The medical field is teaching the industry lessons in innovation!
The medical sources cited are from an article in UsineDigitale : “From digital dictation to ambient AI: How medicine is discovering the promises and limitations of generative voice assistants.”
Download the article in PDF format
The healthcare sector is massively adopting generative voice AI solutions and identifying unprecedented gains. These gains are human, financial, and related to the reliability of the data processed.
This news from the health sector validates the strategy of SPIX industry (specialist in generative voice AI for industry) in partnership with Zenidoc (specialist in generative voice AI for health).
In this article, SPIX industry offers a forward-looking overview of the findings made in the healthcare sector and how the industry could draw inspiration from them to accelerate its adoption of generative voice AI solutions for the benefit of these business processes.
The observation
The observation is undeniable: the healthcare sector is far ahead in the adoption and use of generative voice AI solutions.
According to the French Hospital Federation , ” 65% of hospitals already use at least one artificial intelligence solution in production. Document writing and report management are among the primary use cases. ” The industry is therefore very behind in this area, and yet, the writing of inspection reports, quality control reports, and project progress meeting reports should be among the first industrial use cases to be deployed without delay.
This advancement in healthcare is also linked to the growing awareness among stakeholders of the revolutionary potential of the “generative” component of voice AI solutions. According to UsineDigitale , ” where digital dictation delivered raw text requiring review, generative AI delivers a nearly finalized document, which the physician simply needs to validate .” This revolution, driven by the generative component, empowers healthcare professionals and eliminates the frustrations associated with errors and the lack of synthesis inherent in pure speech recognition. In the industrial sector, speech recognition solutions have long been criticized for their impracticality. It’s time for the industry to reconsider its position on this issue!
The industry remains wary of hallucinations from generative AI solutions. Yet the healthcare sector is critical! The French Hospital Federation (FHF) reported error rates of 1.5% for hallucinations and 3.5% for omissions in a 2024 study. Are we certain that an industry professional would have lower error rates when writing a report? And even if they did, the FHF clarifies: ” Even with only 1% errors, careful professional review is necessary before validation.” “. For SPIX industry , the validation step by the industrial operator is essential, for several reasons:
- Maintaining operator competence , especially among new generations.
- Empower the operator to validate the content of their report.
- Promote communication between operators at the time of the report’s release.
In short, everything is in place for the industry to massively adopt generative voice AI solutions. The healthcare sector has even validated the benefits associated with these technologies.
The benefits
Industry has two obsessions: ” increasing metal time ” and ” ensuring end-to-end digital continuity ” of processes. The healthcare sector has demonstrated that the widespread deployment of generative voice AI solutions can achieve significant benefits in both areas.
For the French Hospital Federation , ” these technologies represent a potential lever for freeing up medical time and reducing administrative burdens.” ” If we replace ” “Medical time ” versus “ metal time “… we’re not far from the holy grail sought by manufacturers!
Doctolib goes further and quantifies the benefits its clients have seen from using generative voice AI: “50% less screen time per day, an average of two hours saved, and five times more documented files.” What industrialist hasn’t dreamed of having truly documented field intervention reports, and not just “Nothing to report” in the comments section? Companies in the energy, chemical, and defense sectors are striving for “zero defects” and “100% traceability.” Wouldn’t the widespread deployment of voice and voice assistant technology with generative voice AI solutions solve this problem quickly?
Finally, let’s not forget human communication, a fundamental aspect of many activities. In the healthcare sector, Doctolib reports that ” the most common feedback we receive is: ‘I rediscovered my patients’ faces,’ or conversely, ‘patients say I finally know my doctor’s eye color.'” ” Could generative AI voice solutions bring about a revival of communication between industry professionals? Who hasn’t experienced technical meetings where participants are glued to their screens, focused on taking notes, and not on communicating with each other?
The benefits of generative voice AI solutions are therefore clearly identified. They are distinct from and complementary to those linked to digitalization, as is often believed.
So now, how do we get into the industry, and why?
The outlook
From Google to Meta and OpenAI , the major players in tech regularly predict the end of screens and a new reign of spoken language. The question is how to achieve this in industrial sectors where everything is complex. Would it be simpler in healthcare? Not necessarily, and yet this sector is ahead of industry.
