JAPAN VIDEO REVIEW · UNMANNED NIGHT

The 24-Hour Unmanned Factory That "Profits on a Single Piece": A Guide to the Kyoto HILLTOP Tour Video

Q: Eight in ten of HILLTOP's orders are just one or two pieces—how can it possibly go unmanned?

Because what it makes unmanned isn't the 'coping with variation' part, but the 'execution' part. Variation (new drawings, new programs) is handled by the daytime digital production system and people; at night the machines only execute already-verified programs. Separate preparation from execution, and drive preparation cost down with a system, and high-mix low-volume and unmanned operation are no longer contradictory.

Q: For a Taiwan shop that wants to learn from HILLTOP, what's the first step?

Not buying equipment, but turning the 'drawing-to-program' knowledge into software: getting the drawing reading, operation sequence, program and verification record of every job into a system, reusable. This step can now start with an AI drawing-to-code tool, at far less cost than building your own system. Once the preparation flow is smooth, then talk about extending the machines' unmanned execution window.

BestAI CAM Manufacturing Intelligence · Published 2026-07-19 · ~9 min read · For: shop owners, second-generation successors, production-control and process managers

The 24-hour unmanned factory that 'profits on a single piece': a guide to the Kyoto HILLTOP tour video—article cover

TL;DR HILLTOP, in Uji, Kyoto, is a legendary case study of Japanese manufacturing transformation: its mainstay isn't mass-production parts but prototype and development jobs—about eight in ten orders are for just one or two pieces. By rights this is the hardest kind of business to automate, yet it has built a 24-hour lights-out operation that doesn't rely on night shifts. The channel XU's factory-tour video (60,000 views in a month) breaks down its secret: an in-house digital production system that accumulates and reuses machining data and know-how—letting even young employees work, and using software capability to have the machines produce on their own at night. Order is everything: turn machining knowledge into software first, and only then is there anything for the unmanned hours to run. High-mix low-volume Taiwan shops, take note: HILLTOP proves "profits on a single piece" and "going unmanned" can hold true at once, but the starting point isn't buying robots—it's turning the knowledge from every drawing to every program into a system.

01What this tour video is about

HILLTOP, located in Uji, Kyoto, was once a traditional mass-production machine shop, and is now the transformation case Japanese media and business schools most love to cite. This June-2026 factory-tour video from the channel XU (クロスユー) racked up 60,000 views in roughly a month after going live^[1]:

XU visits HILLTOP in Kyoto: eight in ten orders are one or two pieces, yet it achieves 24-hour lights-out operation through a digital production system (open on YouTube)

The video's description spells out the company's contrast plainly: its mainstay is small-batch machining of prototypes and development jobs, with about eight in ten orders for just one or two pieces. It has systematized (turned into a repeatable structure) this business that "looks the hardest to make efficient" through its own digital production system, accumulating and reusing machining data and know-how so that young employees can thrive, and achieving 24-hour lights-out operation that doesn't rely on night shifts through applied software technology^[1].

02A contradictory combination: high-mix low-volume × unmanned

Anyone who knows the shop floor will frown the instant they see "eight in ten orders are one or two pieces" plus "24-hour unmanned," because in traditional logic these two things are mutually exclusive:

The traditional premise of going unmanned is repetition: run the same program and the same fixture thousands of times, and only then does the upfront investment amortize—automation theory ranks batch size as the primary criterion for automation investment^[3].
The essence of high-mix low-volume is variation: every order is a new drawing, a new program, a new setup—each "change" demands human judgment and preparation.

In the language of manufacturing-systems theory: there's a natural trade-off between flexibility and efficiency, and to have both at once you must push the "cost of coping with variation" down to near zero^[2]. HILLTOP's very existence proves this can be done—the question is how.

03The answer that resolves the contradiction: turning knowledge into software

The video's answer isn't a more expensive robot, but a digital production system: accumulate machining data and know-how in the system, let preparation be done in the office, and have the machines execute the daytime-prepared programs at night^[1]. Broken down, it's three interlocking links:

Knowledge goes into the system, not into individuals: for every job ever done, its data and tricks are retained and reusable—so the preparation cost of the next similar order keeps declining. This is the machine-shop version of the same logic in the Kawasaki Heavy Industries piece.
Preparation and execution are separated: during the day people do the judgment and preparation (programming, simulation, scheduling); at night the machines purely execute—the unmanned hours run already-verified programs, which is the safety premise that lets going unmanned stand up.
Young people can get up to speed too: with knowledge in the system, newcomers can work through the system—the video especially stresses that HILLTOP values talent development; unmanned operation and cultivating people aren't opposed, because the night-shift labor saved is invested in higher-value work^[1].

