Nurse

In Valais, the nursing shortage is structural. AI will not solve it, but it can shift the centre of gravity of the profession toward what only a human caregiver can give: presence, clinical judgment, and relationship.

The profession today

The Hôpital du Valais (formerly Réseau Santé Valais) comprises six hospital sites: Sion, Viège, Brigue, Martigny, Monthey, Sierre. Alongside these are dozens of nursing homes and a network of community health centres that provide home care across the canton, including in the most remote mountain municipalities. Nurses work here in very different settings, acute care, long-term care, home support, psychiatry, emergency medicine, often under constraints of distance and isolation that large urban centres never face.

The daily work covers a broad spectrum:

• Clinical assessment and monitoring: vital signs, tracking the patient's condition, catching early signals of deterioration
• Administering care and treatment: medication, dressings, injections, IV lines, technical procedures
• Documentation in the electronic patient record: traceability of every act, handover between teams, recording clinical observations
• Coordination with the multidisciplinary team: physicians, physiotherapists, social workers, family caregivers
• Relationship with patient and family: information, psychological support, delivering difficult news, end-of-life accompaniment
• Triage and prioritisation: in the emergency room, in nursing homes, in community health centres, deciding who to see first and why
• Care planning: drafting, adjusting, and reassessing individualised care plans

Swiss studies put the share of nursing time spent on administrative tasks and documentation at 30 to 40 percent¹, time that is not direct patient care.

What AI is preparing

The most immediate and best-documented gain concerns clinical documentation. Today, writing handover notes, updating the electronic patient record, and completing admission and discharge forms represents a workload that is often carried out outside actual care time, during breaks or after shift. Voice-input tools and automatic structuring of the patient record will reduce this burden to a quick validation of a generated summary. There is nothing magical about it: contextual voice recognition trained on clinical vocabulary, paired with a model that knows the structure of the record. The nurse dictates, validates, corrects where needed. Manual entry drops sharply; the time shifts toward validation, correction, and traceability.

Monitoring and early detection. Embedded sensors, connected beds in nursing homes, monitoring wristbands, fall-detection devices, feed predictive models capable of flagging deterioration hours before it becomes clinically visible. AI produces an alert and a probability, never a diagnosis. The caregiver decides what to do with it.

Support for care planning. Drawing on data from the patient record, medical history, current protocols, and available resources, an AI assistant can propose a structured care plan for each patient at the start of treatment. This is a first working draft, which the nurse adapts, refines, and validates against what the data cannot show: the patient's emotional state that morning, the family's unspoken need, the team's fatigue.

Medication preparation and interaction checks. Prescription-support systems, already partly deployed in some Swiss hospitals, will be extended to real-time verification of drug interactions, contraindications, and dosages. The nurse remains the last human check before administration. This automated verification layer reinforces that check, flagging what a tired eye at 3 a.m. might miss, without ever replacing it.

Protecting health data: a specific regime

Health data qualifies as sensitive personal data under the revised Federal Act on Data Protection, in force since 1 September 2023². Its processing by AI systems carries requirements that go beyond what applies to municipal administration.

Medical confidentiality. Article 321 of the Swiss Criminal Code protects professional secrecy, and this protection extends to nurses, not only physicians. Any patient data processed by an AI tool must remain within the perimeter of that secrecy: hosting in Switzerland, strictly limited access, complete traceability of every access.

Hosting. In the absence of an explicit legal basis, any transit of health data through infrastructure outside the Swiss or European perimeter is excluded. For Valais's cantonal hospitals and public nursing homes, local hosting or a certified Swiss cloud is a compliance requirement, not an option.

Clinical accountability. An algorithm can produce an alert or a recommendation. The clinical decision, and the accountability that comes with it, remains named to the caregiver. This chain of accountability must be documented in the patient record: who received which alert, when, and what decision followed from it.

What rises in importance for judgment

When documentation happens largely on its own, when alerts arrive before symptoms become visible, and when care plans are proposed automatically, what remains irreplaceable takes on all the more weight.

Holistic clinical assessment. Sensors measure vital signs. They do not see that the patient has refused food since yesterday, that their gaze has changed, that the way they answer questions is no longer quite the same. This broader reading, which draws together the verbal, the non-verbal, the person's history, and the family context, sits at the heart of nursing judgment. It cannot be automated.

The therapeutic relationship. Presence, touch, voice, simply being there: these are caregiving acts in their own right, documented as such in nursing science. AI can schedule a visit; presence itself remains beyond its reach. In a mountain nursing home where some residents no longer have close family, that presence is sometimes the only real social bond they have. It carries clinical and human weight that nothing replaces.

Handling ethical situations. How far should treatment continue? How should the patient's wishes be weighed against the family's? These decisions, like the decision to urgently request psychiatric support, are made under high pressure, with incomplete information, against the clock. AI can inform; the decision belongs to the caregiver and the team.

Coordination under uncertainty. When several things happen at once, a patient deteriorating, a family in crisis, a short-staffed team, the ability to prioritise, delegate, reassure, and hold the coherence of care together draws on a situational intelligence that language models do not reproduce.

Who keeps the final word?