For the young healthcare company Medadom , ” Eventually, all doctors will have an assistant by their side. It’s a necessary step, just like the computerized medical record was in its time. “. The parallel with the digital transformation of industry is clear. As digital systems become increasingly complex, all industrial operators will need an assistant to generate or access information simply and efficiently. The digital transformation of industry is underway, and some processes are more difficult to digitize than others. For example, field operators in quality control, inspection, or maintenance processes still struggle to use digital tools. Generative voice AI solutions free up their hands and eyes, allowing them to focus their attention on the tasks at hand.
Looking further ahead, and considering the new generations who will come to work in industry, the healthcare sector also offers opportunities. Doctolib specifies that AI will profoundly transform the way we practice medicine. The consultation assistant is just the first step. Medical knowledge doubles every five years, and AI can make this data even more accessible. ” This observation also holds true for industry. How can we quickly upskill new employees on increasingly complex or critical industrial processes? How can we make increasingly rich, and sometimes outdated, documentation accessible on long-life installations? At the same time, we already know that younger generations will no longer be willing to spend hours reading technical documents to find the answer to their questions. Generative voice AI solutions could resolve this situation if they are deployed on a large scale across all processes.
Finally, the industry is hesitant regarding the security and confidentiality of data associated with its critical processes. Once again, the healthcare sector, which manages highly confidential data, has taken the lead on this issue. The French Hospital Federation is considering the necessary steps to address the deployment of generative voice AI solutions. It specifies that ” the framework for evaluating healthcare AI should be structured around three pillars: systematic medical validation, auditability of models and data sources, and transparency regarding the infrastructure used, particularly concerning the hosting of health data .”
For SPIX industry , these three points can be translated for the industry as follows:
- Validation of LLM (or SLM for embedded systems) model results by industry experts is essential to ensure the reliability of information exchanged between the operator and a generative voice AI system. SPIX industry has the skills and resources to perform this validation for each manufacturer, confidentially and without cross-referencing data between manufacturers.
- Models must be sovereign , resource-efficient, and locally controlled. Several technologies exist today for implementing generative voice AI solutions for industry. The key challenges now are to control their implementation in sometimes complex industrial environments, ensure the sovereignty of the models used, reduce dependence on potentially intrusive actors, and limit the dissemination of manipulated data on uncontrolled servers. SPIX industry has mastered all these aspects to meet the requirements of critical industries such as energy, chemicals, and defense.
- To ensure data confidentiality and service reliability in any industrial environment, the models must run on internal company servers (on-premises) or in offline embedded mode . All SPIX industry generative speech AI models are operational offline, in offline embedded mode, or on-premises when a connection to internal company servers is possible, to ensure service reliability and data confidentiality in any context.
Conclusion
Today, the healthcare sector is leveraging the full range of generative voice AI solutions to optimize medical time, improve the patient-physician relationship, and ensure the transmission of high-quality information. The deployment of these technologies has been achieved despite the complexity associated with the highly confidential nature of the health data handled.
Why is the spread of this type of approach not yet a reality in industry? Yet, some critical industries share many points with the healthcare sector: complex processes, confidential data, a requirement for total traceability, etc. Three reasons can be put forward: 1- the medical field has been using voice (dictaphones, voice dictation solutions) for more than a decade and therefore has a fairly long feedback loop, 2- major tech players ( Nuance/Microsoft, Doctolib for example ) have massively invested in this area for healthcare, 3- the scarcity of human resources in the healthcare sector creates a requirement to optimize medical time.
In summary, the industry is 10 years behind the healthcare sector in the use and deployment of voice AI solutions for process optimization. It’s never too late to catch up! Adding the “generative” component will undoubtedly allow this gap to be closed quickly and industrial processes to be optimized with operational generative voice AI solutions.
Download the article in PDF format
Point of contact
André JOLY – Managing Director
Tel.: +33 (0)6 25 17 27 94
Email: andre.joly (at) spix-industry.com
Legal entity
Website: spix-industry.com
Linkedin: linkedin.com/company/spix-industry
Simsoft3D SAS – 40 rue du Village d’Entreprises – 31670 Labège (France)
“Voice Experience “, ” SPIX ” and ” SPIX industry ” are registered trademarks of Simsoft3D SAS.