Academically, this is a textbook instance of a cyber-physical system "driving the physical layer from the information layer"^[4]: the protagonist of going unmanned was never the robotic arm—it's the information system that quickly turns every new drawing into a reliable program.

04Lessons for Taiwan shops: order matters more than ambition

Taiwan's job-shop machining is highly isomorphic to HILLTOP's situation: high-mix, low-volume, tight lead times. Its roadmap is worth copying precisely because of the order:

Step	Content	The common wrong order
① Turn knowledge into software	Get drawing→operations→program knowledge into a system, reusable	Buy the robots and automated storage first, only to find every new order still needs someone up all night programming—the equipment is waiting on the program
② Automate preparation	Programming, simulation and verification done quickly in the office
③ Make execution unmanned	Run verified programs at night, gradually extending the unmanned window

Steps ① and ② are exactly the segment BestAI CAM covers: AI reads the 2D drawing, builds the 3D model, generates G-code within the shop's tool library and machine limits, and has built-in 3D cutting simulation plus independent-AI dimensional cross-verification—the preparation cost of every new drawing is compressed, knowledge automatically accumulates into a shop-floor asset, and people focus on judgment and oversight (see the complete guide to CNC auto-programming; for the overall order of transformation, see the first step of smart manufacturing). HILLTOP spent years building its own system; today's tools let small and mid-size shops stand on the same logic and start with a far smaller investment—get knowledge into the system first, and only then does the unmanned night have anything to run.

05FAQ

Eight in ten of HILLTOP's orders are just one or two pieces—how can it possibly go unmanned?

Because what it makes unmanned isn't the "coping with variation" part, but the "execution" part. Variation (new drawings, new programs) is handled by the daytime digital production system and people; at night the machines only execute already-verified programs. Separate preparation from execution, and drive preparation cost down with a system, and high-mix low-volume and unmanned operation are no longer contradictory.

For a Taiwan shop that wants to learn from HILLTOP, what's the first step?

Not buying equipment, but turning the "drawing-to-program" knowledge into software: getting the drawing reading, operation sequence, program and verification record of every job into a system, reusable. This step can now start with an AI drawing-to-code tool, at far less cost than building your own system. Once the preparation flow is smooth, then talk about extending the machines' unmanned execution window.

Won't going unmanned leave young employees with nothing to learn?

HILLTOP's video gives the opposite answer: putting knowledge into the system actually lets young people get up to speed faster (they can work through the system), and the repetitive labor and night-shift hours saved are invested in higher-value work and talent development. The key is to allocate the dividend of going unmanned to learning and improvement, not simply to cutting headcount.

Subscribe to the tech-blog newsletter (newsletter in Chinese)

Get notified when new articles and video reviews go live—no inbox flooding, one-click unsubscribe.

KNOWLEDGE FIRST, UNMANNED NEXT

The first step to going unmanned is getting knowledge into the system first

AI reads the drawing, models, generates code, simulates and verifies—drive down the preparation cost of every new drawing, and the unmanned night finally has a program to run.

Contact an onboarding advisor Training courses

📋 Download the free "CNC Outsourcing Checklist" (printable) (中文)

← Back to the BestAI CAM product overview

06References

XU (クロスユー) (YouTube). 【前編】24時間無人で動く工場｜「1個だけ」の注文で利益を出す日本企業の秘密【HILLTOP】 (published 2026-06-01). youtube.com/watch?v=HsI7MdsTdPk
Chryssolouris, G. (2006). Manufacturing Systems: Theory and Practice (2nd ed.). Springer.
Groover, M. P. (2019). Automation, Production Systems, and Computer-Integrated Manufacturing (5th ed.). Pearson.
Monostori, L., Kádár, B., Bauernhansl, T., Kondoh, S., Kumara, S., Reinhart, G., Sauer, O., Schuh, G., Sihn, W., & Ueda, K. (2016). Cyber-physical systems in manufacturing. CIRP Annals, 65(2), 621–641.