Composite illustration. A night nurse in an Upper Valais nursing home receives an alert from the monitoring system at 2:30 a.m.: a slight rise in heart rate and unusual agitation in an 84-year-old resident. Vital signs are within normal range. The algorithm classifies the situation as requiring enhanced monitoring. She enters the room and immediately senses that something is wrong: the man is conscious but not responding normally, his gaze is evasive, he is sweating slightly. She suspects a stroke, activates the protocol, and calls the on-call physician. The patient is treated within 40 minutes. The AI alert set the caregiver in motion. The clinical assessment itself was made by the light of her torch and her experience. (Fictional situation, composite of cases encountered in geriatric settings.)

Job profile 2030

Three new competencies will need to appear explicitly in job profiles; they are not taught in initial training today.

The first is clinical validation of assisted alerts: knowing how to interpret a signal produced by a monitoring system or a predictive algorithm within the patient's overall context, deciding whether it is relevant, and documenting that decision. This is augmented clinical judgment, far more than a technical skill, and it requires a solid grasp of algorithmic bias in healthcare.

The second is governance of the augmented patient record: mastering the data flows between the various tools, the patient record, monitoring, prescription-support systems, ensuring the coherence and traceability of clinical information, and catching errors or omissions in automatically generated summaries. The nurse becomes, in part, an auditor of clinical information quality.

The third extends a role already present but set to grow substantially: ethical and relational mediation in a technologically dense environment. Explaining to a patient or family why an alert triggered an intervention at 2 a.m. Preserving the human dimension of care within a context of continuous monitoring. Standing as visible proof that behind the sensors, there is a person carrying both responsibility and presence.

What AI does not do

AI does not fix the nursing shortage. It replaces neither adequate staffing, nor sustainable schedules, nor a retention policy. Poorly deployed, it can even add a further layer of surveillance, alerts, and validation that burdens the work rather than lightening it. Its value only shows up if it genuinely takes administrative work off caregivers' hands, without shifting the load onto an invisible form of digital accountability.

To go further. The deeper question extends beyond AI assisting the existing nursing role: should Valais create an intermediate role between nurse and physician, AI-augmented, to cover areas without a family doctor? This model already exists in Quebec, in the United States, in the Netherlands, and since 2017 in the canton of Vaud. Valais has no equivalent legal framework yet. → Advanced practice nursing in Valais: the missing link between the medical shortage and clinical AI

The Valais context requires plain speaking on this point. According to Obsan projections commissioned by the State of Valais, nursing staff numbers grew by 17 percent between 2012 and 2019, and a 42 percent increase would be needed between 2019 and 2030 to meet needs linked to population ageing⁴. No technological tool closes a gap of that size. What AI can do is change the structure of the work so that every caregiver present spends more time where their presence matters, and less time absorbed by tasks the machine can take on.

Territorial anchoring

Valais is ageing faster than the Swiss average³. Its mountain municipalities have high proportions of isolated elderly residents, poorly served by public transport, with travel times to care that remain a geographic reality. The nursing shortage is felt more acutely here than in Lausanne or Geneva: recruiting for a nursing home at 1,400 metres altitude, in a municipality of 600 people, bears no resemblance to filling a post at Geneva's university hospitals.

If AI reduces a significant part of the documentation burden, even partially, that translates into direct presence time potentially regained: with patients, with families, in the moments where the quality of care truly plays out. In chronically understaffed institutions, that gain is not incidental. It can be the difference between a nursing home that maintains a sufficient human presence and one that outsources, merges, or closes.

That AI is transforming nursing is no longer up for debate: it is already under way. What is at stake for Valais is whether its institutions, large and small, will be equipped and trained to make the most of it before demographic pressure becomes an emergency with no solution.

What the decision-maker must do now

For a hospital or nursing-home management team (Hôpital du Valais, private institutions)

Launch an audit in 2026 of the actual documentation burden in care units: how many hours per full-time equivalent per week go into entering data in the patient record, writing handover notes, filling out administrative forms? This figure is the basis for calculating the potential gain from an AI documentation tool, and the main argument for convincing care teams that the rollout gives them back time rather than adding another layer of surveillance.

For an HR manager in the health sector

Build the three new competencies into evaluation grids and development interviews without waiting for them to appear in HES-SO curricula. Nurses who already practise reflectively with clinical digital tools are levers for internal transformation. Identify them, train them, give them responsibility for supporting their teams.

For a cantonal official (Public Health Service)

Require that any AI deployment in care across Valais come with a data-governance framework compliant with the Data Protection Act and publicly documented. Patients' trust, and caregivers' trust, in these tools rests on transparency: who has access to which data, within what perimeter, with what traceability. This framework cannot be left to each institution's discretion.

---

¹ Data from studies by the Swiss Nursing Association (SBK/ASI) on nursing administrative workload, 2022–2024. ² Federal Act on Data Protection (FADP), AS 2022 491, in force since 1 September 2023. ³ Federal Statistical Office, Demographic trends by canton, 2025–2050 projections. ⁴ State of Valais / Obsan, Health workforce needs projections, Valais 2019–2030.

Jérôme Deshaie is the founder of MCVA Consulting SA, an agency specialising in the AI transformation of organisations in Valais, and the author of Bisse Cognitif.