AI in R&D: software transformed into a medical device
AI is fundamentally transforming the development of new medical devices. Software as a Medical Device (SaMD) – i.e. software designed for medical purposes as an independent product – is particularly booming. Hundreds of AI-based diagnostic and therapy software solutions are being developed and approved worldwide. The US regulatory authority alone (the FDA) had already authorised close to a thousand AI-assisted medical devices by mid-2024, around three quarters of which are in the field of radiology. From imaging methods for detecting tumours to algorithms for predicting heart disease, digital health solutions based on AI are literally springing up everywhere.
What is striking about this is that many of these innovations come from newcomers. Established medtech companies, meanwhile, are only just beginning to exploit this potential. Nevertheless, several pioneers are showing the way. Brainlab, for example, employs over a third of its staff in research and development – showing a clear commitment to innovation leadership – while GE HealthCare and Siemens Healthineers are amongst the leading providers of AI- and software-based medical devices, with numerous FDA authorisations.
AI shortens and improves the development process on several levels. In product development, generative design algorithms can come up with thousands of variants of an implant or device in a matter of minutes – something that used to take engineers months. AI helps with simulation and validation too: instead of carrying out time-consuming physical tests on prototypes, scenarios can be run through in virtual environments, evaluating everything from material stress to behaviour in actual diagnostics. Furthermore, it can reduce the number of laborious simulations required by “learning” the behaviour of complex physical processes. Even clinical trials benefit from AI, for example by speeding up the evaluation of study data and using pattern recognition to identify suitable patient groups. In the case of SaMD, in particular, the product itself is a learning algorithm. Here, the benefit of AI is twofold: on the one hand, it can be used to create a new digital product (e.g. an AI app for diagnostics), while on the other hand, the company can use AI to make that product ready for launching on the market more quickly.
One factor that often slows down medtech is regulation – authorisation processes are complex and time-consuming. However, AI can offer support here too. For instance, it can be used to automatically evaluate regulatory requirements and generate documentation for authorisation based on previous successful submissions. AI-supported tools help to avoid errors in documentation and ensure compliance even during the development stage. The ultimate aim is to get from the initial idea to final authorisation in a fraction of the time it currently takes – without compromising on safety and efficacy. The first AI-based medical devices show that this is feasible. The task now is to establish these approaches more broadly in research and development.
AI in production and supply chain management: intelligent manufacturing and increased efficiency
Production, and the operational processes involved, is where AI really demonstrates its potential as an efficiency driver. Medtech companies are caught in a battle between conflicting forces here: the need to meet high-quality standards and regulatory requirements is up against cost and time pressure. AI can help to navigate this tricky balancing act. Intelligent manufacturing (smart factory) means that machines, systems and products are constantly exchanging data and learning. Causal AI offers a promising approach: sensors on production systems continuously record data and AI models look for causal relationships in the data to uncover the underlying causes of failures. Systematically eliminating these causes before a failure occurs can significantly increase system availability. This can drastically reduce unplanned downtimes – offering a crucial advantage when you consider that the costs of downtimes in highly automated production lines can rapidly mount up to the six- or seven-figure range.
Another area where AI can be leveraged is quality control. AI-assisted image processing can detect even the smallest defects or deviations during the production process – far more reliably than the human eye. This reduces rejection rates and prevents expensive recalls. During assembly, algorithms can optimise the flow of production, for example by dynamically adjusting the sequence of work steps in the event of a bottleneck. This is known as a self-optimising production system. Moreover, generative AI can be used to automatically produce work instructions and training documents for employees, individually tailored to each user’s language and level of experience.
AI also comes into its own outside the factory floor, in areas like supply chain management and logistics. Sales forecasts are becoming more precise because machine learning models can take countless influencing factors into account – from historical orders to weather data. This in turn means that inventories can be optimised, avoiding both overstocking and supply shortages. Some medtech companies report that AI-based demand forecasting has led to double-digit percentage improvements in the efficiency of their inventory management. Overall, AI is paving the way for greater transparency and controllability in operations.
What is the vision for AI in an operations context? Imagine a production manager who can visualise the entire value chain – from the raw materials to the finished product – on a dashboard, with the added bonus of AI-generated recommendations for action. Deviations from the plan are spotted immediately and AI suggests solutions in real time – like a navigation system that offers an alternative route when traffic problems arise. This is how the factory of the future will be managed: in a data-based, predictive and agile way. Companies that invest in these resources can reduce their costs, shorten throughput times and ensure high product quality – giving them a competitive advantage in a market that increasingly demands speed and efficiency.
AI in marketing and sales: personalisation, pricing and a patient-centred approach
AI is proving a game-changer not just in development and production, but in marketing too. In the commercial sector – i.e. marketing, sales, pricing and customer service – GenAI is not the only innovation offering brand-new tools for boosting sales and serving customers better. AI can be used to automate the creation of personalised content for different target groups, for example in the form of tailored email campaigns for doctors and buyers.
Pricing is another key area. Traditionally, prices for medical devices are based on cost calculations and experience – but AI enables dynamic pricing that adapts to market conditions in real time. Algorithms continuously analyse market data, competitors’ prices, demand forecasts and even perceived value in the eyes of customers. Equipped with this information, a company can optimise its prices in a flexible way – either cutting them to increase sales or adding surcharges where the market allows. Studies show that such AI-assisted pricing strategies can push up sales and margins significantly.
