Consulting engineer

The Valais consulting engineering firm sizes what holds the territory together: structures, avalanche barriers, networks, turbines. AI will not sign the drawings: it will generate the calculations and the variants, and concentrate the engineer's value on what the standard leaves unsaid.

The profession today

The canton lives with its engineering firms like few others. Civil engineers, geotechnicians, hydraulic specialists, HVAC and electrical consultancies: mostly small or mid-sized outfits, rooted in the valley towns, whose order books tell the story of the territory. Protection against natural hazards (avalanches, debris flows, floods), watercourse management, mountain roads, land-use planning, and hydropower, the canton's electrical backbone: few regions offer an engineer terrain this demanding, and this unforgiving.

A firm's daily work spans the entire project chain:

• Preliminary studies and variants: feasibility, sketches, comparison of solutions
• Sizing: structural, hydraulic and energy calculations, pipework to SIA standards
• Drawings and reports: permit files, technical reports, impact statements
• Tenders: bids, specifications, offer analysis
• Site supervision: works management, inspections, quantity surveys, amendments
• Expert assessments: natural hazards, structural pathology, damage claims

What AI is preparing

Calculations and variants. Generating five sizing variants where the study budget once allowed only two, pre-calculating cross-sections, exploring geometries, optimising a structure: exploratory work is shifting scale. The signature, though, is not changing hands. SIA standards place the responsibility for sizing on the engineer who validates it; a generated calculation remains a proposal until a professional has checked its assumptions, its support conditions and its load cases.

Tenders and reports. Bid files, technical reports, impact statements, site minutes: a firm's documentary output is massive and largely standardised. On this kind of task, the author observes a productivity factor of four to five among his clients. A very concrete consequence: the small firm that used to give up bidding for lack of hands to produce the file regains access to public tenders.

The firm's technical memory. Decades of projects lie dormant in binders and on servers: geotechnical reports, calculation notes, site photographs, damage files. Indexed and searchable, this memory becomes the differentiating asset of the regional firm. A new hire can ask what the firm knows about a given slope, structure or foundation type, and get in minutes what only the senior partner used to be able to retrieve.

The competitive rebalancing. Large urban firms fielded production teams that regional outfits could not match. As documentary output and pre-calculation compress, the scale advantage narrows; what remains is knowledge of the terrain, proximity to the client, and responsiveness, where the local firm moves ahead. Chapter 10 describes this shift running both ways: large firms, equally well-equipped, will move down into alpine mandates too. The window to get equipped is narrow.

Structural data: the prerequisite

A consulting firm handles data that goes well beyond the revised Federal Data Protection Act (nFADP, in force since 1 September 2023): drawings of hydropower plants and critical infrastructure, data belonging to public clients, natural hazard reports whose premature disclosure could shake a property market or a communal assembly. Three requirements before any deployment: hosting and tools kept under Swiss jurisdiction for anything touching critical infrastructure; confidentiality clauses explicitly extended to AI processing, so the client knows what is being processed, where, and by what; traceability of assumptions (calculation version, input data, validator), since the evidentiary value of a structural file can be tested decades after it was produced.

What judgment increasingly requires

Responsibility for sizing. Checking a generated calculation requires understanding what the tool assumed: support conditions, load combinations, safety factors. A gross error is visible. A silently optimistic assumption has to be hunted down. The signature, in the sense the SIA standards give it, becomes the central act of the profession, and it requires engineers capable of retracing the calculation, at least mentally.

Alpine terrain judgment. The standard sizes a wall; it stays silent on the slope that will have to carry it. Water seepage that reappears after two wet winters, moraine that creeps, a torrent that carries debris differently since a hillside fire: this knowledge comes from job sites, from damage claims and from old-timers, never from the training data of a model fed on lowland structures. On an unstable slope, local experience quietly corrects the standard.

Weighing the variants. Five costed variants call for a decision: cost, residual risk, durability, thirty-year maintenance, acceptability to the commune and to property owners. The tool widens the range; the engineer carries the choice before the client and defends it, sometimes at a public assembly.

Explaining risk. Telling a communal council that a structure protects up to a certain event, and no further, without jargon and without false reassurance, remains an entirely human exercise. Public clients' trust is built there, project after project, flood after flood.

Passing on the eye. A young engineer used to learn by doing the calculations, getting them wrong, starting over. If the machine produces correct calculations on the first try, the firm has to organise the training of doubt differently: site visits, pedagogical counter-calculations, damage-claim inspections commented on by seniors. Succession is already the top concern of Valais firms; training judgment becomes its core.

Who keeps the final word?

Composite illustration. A regional firm is studying a protective structure at the foot of an unstable slope. The generated variants are all compliant with standards; the cheapest relies on anchors whose calculation note assumes homogeneous ground. The engineer who has followed this sector for fifteen years remembers water seepage on a neighbouring site and finds, in the firm's memory, an old geotechnical report that mentioned it. He orders two additional soundings: the ground is layered, and the anchors would have held only on paper. The chosen variant costs more, and will rest easy. (Fictional, composite situation; to be replaced with a real case during the embodiment pass.)

Job profile 2030

Three competencies will be added to the diploma and the register listing.

The first is validating generated calculations: reconstructing the assumptions behind a tool-produced sizing, spotting its blind spots, documenting the check before signing. Schools train engineers to calculate; they will also need to train them to counter-calculate.

The second is technical-memory engineering: structuring the firm's archives, tracing assumptions and lessons learned, turning thirty years of projects into a daily working tool rather than a stack of binders. In a canton where every valley has its own history of events, this memory is worth a fortune.

The third is augmented project leadership: orchestrating generated variants, tenders and reports while keeping a hand on the decisions, and explaining to the client what comes from the tool and what comes from the engineer. Transparency about method will become an argument for trust.

Territorial anchoring

The alpine territory is both client and school for Valais firms: natural hazards, mountain planning and hydropower form an order book that climate change makes heavier year after year. Keeping this engineering expertise within the canton is a matter of territorial autonomy: a slope is better watched from the district than from an urban tower, and the experience of past events transfers poorly through a one-off mandate. The compression of documentary tasks gives regional firms the means to hold their ground, to bid more widely and to retain mandates that used to drift toward the major centres. It also makes the canton more attractive to young engineers: fast-tracked responsibility, terrain twenty minutes from the office, tools on a par with those of large firms.

What decision-makers must do now

For a partner in a consulting engineering firm

Before the end of 2026, open two fronts: indexing the firm's technical memory (projects, reports, damage claims) and a written rule for validating generated calculations (who checks what, with what record). Train senior engineers first: they are the ones who know what a calculation note ought to be worth, and it is their judgment that needs capturing before retirements.

For SIA, Valais section

Clarify with the central body the doctrine of use: what signing an AI-assisted sizing commits to, what traceability requirements apply, where the limits lie. And champion a "consulting engineering" track within the alpine campus (the cantonal training scheme proposed under action plan PA-I1): validating generated calculations, governance of structural data, case studies on fictional projects, pooling across firms what none can afford alone.

For the canton's construction departments (roads, natural hazards, energy)

Publish clear requirements for AI-prepared files in public procurement: traceability of calculations, signatory responsibility, deliverable formats. And adjust award criteria so the new documentary speed does not degenerate into a race for the thickest file: the canton, the territory's largest client, will effectively set the practice.

---

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.