# The Pragmatic Middle Ground: Why Hybrid Autonomy Gets SMEs Further Faster

> Author: Chris Jon Graf (AI Strategist & CEO)
> Updated: 2026-08-24
> URL: https://ai-outsourcing.ch/insights/the-pragmatic-middle-ground-why-hybrid-autonomy-gets-smes-further-faster

## Summary

Hybrid autonomy – one human lead plus autonomous followers – is already live in China's logistics: 45,000 kilometres driven, 29% lower freight costs, without waiting for perfect full autonomy. The reason for this success: the pragmatic middle ground uses what's technically mature today, gathers data for tomorrow, and avoids the trap of endless pilot projects. For Swiss SMEs, the same logic applies: don't automate everything at once; capture real value step by step – and move faster than those waiting for the infallible system.

## Why China's logistics runs the 1+N model – and what SMEs can learn

Since early 2025, China's highways have been running convoys that show how automation actually works: one human steers the lead truck, followed by up to four driverless trucks in formation – autonomously guided, but not left alone. Pony.ai, KargoBot and SANY have deployed this 1+N platooning on routes longer than 400 kilometres, now covering 45,000 kilometres and transporting nearly 500 TEU of freight. The result: 29% lower cost per kilometre, 195% higher profit per vehicle, 80% less personnel on predictable routes.

The trick isn't the tech alone – lidar, cameras, V2V communication exist elsewhere too. The trick is the decision not to replace the human, but to position them smartly: in the lead truck, handling construction zones, chaotic toll plazas and all the long-tail scenarios that still overwhelm autonomy today. The follower trucks handle the standard scenario – straight ahead, lane-keeping, distance control – with L4 precision. The system isn't an end state, but the economically sensible intermediate step: it pays off today while gathering the data that enables the next step.

## The economic core: why the hybrid reaches profitability sooner

Fully autonomous systems promise maximum efficiency eventually – but eventually costs money, time and nerves. Hybrid models, however, lower deployment risk immediately: Accenture and Logistics Viewpoints confirm that onboard automation with remote intervention on demand builds trust and cuts operating costs sooner. Even in the most pessimistic scenario, autonomous trucks save between 1,415 and 2,222 US dollars per vehicle per month – and that's with systems that aren't yet perfect.

**29%** — cost reduction per kilometre with 1+4 platooning (SANY/Pony.ai)

KargoBot has now clocked over 35 million kilometres with L4 fleets and moved 1.2 billion tonne-kilometres. The message is clear: logistics is where autonomy pays off today – not in perfect full-autonomy, but in smart hybrids that deploy humans where they're unbeatable and machines where they perform reliably.

## The parallel to Swiss SMEs: end pilot paralysis, scale value

Many Swiss companies are stuck in the pilot trap: they test AI in isolated projects, gather insights, celebrate internal wins – and never scale. The reason is often an all or nothing attitude: either the system runs fully automatically and flawlessly, or we wait. Hybrid autonomy flips this logic: you start where the value is tangible, keep the human in the loop where judgement is needed, and expand step by step. This approach avoids endless waiting for the perfect solution and delivers real value while you learn.

> **The pilot project trap**
>
> Pilots that never scale cost more than they deliver. The hybrid approach forces you to think productively from day one: what can run today? Where does the human sensibly remain? How do I gather data for the next step?

The Swiss AI podcast has analysed this phenomenon precisely: Ralf Blaschke from Accenture explains why AI pilot projects are dying in 2026 and what companies must do instead to capture real business value. The core message: scaling isn't a later step; it must be designed in from the start.

## Middle-mile autonomy as blueprint: where hybrid systems run today

Why does the 1+N model work so well in logistics? Because the middle mile – the predictable long haul between two hubs – is the ideal deployment ground for autonomy: limited variance, high repetition, clear parameters. First mile (collection) and last mile (delivery) remain human, because chaos, narrow streets and spontaneous decisions dominate there. Transfer-hub networks, as described by ArXiv 2023, use exactly this split: autonomous on the route, human at the edges.

- Identify the middle mile in your processes: where is the flow predictable, repetitive, clearly structured?
- Automate there, keep the human at the interfaces where judgement, context and flexibility are needed.
- Gather data in live operation: every hybrid run shows you where the machine performs stably – and where it doesn't yet.
- Expand incrementally: when the machine reliably masters a new situation, give it more responsibility.

DeepWay in Hefei completed a 200-kilometre route with unmanned followers – on national highways, not in a lab. The system isn't perfect, but it's productive. And that's exactly the point: productivity beats perfection when you're learning as you go.

## Why bad processes don't improve through AI

Another reason hybrid systems succeed: they force you to clarify the process first. If you don't know what the human should do and what the machine should do, neither will work. Martin Jäger explained in the AI podcast why AI is a leadership matter: bad structures don't improve through AI – first comes the process, then the tool.

