Artificial Intelligence (AI) has become a pervasive topic, with narratives frequently oscillating between its potential to enhance the quality of life and its potential to displace human labor. It is probable that AI will, in practice, lead to both outcomes. Within MedTech engineering organizations specifically, this represents a recurring pattern observed over several decades. The demand for engineers would be considerably greater if powerful enabling technologies, such as integrated development environments (IDEs), computer-aided design (CAD), and Electronic Design Automation (EDA) tools, were not available.  While these tools evolved at a more gradual pace, their ultimate effect was the same: they augmented the efficiency of engineering tasks, thereby enabling organizations to achieve greater output with fewer personnel. Beyond efficiency gains, these tools also yielded significant improvements in the quality of engineering deliverables as well as expanded the realm of what is possible to achieve in healthcare.

Consider an engineering team attempting to develop a product without the assistance of an IDE, CAD, or EDA system. Such a team would be incapable of producing deliverables with a level of complexity comparable to that of their modern counterparts. In a similar vein, engineering organizations that delay the adoption of AI will find themselves at a significant disadvantage, as AI is becoming an indispensable tool for modern MedTech engineering. Although identifying the optimal application of AI within the product development lifecycle may be less obvious compared to implementing tools like an IDE, it is of equivalent critical importance.

AI represents the latest iteration in this pattern of advancement of engineering tools. Its emergence is characterized by greater velocity, and, unlike preceding tools, its applications possess a far broader scope, extending beyond the enhancement of a single discipline. Consequently, a pivotal question arises: in which areas can AI be most effectively applied within an engineering organization?

An engineering process, or more comprehensively, a product lifecycle, comprises a series of interconnected business processes. These processes inherently possess imperfections, which in turn generate inefficiencies. For engineering organizations, inefficiencies typically materialize as lost time, unnecessary effort, or the necessity for rework. AI presents a new tool for the optimization of these processes.

Lean Management is a systematic methodology focused on maximizing value while concurrently minimizing waste (aka inefficiencies). Its principles offer a foundational framework for initiating an AI transformation.

Lean AI

The tenets of Lean Management provide a guide for identifying potential domains where AI can accelerate development efforts. The Lean methodology usually starts with a value stream mapping exercise. While FDA regulated development organizations generally possess well-documented procedural frameworks detailing what is to be done, this exercise necessitates a more detailed investigation into how these tasks are executed. For instance, an inquiry might explore the structure of meetings convened to generate requirements or assess the number of iterations typically required to translate a user’s need into an approved set of system requirements.

Such detailed inquiries are essential for pinpointing areas of inefficiency, or “friction,” within the development process. The objective is to construct a detailed map of the workflows through which development deliverables are produced. With such a map, points of friction become easy to spot, either through quantitative measurement or through the articulated frustrations of team members.

Subsequent to compiling an inventory of these areas of friction, it is important to conduct a thorough analysis to understand their underlying causes. In some instances, friction arises from an inherently inefficient process; in those cases, the remedy may involve the removal of non-essential steps or the addition of missing steps. However, other sources of friction are indicative of the inherent complexities of medical product development. It is in these areas that AI is most likely to yield a substantial impact. An examination of various stages of product development reveals several common sources of friction.

Early Development

The initial phases of any development effort are frequently characterized by considerable friction. The project team may need to acquire familiarity with a new product, new patient condition, and new team members. Furthermore, substantial user, market, and competitive information are required to define a product that will yield commercial success.

A primary area for consideration is the onboarding process for new team members joining a project, which is a recognized source of friction. Symptoms can include protracted onboarding periods, sometimes extending over months, or the harder to observe, increased error rates among engineers lacking a complete understanding of their operational context. A key strength of AI lies in its capacity to consolidate information into digestible summaries. For example, summarizing historical defect data or product requirements from previous iterations can enable new personnel to attain a more rapid and thorough understanding of the problem domain. It is crucial to acknowledge that such summaries do not obviate the need for a review of detailed information, but they do serve as an accelerant in the learning process.

Another domain requiring careful scrutiny is that of initial product definition activities. The development of requirements and preliminary clinical risk analysis are both areas that often start slowly. While AI is not a substitute for the execution of these activities by qualified engineers, what proportion of product requirements are entirely novel to your specific product? Requirements that are not unique could potentially be generated quickly by AI. Engineers must retain responsibility for the review and approval of these AI-generated requirements, a standard that also applies to requirements authored by human personnel.

Finally, a significant effort in the early stages of a program can be devoted to establishing DevOps pipelines or constructing the initial software framework. Although AI currently faces limitations in complex code development, it demonstrates efficiency in generating the boilerplate code necessary for integrating components within a DevOps pipeline, for instance.

Mid Development

The middle phase of a development effort often provides a sense of stability, yet it is not without its own elements of friction. In this stage, inefficiencies, predominantly in the form of rework, may be more subtle to detect.

Consider the timing of unit test development by engineers. It is a suboptimal practice, yet frequently observed, that many engineers defer this activity. An eagerness to begin work on additional components or features can lead to the neglect of early unit testing efforts. AI can be highly effective in generating unit tests and associated test data, accelerating an activity that is often postponed and generally perceived as less engaging than feature development.

Another area that typically lacks transparency is coding efficiency. The amount of time personnel might spend consulting online forums or other resources to find solutions for relatively common problems is likely greater than often acknowledged. AI excels at information consolidation. If an initial AI-proposed solution proves inadequate, the prompt can be refined to elicit a more precise result. Ultimately, this provides engineers with a tool to obtain solutions more quickly, allowing them to concentrate their efforts on innovation.

Conclusion

Lean is a journey of continuous improvement.  An inventory of opportunities establishes the foundational elements of an improvement roadmap. Areas exhibiting the highest levels of friction correspond to those with the greatest potential for AI to eliminate inefficiencies and consequently deliver productivity enhancements. These improvements in productivity go beyond mere cost savings; they create opportunities for the development of higher-quality products delivered more quickly.

While enhanced efficiency serves as one metric of success, perhaps of greater significance is the enablement of engineering talent to dedicate their intellectual efforts towards innovation. Innovation is a primary driver of growth within medical technology organizations, and the reduction of friction within the product development process enables patient access to life-enhancing or life-saving technologies.