Up on the Hönggerberg, ETH Zurich’s Robotic Fabrication Lab spent the better part of fifteen years dragging construction robots from file-to-factory to file-to-site — the shift Gramazio and Kohler document in The Robotic Touch, where the machine stops sitting in a clean prefab hall and starts laying real material in the dirt. We read that book the way most small offices did: as a beautiful research programme that had nothing to do with next Tuesday’s deadline. This week a lift company quietly proved us half wrong.

The signal: Schindler — the Swiss group out of Ebikon, near Lucerne — has grown its fleet of R.I.S.E self-climbing elevator-shaft robots to seven units worldwide, per Construction Management‘s Digital Construction desk. Launched five years ago, the robots have now worked 36 job sites and set roughly 50,000 anchor bolts — the unglamorous structural fixings that carry landing doors, divider beams and wall brackets — across Austria, Poland, India, the UK, Brazil, China and Singapore. Schindler claims up to 40% time saving in the shaft, and a specialist operator drives it from a remote panel. Elevatori Magazine adds the detail that matters to a practice: these machines aren’t sold, they’re dispatched from hubs in Switzerland, Hong Kong and Australia.

That last point is the real story, and it isn’t about robotics at all. It’s a fleet-and-hub model — the same logic as a crane on hire, not a plugin on your licence. Nobody is asking a 14-person studio to buy a self-climbing robot; they’re asking the general contractor to book a slot. Which means the question lands back on our side of the desk in a familiar shape: who owns the model the robot reads?

←TODAY: Seven R.I.S.E robots, 36 sites, ~50,000 anchor bolts, dispatched from three hubs. →3012: The building delivers its own setting-out; human hands move up the value chain, not out of work. Fulcrum: The robot only earns its 40% if the shaft it climbs was modelled cleanly enough to trust — automation rewards the office that already had its coordination in order.

Because a self-climbing robot drilling anchor positions is only as good as the setting-out it’s fed. On the projects we’ve coordinated, the lift shaft is exactly where the model goes soft: a generic Schacht from the architect, a structural opening from the engineer, the actual landing-door geometry living in a supplier’s PDF nobody re-modelled. The week we finally trusted our IFC round-trip on a mixed-trade Wettbewerb was the week we stopped treating the shaft as a void and started treating it as a coordinated assembly with real Z-heights for every bracket.

Atelier: Treat the lift core as a federated sub-model, not a hole — pull the shaft into a Speckle stream or a clean IFC4 export with the landing-door datums, divider beams and bracket positions as actual geometry, so the contractor’s robot operator is reading your coordinated heights and not re-surveying on site. The 40% Schindler advertises is only banked downstream if the upstream model didn’t lie.

Hack: This Hack teaches you to size a robot’s payoff per shaft before you sign the rental slot, so the decision is a number and not a vibe. The medium is one SQL query against your quantity take-off; the domain is Databases. Assume your take-off table holds an anchor count per shaft and a manual rate of ~4.5 min/bolt:

SELECT shaft_id,
       anchor_count,
       ROUND(anchor_count * 4.5 / 60.0, 1)        AS manual_hours,
       ROUND(anchor_count * 4.5 * 0.40 / 60.0, 1) AS hours_saved
FROM   elevator_shafts
WHERE  project_id = 'VICTORIA-105'
ORDER  BY hours_saved DESC;

Run it once and the shaft with 900 bolts that justifies the robot separates instantly from the two-stop core where it never will.

The honest trade-off: the fleet model spares you the capex and the maintenance headache, but it also means the robot leaves with the hub van and takes none of its competence into your office — the repeatability stays with Schindler, not with you. The practices that got hurt over the past decades weren’t the late adopters; they were the offices that rented a different clever thing every project and never wrote down how their own model fed it, so the knowledge walked out with each Praktikant. Skanska, the first UK contractor to run R.I.S.E at 105 Victoria Street in London, gets a robot for a season; what it keeps is whatever it documented about its own shaft model.

PAZ Takeaway: The capability worth building here isn’t the robot — it’s a reusable lift-core detail kit and a disciplined Grasshopper↔Archicad shaft export, the exact gap the PAZ Grasshopper↔Archicad Library exists to fill, so that whoever’s robot or crew turns up next year is reading geometry your office still owns and can repeat.

So before the next contractor offers you a shaft robot, do the cheaper thing first: model your lift core as if a machine were going to read it — because soon enough, one will.

Source: Construction Management — Digital Construction