Winegrower-Winemaker

The Valaisan vigneron-encaveur lives off their signature as much as off their grapes. AI will not make the wine in their place: it will shift the battle to where it is now actually won — to everything that comes before the vine and everything that follows the cellar.

The job today

Valais is Switzerland's leading wine region: around 4,600 hectares in production, held by some 18,000 owners across nearly 73,000 parcels, spread over 62 wine-growing communes¹. Within this fragmented landscape, the vigneron-encaveur occupies a distinctive position: growing the grapes, vinifying them, and selling the wine under their own name. Their estate rarely exceeds a few hectares, their team can be counted on one hand, and their economics rest on signature rather than volume (Chapter 7 of the essay makes this point directly: no one will pay for a Valaisan Cornalin the price of a great Burgundy if it is merely one Alpine red among others).

Their year spans a range that reaches far beyond the vine itself:

• Vineyard work: pruning, disbudding, leaf-thinning, treatments, harvest, often across steep and fragmented parcels
• Winemaking and ageing: pressing, tank monitoring, blending, bottling
• Regulatory administration: transport permits, harvest inspection, AOP traceability, cellar declarations
• Sales: direct sales, tastings, trade fairs, relationships with restaurateurs and private clients
• Communication: tasting notes for each vintage, website, newsletters, social media, often in several languages
• Export and competitions: entry files, correspondence with importers and journalists
• Business management: invoicing, inventory, seasonal staff, investment

The 2024 harvest, one of the three smallest in fifty years (34 million kilograms pressed)², is a reminder that this trade remains, above all, exposed to the climate. No tool will change that. Which is precisely why time spent at a desk rather than in the vineyard or the cellar costs twice as much.

What AI is preparing

Multilingual content for every vintage. Tasting notes in French, German, and English, presentation copy, cuvée descriptions for the website and for retailers: work that used to call for a communications agency or entire evenings can now be handled in-house in a few hours per vintage, in the estate's own voice — provided the winegrower defines that voice and keeps watch over it. AI drafts; the winegrower tastes, corrects, and signs off.

The files that open doors. Entries for international competitions, files for importers, responses to restaurant-trade tenders: standardized, demanding, multilingual documents where drafting time used to be the real barrier. Pre-processed by AI from the estate's own data, they now come together in a fraction of the time; the balance of power with the large houses, which had whole teams for this kind of work, starts to even out.

Pre-filled regulatory paperwork. Transport permits, declarations, AOP traceability: repetitive workflows well suited to automatic preparation under supervision, in keeping with the requirements of the cantonal harvest inspection service. The winegrower validates and remains responsible for every declaration.

The estate's memory. Cellar logs, plot histories, blending decisions, tasting notes accumulated over twenty vintages: once digitized and indexed, these become a searchable memory that can be passed on to the next generation. For indigenous grape varieties (Petite Arvine grew from forty to two hundred and fifty hectares over thirty years; Cornalin, Humagne Blanche, Amigne, and Rèze follow more fragile trajectories³), this fine-grained documentation is an investment in cultural sovereignty as much as a commercial tool.

Estate data: the prerequisite

The cellar log, the land registry of plots, the client file, and treatment histories make up the estate's informational capital. Three rules apply before any deployment: client data falls under the nFADP (hosting and subcontracting arrangements must be documented); operational data must not, through poorly configured free tools, end up feeding competitors' or platforms' training corpora; and the estate's voice (copy, descriptions, philosophy) deserves to be set down in an instructions file the winegrower controls, rather than scattered across chat histories.

What rises in importance for judgment

Taste and blending. Deciding that a cuvée is ready, that a blend holds together, that a difficult vintage will be owned rather than corrected: this sensory and stylistic judgment is the estate's signature in its purest form. No model carries it.

The authenticity of the voice. A generated text can be accurate and still ring false. A winegrower who lets AI dress up a frost-damaged vintage loses, in a single newsletter, credibility built over twenty years. Validating the tone becomes a commercial act of the first order.

