powerful concepts of the lean model is the relentless pursuit of identifying and exposing problems. The very origins of lean take us back to this concept.
The famous “jidoka pillar”, an inspiration that Toyota brought from the loom factory and deeply incorporated into its vehicle production practices, essentially refers to the capability that machines and equipment should have to be able to judge “autonomously” whether or not the product has quality.
In the digital age, with the support of AI, we should be able to reach new levels of quality at decreasing costs. If our systems, machines and equipment are designed to help us distinguish errors, failures and defects, we will open the way for the speed and effectiveness of improvements to be leveraged tremendously. For example:
In anomaly detection and diagnosis: AI can be programmed to constantly monitor production or service data in real time, detecting patterns that deviate from normal. Machine learning algorithms can identify failures, defects or quality problems before they become more serious. Once the problem is detected, the system can automatically pause the in-line operation, following the “jidoka pillar”.
In predictive analytics: Using large sets of historical data, AI can predict potential failures in equipment or processes. These predictions help with predictive maintenance, which aligns with “jidoka”, by avoiding unplanned interruptions and quality defects, optimizing equipment uptime and process efficiency. Industry 4.0 uses these mechanisms through RFID and IoT sensors to capture and collect data, and makes use of Deep Learning to feed AI with the intention of performing predictive analyses. For example, in the case of Level 5 autonomous cars.
In computer vision: On assembly lines or in quality control, AI equipped with computer vision (image recognition) can inspect products at much higher speeds and with greater precision than humans. This not only improves defect detection, but also allows rapid intervention, in line with jidoka principles. A well-known case in agribusiness is crop tracking by drones equipped with cameras and AI mechanisms that use image recognition to spot pests, diseases and other production problems.
In responsive automation: AI-equipped systems can be programmed to make automatic decisions based on the information collected and analyzed. This includes not only stopping the system when necessary, but also suggesting adjustments to processes or to machine and equipment settings to prevent defects from recurring. Along these lines, one example is the well-known delivery to production of features of a system or digital product which, when enhanced with AI, can be configured to implement contingency plans such as roll-back, if errors are detected in the deployment, to protect the last working version of the solution.
In feedback and continuous improvement: AI can help collect and analyze feedback from processes and products on an ongoing basis, facilitating improvement. For example, if a specific problem is frequently detected, AI can analyze the root causes and suggest changes to the process or product design. This is very common on streaming platforms such as Netflix, where AI continuously reads patterns of behavior and viewing to offer the best user experience, bringing content suited to each individual.
In integration with enterprise systems: By connecting AI with enterprise management systems (ERP, for example), it is possible to optimize the response to detected problems. AI can automatically trigger work orders, adjust supply chains or realign production schedules. Today, the declaration of standards in the digital environment is enhanced by AI, since from these standards it is possible to establish preconditions and requirements to keep the delivery of value level in the daily routine.
Speed of improvement and problem solving
Until AI, making improvements was an exclusively human task. But perhaps we are facing profound changes in this respect as well. As we saw in the topics above, we increasingly have “autonomous” means capable of identifying anomalies and signaling the existence of problems in real time. This allows us to dramatically accelerate the response time to problems, increasing the precision of interventions aimed at improving processes and the organization of work.
Revealing and solving problems remains a fundamental human capability and perhaps always will. After all, it is still real people who hold responsibility for decisions and the critical eye on performance and business needs. However, a “lean AI” can help a great deal with this challenge, and in a way never imagined.
This may be a possible “ideal state” of the improvement dynamic, with AI as a great ally in identifying performance gaps in real time, pointing out paths and alternative countermeasures. May it become responsive and develop the ability to self-correct inconsistencies at the process level. Not only analyzing and providing complex data and information, but also “proposing” and “making” the necessary transformations.
Asking the right question
Currently, most interactions between humans and AI take place through so-called “prompts”, which are guided by commands we send to the bots with our specific requests. In this context, it is essential, for example, to understand how to ask the “right question” of AI platforms – such as ChatGPT, Gemini, Copilot and so many others.
It is necessary to understand how to think about and format “lean prompts” that will extract and analyze information effectively, as well as offer viable alternatives for starting, developing and enhancing the solution to a given problem or challenge. Without the right question, we will get imprecise answers that can lead us to wrong decisions.
This even involves deeply understanding customers' needs and their respective “pain” points. Consequently, identifying the strategic elements in which AI can bring significant improvements. Then, formulating questions to the AI using specific, clear and direct prompts. And evaluating, based on the answers, how we can refine and align management practices, contributing to strategic objectives.
A “lean AI”?
In-depth discussions at a series of major recent global gatherings on the subject – such as Brazil at Silicon Valley, SXSW and South Summit – give us clues to the enormous potential of a “lean AI" to transform, in ways never imagined, the management of companies, the evolution of markets and relationships with customers.
They can, for example, personalize user experiences, predict future market needs and optimize service delivery. In addition, they will certainly generate new business models and create competitive advantages through continuous innovation and rapid adaptation to social change. They will be able to translate patterns of behavior to further align specific demands, organizational configurations and people development, strengthening economic ecosystems.
In this context, it is likely that, in the near future, we will see not just a "lean system", but a "lean AI system", which will integrate management and artificial intelligence as essential and indivisible parts of a company's operations and management.
Its foundations would include a deep integration of AI with processes and human work, in continuous optimization and learning. Also, generating adaptive structures that allow rapid responses to change while maintaining the focus on value creation. And strengthening an increasingly advanced use of data and analytics to support strategic and operational decisions.
The future has arrived. Companies need to understand how to think about and execute AI on a lean basis. It is not something that is yet to happen. Perhaps we are already facing great potential levers to accelerate progress.