Hybrid autonomy is therefore also an organisational test: if you can't document the process cleanly enough to define where the machine takes over, your problem isn't missing AI – it's missing clarity. The advantage: this test costs you nothing but thinking time, and it shows you where you truly stand.

## Three principles for the hybrid entry in your SME

> **Start lean, scale deliberately**
>
> You don't need a perfect system to start. Automate a clearly bounded sub-process, keep the human in the loop where judgement is needed, and expand as soon as the machine performs reliably.

First: define the lead and the followers in your process. Where is human judgement, context, empathy needed? That stays with the human. Where is the flow repetitive, rule-based, predictable? That goes to the machine. Second: accept that the machine can't do everything – and plan for it. The human in the lead truck isn't a stopgap, but part of the design. Third: measure, learn, expand. Every run shows you where the boundary between human and machine can be shifted.

1. Identify a process with clear repetition (data entry, categorisation, standard inquiries, report generation).
2. Automate the standard case, keep the human for exceptions, context, strategic decisions.
3. Gather feedback: where does the machine run stably? Where does it still need intervention? Expand autonomy where it proves itself.

This approach isn't glamorous, but it works. And it delivers something pilots never do: real, measurable value from day one.

## What distinguishes hybrid autonomy from full automation

Full automation promises maximum efficiency – eventually. Hybrid autonomy promises less, but delivers sooner. The difference lies not only in the tech, but in the mindset: full automation treats the human as a problem to be eliminated. Hybrid autonomy treats the human as an asset to be deployed smartly. In a world where tech is developed faster than it's understood, the latter is the more realistic path.

> The most successful path to automation isn't fully autonomous or nothing at all
>
> — 1+N principle from China's logistics

China's 1+N convoys show that the pragmatic middle ground is not only faster, but also more economical. You don't have to wait until the machine is perfect – you can start today where it's good enough, and gather the data that makes it better as you go.

## The next step: from knowing to doing

You now understand why hybrid autonomy works. You see the numbers, the logic, the examples. The next step is the question: where in your company is the middle mile – the process that's clear enough to be partially automated, and valuable enough to justify the effort? Answering this question isn't trivial. It demands process clarity, technical understanding and the ability to estimate value realistically. This is precisely where general advice ends – and tailored strategy begins.

## FAQ

### What is hybrid autonomy and why is it more pragmatic than full automation?

Hybrid autonomy combines human control at critical points with machine automation in standard cases. It delivers measurable value immediately, lowers risk and gathers data for the next step – instead of waiting for perfect full automation that comes later and costs more.

### What does the 1+N model from China's logistics mean concretely?

One human driver steers the lead truck, followed by up to four driverless trucks autonomously in formation. The human handles complex scenarios (construction, chaos), the machines take over the predictable part. Result: 29% lower costs, 80% less personnel on the route, productive from day one.

### Why do so many AI pilot projects fail in Swiss SMEs?

Because they bet on all or nothing: either the system runs perfectly, or you wait. Pilots remain isolated, aren't scaled, deliver no business value. Hybrid approaches start where value is tangible and expand step by step – with real ROI from the start.

### How do I find the middle mile in my company?

Look for processes that are repetitive, rule-based and predictable: data entry, categorisation, standard inquiries, report generation. Where little context or judgement is needed, that's your middle mile – and thus your first automation candidate.

### What role does the human play in hybrid systems long-term?

The human remains where judgement, context, empathy and strategic decisions are needed. Hybrid autonomy treats the human not as a problem, but as an asset – and shifts the boundary between human and machine only where the machine performs reliably.

## Sources

- [China Daily: Autonomous truck convoys put Ordos at smart logistics forefront (Juni 2026)](http://regional.chinadaily.com.cn/ordosen/en/2026-06/09/c_1189409.htm)
- [Pony.ai: First Company in China Approved for Autonomous Truck Platooning Tests (Januar 2025)](https://ir.pony.ai/news-releases/news-release-details/pony-ai-inc-becomes-first-company-china-approved-autonomous)
- [SANY/Pony.ai: Mass-Production Readiness of Fourth-Generation Autonomous Heavy-Duty Truck (2026)](https://www.sanyglobal.com/press_releases/4826/)
- [Accenture: Making Self-Funding Supply Chains Real (2026)](https://www.accenture.com/content/dam/accenture/final/accenture-com/document-4/Making-Self-Funding-Supply-Chains-Real-Report-Final.pdf)
- [Logistics Viewpoints: Autonomous Trucking Is Fragmenting Into Distinct Market Entry Models (April 2026)](https://logisticsviewpoints.com/2026/04/14/autonomous-trucking-is-fragmenting-into-distinct-market-entry-models/)
- [MDPI: Cost-Effectiveness of Introducing Autonomous Trucks (September 2023)](https://www.mdpi.com/2076-3417/13/18/10467)
- [ArXiv: Optimizing Autonomous Transfer Hub Networks (Mai 2023)](https://arxiv.org/html/2305.03119)