RSD-008.6 : Field note
Pilots to scale: what this year's robotics rollouts mean for advisors
Commercial robotics spent a decade running pilots. This year a run of them turned into rollouts: a restaurant chain, a car plant, and a global logistics operator each moved a robot from one site to many. This note reads those announcements for the pattern underneath, and lays out what changes for an advisor when the client's question stops being whether it works and becomes how fast it can be rolled out.
Robotics SD : September 1, 2026
The short version
- The 2026 rollouts share a shape: a narrow task with a known labor cost, a pilot that ran long enough to cross a return threshold written down in advance, then a multi-site expansion.
- The robots that scaled are specialists, folding napkins, scrubbing floors, moving totes, sequencing parts, not general-purpose machines.
- Scale is increasingly bought as a subscription: the IFR reported the Robots-as-a-Service fleet grew 31% in 2024, against 9% growth in outright service-robot sales.
- For an advisor the job moves earlier and lasts longer: register before the pilot, help the client write the exit test, and plan the rollout with the supplier's engineers from day one.
Field note
What changed this year
The numbers turned first. In June 2026 the International Federation of Robotics (IFR) reported preliminary figures showing US industrial robot installations rebounded to about 38,000 units in 2025, up 11%, with the food and beverage industry up 30%. In August, A3, the Association for Advancing Automation, reported North American orders of 8,940 robots worth $622 million in the second quarter of 2026, and for the first time in a long while non-automotive buyers accounted for 56% of the units. The demand is broadening beyond car plants, which is where the mid-market lives.
Then the case studies arrived, and they are more useful than the numbers. Dyna Robotics wrote in August that Din Tai Fung, after roughly a year of piloting a napkin-folding robot in its restaurants, is rolling the robots out across its network. Dyna reported the robot went from 35 napkins an hour at a 75% acceptance rate to 95 an hour at 93%, enough to cover a location's daily need in a shift, and it said the pilot had crossed the return-on-investment threshold before the rollout was approved. The company's stated plan is a deployed fleet in the hundreds by the first half of 2027, across every customer it serves.
On the industrial side, BMW said in June that it moved a humanoid robot from an eleven-month body-shop pilot at its Spartanburg plant into a production logistics task with the next model. Agility Robotics, in its June announcement of a plan to go public, disclosed more than $300 million in multi-year orders for its Digit robot, subject to contractual milestones, and more than 65,000 hours of operation across nine customer facilities for names including Amazon, GXO, and Schaeffler. And Standard Bots, after a $200 million round in June that valued it at $1 billion, said it is "on pace to deliver 10% of new U.S. industrial robot deployments by next year." That is the supplier's own claim, not an audited number, but it tells you what the well-funded end of the market is building toward: volume, not demos.
Field note
The shape of a pilot that scales
Read the rollouts side by side and the pattern is hard to miss. Every one of them started with a task narrow enough to measure: napkins folded per hour, totes moved per shift, parts sequenced per line. None of them started with a general-purpose robot looking for a job. The task had a labor cost the operator already knew, so the robot's cost could be weighed against something real from the first week.
Every one of them also ran long enough to be boring. The BMW pilot took eleven months; the restaurant pilot took about a year. A pilot that ends after two impressive weeks proves the demo, not the deployment. What the long pilots produced was the unglamorous evidence a finance team actually needs: uptime across seasons, support response when something broke, and throughput that improved as the supplier tuned the system. Dyna's own numbers, from 35 an hour to 95, are the kind that only exist when someone kept measuring.
McKinsey put a name on the failure mode in October 2025: pilot purgatory, the state where a robot works well enough to keep but never well enough to expand. In our experience the cause is rarely the robot. It is that nobody wrote down, before the pilot started, what result would end it. The deployments that scaled this year had an exit test: a throughput number, a cost per unit of work, an uptime floor. When the pilot hit it, the rollout was already approved in principle.
Field note
Scale is bought as a subscription
The second half of the pattern is how the expansion gets paid for. In its 2025 service-robot report the IFR found that the Robots-as-a-Service fleet grew 31% in 2024, to more than 24,500 units, while sales of professional service robots grew 9%. Fleets are growing faster than purchases, because a subscription is what lets a second site, then a tenth, be approved without a fresh capital request each time.
