The Robot Question Is Not Whether to Automate. It’s What to Automate First.

Insights from the Scaling Business Summit 2026, Ho Chi Minh City.

The stage had no sofas. No gentle warm-up questions. Kien Nguyen, Business Development Director at Gradion, took the podium with a stack of provocations and a one-minute timer. The format was deliberate: throw a thesis, let the panel fight it out.

On stage: David Will, General Director at DRÄXLMAIER; Nguyen Hong Phuc, Director of Digitalization Strategy at Schaeffler Asia Pacific; Huynh Phong Phu, General Director at KUKA Vietnam; and Dr. Lennart Bochmann, Co-Founder and CPO at SYNAOS. Four people who build, buy, deploy, and manage automation at scale with no prepared answers and no filter.

What emerged was not a debate about whether automation matters. Everyone in the room already knew it does. The real argument was sharper: what to automate, when, how to avoid the traps, and what the factory floor actually looks like when you get it right.

1. Stop Choosing Between People and Robots

The first thesis landed fast: Manual labor scales faster. You can hire 50 people tomorrow. A robot takes six months to arrive. True or false?

David Will did not take the bait. “Why do we have to choose A or B? Why do we not combine it? We optimize the financial report.” That framing combination over substitution ran through the entire session.

Huynh Phong Phu pointed out the reality that any manufacturer in Vietnam already knows: after every Lunar New Year, factories face the same crisis of workers not returning from their hometowns. The labor flexibility argument has a very specific expiry date. And it comes around once a year.

The more sophisticated model emerging is Robot as a Service (RaaS) - leasing mobile robots the same way you pay workers monthly, with maintenance, upgrades, and support included. It changes the risk calculus entirely. “You lease the robot as if you are paying your workers every month.” David Will added his own line on the worker side: “Leasing workers kills quality.” Owned relationships, trained people, and consistent teams are the standard he holds.

Lesson 1: The automation decision is not a binary. The question is not robots or people. It is which tasks, at what scale, and with what ownership model get the best result.

2. Fix the Process Before You Automate It

One thesis stopped the room: Automating a bad process just speeds up the production of waste.

Every panelist agreed. The reference point was Toyota's Total Productive Maintenance methodology, a system built on the principle that when an error occurs, you stop the entire production line. Not to slow things down. To prevent the waste from compounding. The discipline of stopping is what makes the process clean enough to automate.

Dr. Lennart Bochmann made the implication explicit: “If you think about automating things which have been done manually before, you definitely need to rethink your processes. Depending on what kind of equipment you are going to use, there will be a change and you have to consider this when you plan a project.”

The practical failure mode is common. A company identifies a bottleneck, buys automation, and deploys it on top of a process that was already inefficient. The robot runs perfectly. The waste just moves faster. The audit that should happen before the purchase order, mapping the process, identifying the root causes, eliminating unnecessary steps is the work most teams skip.

Lesson 2: Automation amplifies what is already there. If the process is broken, the robot will break it faster. Fix the workflow before you write the purchase order.

3. The Software Trap Nobody Warns You About

Thesis: Proprietary software traps you. You lose the freedom to choose best-in-class hardware.

This one was personal. David Will described it as a constant internal battle. The issue is not just being locked into a single vendor. It is that the software provider uses the integration point between software and hardware to charge above-market prices for the hardware itself. “It is a constant discussion with our central purchasing department who wants to keep it all together and keep it nice and simple.”

Dr. Bochmann's answer from the SYNAOS side was direct: build for interoperability from day one. Starting with one or two robots feels manageable with proprietary integration. But as the fleet grows and use cases overlap, the technical debt compounds. “Really try to stay independent from the beginning,” he said. “Once you have done this, scaling is really, really much faster than if you do proprietary APIs and integration every time again.”

The counter-argument came from the factory floor: “If it runs well, don't touch it. It still produces money for you.” Not every system needs to be best-in-class. Some need to be stable and profitable. The discipline is knowing which is which and not making the vendor lock-in decision by default.

Lesson 3: Vendor lock-in is a strategic decision, not a procurement detail. Decide your interoperability standard before you deploy the first robot, not after you have built a legacy you cannot escape.

4. Two Brains for the Factory Floor

One of the sharpest exchanges of the session came from a thesis about AI: Factories need boring, deterministic reliability, not creative AI decision-making.

Dr. Bochmann disagreed with the framing, but not the instinct. His answer separated two layers that most companies treat as one: “Operations needs to be really, really deterministic and reliable to have the highest output. But the planning behind it needs to be as dynamic and fast as possible because things change.”

The shop floor itself must be predictable. Robots arrive on time. Routes are optimized. Outputs are consistent. But the system coordinating all of that needs to adapt in real time: a robot arriving late, a driver taking a break, an order surge at 2am. Static planning cannot handle a dynamic environment. AI handles the planning. Determinism handles the execution.

Nguyen Hong Phuc put it simply: “We consider AI just like a tool. When you need support, you use the tool for that.” And on whether AI will ever be a team member: “A team member should have emotion. A machine is a tool.”

Lesson 4: Do not ask AI to run the factory. Ask it to plan the factory. Operations needs reliability. Planning needs adaptability. These are different jobs.

5. The Honest Truth About Lights-Out Manufacturing

The final thesis was the boldest: Lights-out manufacturing, fully autonomous, human-free factories is a marketing myth, not a realistic business goal.

Agreement was quick, but it came with an asterisk. Kien Nguyen dropped it quietly: there is a dark factory operating 35 kilometers from the conference venue, built 12 years ago. It exists. It works. But it has never returned its investment.

Dr. Bochmann drew the more accurate picture of where the industry is heading: “The role of the worker is just changing from being operative or doing operative tasks to supervision and expert-level tasks.” Warehouses are already approaching lights-out in specific configurations. Manufacturing will follow, sector by sector, as the economics shift and the tooling matures. The question is not whether it will happen. It is whether the timing and the investment thesis are right for your specific operation, right now.

David Will pointed to maintenance as the real blocker: more automation means more failure points. You do not eliminate the human. You move them up the skill curve. The factories that win in the next decade are the ones that start building that workforce now, not the ones waiting for a fully autonomous future that replaces the need entirely.

Lesson 5: Lights-out manufacturing is not a myth. It is a timing question. The factories winning right now are not fully autonomous. They are systematically moving human work up the skill curve.

The CEO Execution Playbook: What to Do Tomorrow

1. 1. Map your process before your next automation purchase. Identify every step in the target workflow. Remove the waste first. The robot should inherit a clean process, not an optimized broken one.
2. 2. Evaluate RaaS models for your next deployment. If CapEx is a barrier, Robot as a Service is now a viable option. Model both scenarios before signing a purchase order.
3. 3. Audit your current software stack for vendor lock-in. Ask your automation team: if we needed to swap hardware vendors in 18 months, what would that cost us? The answer will tell you how exposed you are.
4. 4. Separate your AI investment into two buckets. Reliability and consistency on the operations layer. Speed and adaptability on the planning layer. Evaluate each against the right criteria.
5. 5. Decide now what role humans play in your five-year factory. Not headcount reduction, role redefinition. The companies building expert-level human capacity alongside automation today will have a structural advantage when the tooling matures.

→ Watch the full session on YouTube