In terms of sales, innovative companies are opting to use AI-based virtual sales assistants to support their sales teams. These systems scour customer relationship management databases, for example, and identify the most promising leads, including carrying out an automated assessment of which sales pitches could catch on with which customers. In customer meetings, AI can display relevant information or answer questions in real time, leaving sales managers free to focus fully on the conversation.
Tender management is also seeing a huge boost in productivity thanks to AI. A lot of medical technology business – especially with hospitals – is conducted through tenders. AI tools can analyse tender documentation, which often runs to many pages, and pick out key requirements. They can even create initial drafts for bids. This significantly reduces both the effort involved and the risk of missing any important tender opportunities. Companies report significantly higher success rates when AI is applied in tenders, as bids can be tailored more precisely to those asking for them.
Another field with exciting potential for harnessing AI is patient communication and care. Although medtech companies traditionally target doctors and buyers, there is a growing trend in addressing patients directly with the increased emphasis on patient empowerment and the rise of digital therapy companions. Generative AI can be used to create customised information services for patients – such as chatbots that answer questions about a medical device in a clearly comprehensible way, or AI-assisted apps that guide patients through rehabilitation following an operation or through therapy after a diagnosis. One recent example from practice is Roche’s continuous glucose monitoring (CGM) solution Accu-Chek SmartGuide Predict, which uses machine learning algorithms to predict blood glucose values up to two hours in advance and warns patients of impending hypo- or hyperglycaemia at an early stage, so they can proactively adjust their therapy.
Services like these improve the patient experience and, at the same time, highlight a distinctive feature that makes the company’s products stand out. They also provide valuable feedback: AI analyses the most frequently occurring patient questions or problems and reports these findings back to the company, which can use them to further improve their products and services. To sum up, in the commercial sector, AI has the ability to leverage sales potential and consolidate customer relationships. Companies that use AI to enhance their sales and marketing processes will be more agile in responding to market changes and build a closer connection with their customers. This is a crucial advantage when rivals are competing for every last penny in healthcare budgets.
Success factors: operating model and data base
While AI offers all kinds of application possibilities, it is also clear that success cannot be achieved without the right framework conditions. Two factors are crucial in determining whether AI projects at medtech companies really take off: having a suitable operating model and having a solid data base.
An operating model regulates the company’s work with artificial intelligence and embeds it within the organisation. This means, for example, that clear responsibilities and governance are in place for AI initiatives. Leading companies are setting up cross-divisional AI steering committees or centres of excellence that manage strategies, use cases and resources. It is essential that AI projects are not just confined to the IT department. They must be supported by the other business units and specialist departments. Building up an AI portfolio – a pipeline of use cases, prioritised according to value contribution and feasibility – has proven effective. Some of these use cases should be aimed at gaining quick wins to build momentum, while others should be geared towards making strategic breakthroughs. It is important to take a focused approach here: it is better to scale up a few crucial use cases first rather than getting tangled up in multiple pilot projects.
Companies like Medtronic are doing this by using targeted partnerships and acquisitions to boost their AI expertise in their core business areas. For example, Medtronic has set up a platform for AI-assisted endoscopy in collaboration with Cosmo and Nvidia. At the same time, governance also involves risk management: the use of AI in medical technology must be reliable, traceable and ethically justifiable. A robust AI governance model ensures that data security, patient protection and regulatory compliance issues are considered right from the outset. In the life sciences sector in particular, stakeholder trust is essential – an AI tool that makes diagnostic recommendations, for example, must meet the highest quality standards. Appropriate committees and processes for approving AI solutions, similar to validation committees in product development, are therefore a must.
The second success factor is the data base. AI is only as good as the data used to train it. Although many medtech companies are sitting on a treasure trove of data (from clinical trial data to machine sensor data), this often remains untapped: the data is held in silos, incomplete or inaccessible. The solution here is to follow the “FAIR” principles: data should be findable, accessible, interoperable and reusable.
In practice, this means setting up a modern data architecture – such as “data meshes”, which make data available across business areas. It also means investing in data quality and master data management, so that AI algorithms can work with consistent and trustworthy data. According to a recent study by the Massachusetts Institute of Technology, 60% of Chief Data Officers (CDOs) believe that their company is not yet sufficiently prepared for AI as they have not yet got to grips with issues such as data governance and data security. This is reflected in the industry, where more and more companies are creating the role of a data steward or CDO to drive their data strategy forward. Aside from technology and processes, data culture is also essential. Employees must understand that data is a valuable asset that needs to be maintained and shared.
Basically, AI can only fulfil its potential when there is a robust data base available. Clear organisational rules on how AI initiatives are identified, implemented and scaled need to be set out for pilot projects to have a genuine game-changing impact. This preparatory work may seem time-consuming, but it is the key to successfully establishing AI at scale within a company.
Conclusion: medtech without AI is no longer a conceivable scenario
The medtech industry is on the brink of a new era. Artificial intelligence – and not just GenAI – is evolving from hype to reality and bringing genuine added value for patients. Now is the time to be decisive and seize the opportunities on offer. There is no shortage of capital and technologies are readily available – the key to success lies in taking action. Companies that harness the potential of AI throughout their value chain and scale it strategically will shape the future of healthcare.