The largest fleets in ordinary businesses look nothing like the humanoid headlines. Locus Robotics reported in February that DHL Supply Chain had passed one billion picks on its robots across more than 40 sites, with a plan to expand to 5,000 robots. Tennant announced the sale of its ten-thousandth robotic floor scrubber in mid-2025, and Brain Corp, whose software runs many of them, says more than 50,000 robots now operate on its platform. Those are specialists, bought by the fleet, mostly on service terms, doing one task at thousands of sites.
That is exactly the shape our own RaaS note describes: a business subscribes to a working robot instead of buying one, with deployment, support, and updates included, and weighs the subscription against a shift it cannot staff. The 2026 rollouts are that logic applied at scale, and it is the logic an advisor can carry into any client with the same task on its floor.
Field note
What it changes for the advisor
When the client's question was whether robots work, the advisor's job was to get a pilot started. Now that the public record shows pilots turning into rollouts, the job moves earlier and lasts longer. Three things change.
- Register before the pilot, not after. Deal registration records the opportunity against you from the first qualifying conversation. A pilot that succeeds becomes a multi-site rollout, and attribution for the rollout is settled by the registration that predates it.
- Help the client write the exit test. Throughput, cost per unit of work, uptime, support response: the criteria that decide whether the pilot ends in a rollout or in purgatory. Our pre-sales engineers scope the site and help set those numbers before anything ships.
- Plan the rollout with the supplier's engineers from day one. Sites two through ten differ from site one in power, network, floor layout, and staffing. The suppliers who scaled this year had a rollout playbook; the advisor's role is to get the client and that playbook in the same room early.
The money follows the same path it always does: the supplier pays Robotics SD a commission on the revenue a registered deal produces, and Robotics SD pays the advisor from it. A pilot that becomes a rollout produces a great deal more of that revenue than the pilot did, and the registration filed at the start is what keeps all of it attributed to the advisor who opened the conversation.
Field note
Where the mid-market comes in
The rollouts in the headlines belong to large operators: a national restaurant chain, a global automaker, a logistics giant. Their pilots absorbed the risk and produced the playbooks. The same tasks exist at regional restaurant groups, mid-size warehouses, cleaning contractors, and single-plant manufacturers, and the suppliers that just proved a rollout at scale now need the next tier of customers to sell it to. A direct sales team cannot reach that tier economically. Advisors already hold the relationships.
That is the conversation to start now, and our use-case sheets give you the one-page version for nine industries. Bring the client the proof from 2026, register the opportunity, and let the pilot be designed to end the right way.
FAQ
The questions that follow
What does pilot purgatory mean in robotics?
The phrase, used by McKinsey in an October 2025 analysis of humanoid robots, describes a pilot that works well enough to keep running but never well enough to expand. The usual cause is that no one defined, before the pilot started, what result would end it. The fix is an exit test written up front: a throughput number, a cost per unit of work, an uptime floor, and the support response the client expects.
How long should a robotics pilot run?
Long enough to hit the exit criteria across normal operating conditions, and no longer. The public 2026 rollouts followed pilots of roughly a year, because a car plant and a restaurant chain needed evidence across seasons and shifts. A simpler task, such as floor scrubbing on a known route, can settle far faster. The length matters less than having an agreed test that ends it.
Should an advisor register a deal before or after the pilot?
Before. Deal registration records the opportunity against the advisor from the first qualifying conversation, well before a supplier is engaged. If the pilot becomes a multi-site rollout, attribution for the rollout rests on the registration that predates it. Registration costs nothing and commits nobody.
Does the advisor earn a commission on the pilot or on the rollout?
Commissions follow revenue: the supplier pays Robotics SD on the revenue a registered deal produces, and Robotics SD pays the advisor from it. The exact structure is worked out at supplier onboarding. In practice the rollout is where most of the revenue sits, which is why registering before the pilot matters.
Next step
You already know the client for this.
Start one conversation, register the deal, and Robotics SD engineers take it from there. Joining as a Robotics Advisor is free.