Reading the land. Deciding on a treatment, moving the harvest forward ahead of the rain, sacrificing a parcel: weather models and sensors inform the decision, but the winegrower makes the call, drawing on their knowledge of each hillside parcel's microclimate.

The long relationship. The importer known for fifteen years, the restaurateur who stood by the estate through hard times, the private client who comes to help with the harvest: this fabric cannot be automated, and it is precisely what carries an estate through bad years.

Passing it on. Knowing why a particular parcel is pruned differently, why the Rèze from a certain vineyard site deserves replanting: a well-equipped memory helps, but the master remains the one who shows how it is done.

Who keeps the final say?

Composite illustration. An encaveur receives an inquiry from an Asian importer who discovered the estate through its website: a potential order worth twelve thousand francs, with a complete technical file required in English within eight days. AI prepares the file in a single evening, drawing on the cellar log. Reviewing it, the winegrower corrects one point: the text describes the frost-damaged vintage as having "exceptional concentration." True, but incomplete — they add a line about the tiny volumes produced and suggest blending in the following vintage. The importer signs, citing that very candor. (A fictional, composite situation, to be replaced with a real case from a Valais estate once the essay's grounding pass takes place.)

Job description, 2030

Three new competencies will need to appear in the job profile, none of which any current winemaking curriculum covers.

The first is steering the estate's voice: defining, in an instructions file the winegrower controls, the estate's style, its no-go phrases, and its vocabulary; validating every generated text the way one validates a blend. It is an editor's competency applied to one's own signature.

The second is governance of the estate's data: keeping the cellar log, the land registry, and the histories in searchable, transferable formats, and deciding what stays private, what gets shared with the interprofession, and what gets published. The estate's heritage becomes a data heritage as well.

The third is interpreting AI-assisted agronomic signals: cross-referencing alerts from models, sensors, and field observation, while knowing the tools' biases (calibrated on lowland vineyards, rarely on the steep hillside parcels of Valais).

Territorial roots

The fragmentation of the Valaisan vineyard, often described as an economic handicap, becomes in this light a reservoir of signatures: hundreds of estates, each with its own story, its own vineyard sites, its own rare grape varieties to tell, none of which had the means to do so at the level of the major international houses. If every hour saved on administration and communication flows back to the vine, the cellar, or the client relationship, the competitive shift here favors the territory: value moves toward what Valais holds in abundance — the singular, the rooted, the thing no one can relocate. The condition set out in Chapter 7 still applies: an experienced practitioner has to be the one orchestrating; the multiplier effect does not work for an estate that outsources its own voice.

What decision-makers must do now

For an independent vigneron-encaveur

Before the next vintage, set down the estate's voice (style, no-go phrases, vocabulary, the history of each parcel) in a reference document, and test the chain from tasting note to export file on a single cuvée. Measure the time saved, and judge the quality by the same standard as a blend. It is the least costly investment the estate can make this year.

For the Interprofession de la vigne et du vin du Valais

Build a shared corpus on the indigenous grape varieties (reference descriptions, history, validated multilingual terminology) and make it available to encaveurs: pool what is common heritage, leave each estate its own signature. This is the pilot "viticulture" sector that the Alpine Campus action plan (PA-I1) plans to equip.

For the cantonal agriculture service

Examine automated pre-processing of regulatory workflows (transport permits, harvest inspection) on the administrative side: every hour of administrative work saved on both sides of the counter is returned to production, and the AOP traceability framework gains in data quality.

---

¹ Cantonal Agriculture Service, Canton of Valais, vineyard statistics (2024 area: ≈4,600 ha; ≈18,000 owners; ≈73,000 parcels; 62 wine-growing communes). ² 2024 harvest: 34 million kilograms pressed, among the three smallest in fifty years (OFAG data / Swiss press, 2025). ³ Figures carried over from Chapter 7 of the essay, to be sourced with the Interprofession.

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