Lawyer

The typical Valais law firm is small, regional, generalist; it lives off billed hours. AI is not going to plead in its place: it will erode the hours that used to fund it and shift value toward what the lawyer never quite knew how to bill for, strategy and judgment.

The trade today

The Valais bar mirrors the economic fabric it serves: firms of two to five lawyers, generalist by necessity, spread across the valley towns, working for SMEs, private individuals, communes, and bourgeoisies, in the canton's two languages. A lease and a divorce get handled in the morning, construction law in the afternoon. Narrow specialization remains the exception.

A firm's daily work spans a wide range:

• Consultation and advice: initial meetings, legal opinions, dispute prevention
• Litigation: pleadings, hearings before the district courts and the Cantonal Court, appeals to the Federal Supreme Court
• Drafting: contracts, agreements, articles of association, wills, settlements
• Legal research: federal and cantonal case law, doctrine, administrative practice
• Procedure: deadlines, notifications, exhibit lists, correspondence with authorities and fellow lawyers
• Negotiation: out-of-court settlements, talks, mediation
• Running the firm: time-based billing, timesheets, training articling clerks

A substantial share of billed hours comes from intermediate tasks: research, first drafts, file summaries. Chapter 10 describes their erosion over five to ten years; the bar is a textbook case of it.

What AI is preparing

Legal research. What used to take hours, sweeping through the ATF (Federal Supreme Court rulings), published cantonal case law, doctrine, now takes minutes, with argued summaries and supporting references. The trade-off is absolute: every reference must be checked against the source. A hallucinated ruling cited in a pleading constitutes professional misconduct, and foreign courts have already sanctioned lawyers for invented rulings filed as submitted. Verification becomes a professional act in its own right, billable as such.

Pre-drafted work. Standard contracts, procedural letters, template agreements, first drafts of pleadings: prepared from the file and the firm's templates, then reviewed and signed by the lawyer of record. The signature keeps its full weight. It is what the client pays for, along with the responsibility that comes with it.

The firm's economics. This is where the real shock plays out. If research and first drafts compress, so do the hours that used to bill for them. A firm that sells time watches its base shrink; one that sells outcomes (a settlement reached, a risk averted, an estate settled without rupture) converts the productivity gain into margin or volume. Moving from hourly rates to flat fees becomes a decision about survival, on the timeline chapter 10 sets out: five to ten years.

Access to complex cases. The movement also plays out as conquest. A heavy construction dispute, a public procurement matter, a restructuring: yesterday, the Valais SME went looking for a large firm on Lake Geneva or in Zurich. A well-equipped small regional firm can now handle these matters with comparable depth of research, adding a knowledge of cantonal practice, the courts, and the local terrain that cannot be imported.

Case data: the prerequisite

Lawyers are fully bound by professional secrecy under Art. 321 of the Swiss Criminal Code, whose violation is a criminal offense prosecuted on complaint. Feeding a client file into a consumer-grade tool whose servers, jurisdiction, and training practices escape all control exposes the lawyer to that violation. Three requirements precede any deployment: tools operated under Swiss law or offering equivalent contractual guarantees (hosting, documented subcontracting, no reuse of data for training); compliance with the nFADP, in force since September 1, 2023, for the data of clients and opposing parties alike; and clear information for the client about what is being processed, where, and by what. The practical sovereignty discussed in chapter 10 takes its clearest form here: for the profession entrusted with other people's secrets, the jurisdiction of the tools is a matter of professional ethics.

What rises in importance for judgment

Verification that carries liability. Reviewing generated research or a pleading with the eye of the person who will sign it: checking every reference, spotting the invented ruling, the real ruling pulled out of context, the plausible but wrong argument. This skill is distinct from research itself. It is becoming one of the first things taught to articling clerks.

Strategy. Settle or litigate, act now or let the deadline run, set aside a provision or contest: these judgment calls cross the law with knowledge of the courts, the opposing party, and the client's capacity to sustain a three-year proceeding. AI illuminates the scenarios; the lawyer chooses the one to pursue, and answers for it.

Pleading and negotiation. The hearing, the confrontation, the face-to-face remain entirely human exercises: persuading a judge, sensing the moment the opposing party is ready to give ground, framing the offer that unblocks a stalemate. None of it runs through a keyboard.

The relationship of trust. Divorce, inheritance, a dispute between business partners: the client arrives with a legal problem embedded in a human story. Listening, reframing, sometimes advising against the very proceeding the client is demanding: counseling a client to forgo a winnable but destructive lawsuit ranks among the profession's most valuable acts, and among the least visible on a timesheet.

Training the next generation. The tasks through which an articling clerk used to learn the trade (research, first drafts, summaries) are precisely the ones being compressed. Deciding what a young jurist should still do by hand, in order to learn, becomes an explicit pedagogical choice for training supervisors. That question had never been posed in these terms before.

Who keeps the final say?

Composite illustration. A three-lawyer generalist firm takes on a construction dispute the client had originally intended for a large out-of-canton firm. AI-assisted research develops, within two days, a line of argument grounded in recent cantonal case law; the partner checks every reference and discards one, plausible but drawn from a context that does not apply. What remains is the real decision: litigate or settle. She knows the opposing company, its order book, its need to close the dispute before winter, and proposes a settlement the legal file alone would never have suggested. The matter is resolved within three months. (A fictional, composite situation, to be replaced by a real case during the incarnation pass.)

Job profile 2030

Three skills will be added to those covered by the bar exam.

The first is verification of sources that carries liability: systematically checking every reference produced, tracking what was verified and how, turning this discipline into a courtroom advantage (a firm that documents its verification pleads with more confidence than one that cites on faith).

The second is economic management of the case: building flat-fee or outcome-based offers, costing a matter in scenarios, explaining to the client what they are paying for when research is no longer counted in hours. Billing becomes a skill in its own right, on a par with procedure.

The third is governance of tool-assisted confidentiality: choosing, configuring, and monitoring the firm's tools against Art. 321 of the Criminal Code and the nFADP, documenting data flows, informing clients. In a small firm without a dedicated IT manager, this responsibility falls explicitly to one of the partners.

Territorial anchoring

Maintaining a dense regional bar is a matter of access to justice. The valley SME, the commune, the family settling an estate all find, within an hour's drive, counsel who knows cantonal practice, the district courts, and the local fabric, in both of the canton's languages. The erosion of intermediate tasks first threatens the firms that lived off them; the competitive shift offers those same firms a chance to keep complex cases in the canton that used to leave it. Between these two trajectories, the difference will come down to equipment, billing model, and training. Nothing that depends on size.

What decision-makers must do now

For a firm partner

Before the end of 2026, measure the share of billed hours built on research and standard drafting, then settle on a three-year billing trajectory. In parallel, put in place a written verification rule (who checks what, with what record) and choose tools compatible with Art. 321 of the Criminal Code: both efforts protect the firm's liability while its business model transforms.

For the Valais Bar Association

Issue cantonal guidelines on the use of generative tools: conditions for compatibility with professional secrecy, the duty to inform clients, responsibility for generated content. And champion a "bar" track at the alpine campus (the cantonal training initiative proposed under action plan PA-I1), with practical exercises on fictional cases: a shared learning curve costs less than ten individual mistakes before the supervisory authority.

For the cantonal bar supervisory authority and the courts

Publish expectations early: what an AI-prepared pleading must satisfy, how the duty of verification will be assessed, what practices would constitute a breach of professional secrecy. A practice announced in advance spares the bar the disciplinary case law it would otherwise absorb after the fact, and gives it the security it needs to equip itself without waiting for the first sanction.

---

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.