A graduate of CentraleSupélec in artificial intelligence and ESCP Business School in strategy and consulting, Victoire Bach embodies innovation in the service of healthcare. In 2020, she co-founded HOPIA with Matthias Lesage: their ambition is to optimize hospital resource management through AI. The solutions developed by HOPIA have already improved patient care.

Their recent fundraising round of 3.5 million euros was completed with the support of Kurma Partners, specialized in healthcare, and Iris Capital, focused on deep tech.

When asked about HOPIA, Victoire begins with an example illustrating the current difficulties faced by patients.

“This morning, I went to see my doctor for unusual symptoms. A more thorough examination was necessary, possibly urgently. He prescribed a series of tests to refine or confirm a diagnosis. When I left, I found myself facing a problem: where to start? Should I look for an appointment on an online platform?

Fortunately, a nearby hospital has a person whose role is to help patients organize their appointments as quickly as possible, particularly for urgent cases. But this work relies on a manual method: this person calls each department, one by one, to check availability. First challenge, the test constraints: my stress test must be scheduled after my blood draw, but before an appointment with a specialist.

Then, it turns out that the information, if it is up to date, is recorded on sheets of paper or Excel files. These tools still represent, in many cases, the primary means of managing healthcare professionals’ schedules.

Of course, some practitioners use solutions like Doctolib. However, these tools do not generate the initial schedules. They only come into play once the time slots have been manually determined. In hospitals, for example, anesthesiologists’ schedules must integrate multiple constraints: operating room operations, on-call duties, and consultations. Once defined, these slots can be opened on platforms like Doctolib.

However, not all doctors are registered on these platforms. Moreover, hospital schedule management remains mostly manual, with rudimentary tools. Yet, it is an essential strategic lever to ensure work-life balance for healthcare workers, guarantee effective patient care, and optimize the institution’s resources.

This task is highly complex. They must comply with legal constraints, such as maximum working hours and rest periods. They must also take into account collective factors, such as continuity of care, the difficulty of certain positions (for example, not spending the entire day in the emergency room), or equity rules among colleagues. Added to this are individual specificities: skills, medical restrictions, and personal wishes. Schedule managers, whether department heads or nurse managers, sometimes spend up to 90% of their time on this task, often a source of errors and dissatisfaction.”

 

The Hopia Solution

Faced with these difficulties, few software solutions offer a capability to take all these constraints into account. This is where HOPIA comes in, offering technology based on symbolic artificial intelligence.

Unlike machine learning models, this approach relies on explicit rules, guaranteeing fairness and transparency in decisions. This helps reduce systemic biases that can slip into manual schedules or those generated from historical data.

HOPIA thus positions itself as a co-pilot for schedule managers, making the most of their work without replacing them. The technology centralizes data and provides fair responses, while remaining flexible to the specificities of healthcare facilities and staff.

HOPIA integrates all individual and collective constraints. This includes legal rules, work-life balance, as well as patient needs. Victoire emphasizes that these schedules become more accurate and efficient thanks to Hopia.

Victoire acknowledges that certain aspects, such as predicting patient flow, remain challenges. Despite theoretically accurate models, practice shows that available data and organizational constraints limit their relevance. For now, HOPIA is focusing its efforts on improving its core business.

However, Victoire plans to integrate learning technologies in areas such as skills management and training plans. The goal is to better identify the needs of healthcare professionals and anticipate long-term skills evolution. With a team of 19 people, HOPIA plans to double its workforce in 2025.

 

Generative Artificial Intelligence at HOPIA

“We are exploring different ways to integrate artificial intelligence into our internal processes. For example, we use ChatGPT, particularly in its paid version, for several writing tasks: LinkedIn posts, press releases, customer responses, funding applications, and tender proposals. Of course, this use is done within the regulatory framework: we do not share any patient or personal data. I also support the team to maximize the effectiveness of these tools.

Operationally, AI plays an essential role in reprocessing databases. Raw information received from clients is often disorganized. We reformat non-personal data to adapt it to our needs using these technologies. This reduces the time needed to analyze and structure this data.

We are also working on customizing models like DLLMs (Domain-Specific Language Models). The goal is for AI to automatically generate complete configurations from client reports. Then, our teams validate these configurations, which is much faster than creating everything manually. This allows us to free our engineers from repetitive tasks so they can focus on verification and strategic adjustments, thus adding real contextual and human value.

However, AI has limitations. One of the main challenges remains the explainability of models, particularly for large language models (LLMs). They are often perceived as black boxes, which can pose a problem when it comes to justifying their decisions. I believe it is essential to maintain a critical mindset and not become passive in the face of AI. We must continuously verify and refine results to avoid biases or errors.

In terms of collaboration, we regularly share our experiences with generative AI, whether within a team or throughout the organization. We exchange effective prompts and our feedback on best practices. This is particularly useful for team members less familiar with these technologies, who thus discover how they can simplify their work.

Finally, AI has allowed us to save considerable time on heavy tasks. For example, writing a public tender, which traditionally took two weeks, can be reduced to a day or two. Although human proofreading and adaptation remain necessary, this represents a significant productivity gain.

We are also deploying tools like GitHub Copilot for our developers, which improves their efficiency. In summary, integrating AI into our processes has become an essential pillar to support our growth. However, it requires constant vigilance to ensure responsible and effective use.”

 

Change Management and Hopia Adoption by Clients

Another major challenge identified by Victoire is the adoption of Hopia by end users. Many fear that AI will replace their expertise. HOPIA emphasizes training and awareness, explaining that their solution acts as a co-pilot.

Victoire emphasizes that this reluctance is not specific to AI, but concerns all digital innovations. She shares the importance of supporting users so they understand the benefits, while respecting their need for autonomy and their expertise.

 

Advice for Entrepreneurs

When asked about her advice to companies hesitant to adopt AI, she encourages entrepreneurs to try these technologies by concretely evaluating their impact. She insists on the importance of starting with limited tests, while keeping in mind that AI will not solve all problems. For her, AI is an opportunity to improve existing practices, but its use must be relevant and aligned with real needs.

Photo credit Mathieu Puga