[Case Study] A 199-employee manufacturer with no IT dept gains judgment criteria for software outsourcing — Nabel Co., Ltd. × Leach Generative AI Advisor

Mr. Iwai: The company originally started making bellows for cameras, and today we supply bellows used in medical equipment, aerospace, measurement devices, and various other machinery as a core technology. The stretch-and-compress functionality of bellows is the foundation; we build on that to fit our customers' equipment.

Beyond OEM products, we've recently been developing unique businesses centered on customer-intimate services, such as consumable filters and robot condition monitoring (Robot Insight). We also distribute robots from a Taiwanese manufacturer in Japan, and our business gravity is shifting from hardware toward software-adjacent domains.

Mr. Nagai: We were originally a hardware manufacturer. But with new initiatives like robot condition monitoring, software-intensive development has grown rapidly. That brought up questions we couldn't answer with our existing knowledge: how much time things should take, and whether supplier quotes were reasonable. We wanted advice on that — that was the starting point for bringing in the advisor.

Mr. Iwai: As a small and mid-size company, we don't go deep on technology. With AI evolving fast, we didn't have anyone internally who could handle information systems or advanced technical work. Everything was outsourced to partner companies — a common situation for SMEs with limited hiring reach.

Hiring a specialist mid-career is one option, but what we actually wanted was to train our existing people. That's why we asked for a format where we'd receive advice and move forward internally, rather than having the advisor take on the full build.

Mr. Iwai: Internally, we were running much of our operations on Microsoft Access — drawing number issuance, progress tracking of unreleased drawings, work instructions, even drawing templates and annotation frames. Without the discipline of requirements definition or flowcharts, things just accumulated, and bugs and inefficiencies became the norm.

Knowledge was siloed with specific people. Design work also runs on Access in many places, and it's buggy and cumbersome. The voices calling to "get off Access" had been there for a long time.

Mr. Nagai: The hardest part was that we had no way to judge whether the software quotes coming in from suppliers were reasonable. We were signing off without really understanding them. Even looking at the breakdown, we had no yardstick for what was reasonable and what was excessive.

Conversely, features we asked for sometimes turned out, in hindsight, to be unrealistic requirements. Development was running forward while the side making the decisions lacked the knowledge to judge.

Mr. Nagai: It started with the fact that we'd been colleagues at our previous company, and the trust built through that relationship was enormous. From there, through the study sessions you led, the advice you gave on quotes, and your attitude of quickly solving concrete problems like purchase-order automation, I felt: this is someone I want to work with.

"Solve the customer's problem" — that's been our company's value for as long as I can remember. Your stance aligned with that, which made it natural to work together.

Mr. Iwai: From what I hear from the CEO, Nabel deeply values the concept of "ichigo ichie" (treating each encounter as a once-in-a-lifetime opportunity). Together with the Chairman's personality, we have a culture of building deep, long-lasting relationships.

Expertise is a baseline requirement, of course. But a criterion we really cared about was whether we could engage beyond pure business, on a human level.

Mr. Nagai: We started the study sessions because we wanted to do something about the siloed knowledge and the systems fragmented by department. Each department had built what they needed, but the company didn't have a cohesive system. We wanted to see if a unified, end-to-end system was possible. Alongside that, we wanted younger members to start engaging with AI and systems.

Mr. Iwai: What Tominaga-san has also been telling our Statutory Auditor, Tomoko Nagai, is that "if requirements aren't defined, having just the insides of Access isn't enough — you can't cleanly port tangled requirements." Rather than mechanically rewriting Access apps in Python, we first need to agree internally on our requirements framework, and only then turn that into software. Once the requirements are solid, Claude and similar tools can write much of the code — even without hands-on coding skills, software becomes possible. The real challenge is having members inside the company who can define requirements.

Mr. Iwai: During Robot Insight development, we have Tominaga-san sit in on the meetings with a major telecom company. We rely on him for two things.

First, third-party evaluation on the security side. Security issues on the end-customer side are terrifying if they happen. Even large enterprises have had IoT-related incidents, so having a third party ask "is this design's risk profile reasonable?" gives us real peace of mind.

Second, advice on AWS cost reduction and gateway-software cost optimization. For both initial and running costs, we get knowledge on how to build the same functionality more affordably without strain.

Mr. Iwai: Right now, the only person on our team who can take software development all the way through to implementation is AN. He's also the only one who knows theory like requirements definition and flowcharts; the rest of us are at the level of having touched programming in vocational school coursework. Having a structure where Tominaga-san can share his experience when AN hits a wall is a huge help.

Mr. DO DINH AN: I've only worked with Python and similar tools on a self-taught basis — I'm not a professional by any stretch. That's exactly why having Tominaga-san sit in and share his experience when problems come up is the most valuable thing for me.

Mr. Iwai: At first I hesitated — "is this the kind of thing I should even be asking in chat?" — with the cost in mind. But since we're increasing the monthly fee at renewal, I'll be making full use of it going forward.

Mr. Nagai: Without the advisor, we might have approved quotes at those numbers without understanding whether they were reasonable, and we probably wouldn't have caught unrealistic requirements either. The current project is going well, but without the advisor, it might have taken longer, cost more, and honestly — I'm not sure we could have realized it at all.

Mr. Nagai: The purchase-order automation we initially asked for has started working smoothly. On top of that, we want to increase the frequency of the study sessions too, and we're committed to moving this forward ahead of schedule as a company. That's why we proposed, without waiting for the contract renewal, to increase the monthly fee.

Mr. Nagai: The working population is going to shrink — that's certain. Forecasts point to roughly 11 million fewer workers by 2040. That's exactly why non-productive indirect work should be replaced with AI while we still have people available.

Scrambling to get started five or ten years from now will be too late. Embedding AI into the company now is essential for stable operation in the future.

Mr. Iwai: In manufacturing especially, production management systems bloat in cost because standard software gets customized over and over — it's a shared pain. I've seen plenty of companies end up with a modification patchwork they can't maintain.

Cutting fixed software costs (for example, migrating Office 365 to Google Workspace) and expanding what you can handle in-house using generative AI — that's an approach that applies to many SME manufacturers' challenges.

Mr. Nagai: Going forward, I want to actively bring in generative AI tools like Gemini. Using study sessions and support to promote AI adoption internally, and dissolving resistance to new technology — that's the top priority for the moment.

Mr. Iwai: On AWS costs and the volume of raw data Robot Insight accumulates — how to compress, analyze, and store it — we'd like concrete guidance going forward.

And when we bring AI into the company overall, role-splitting matters: IoT and robot domains with different partners, internal DX primarily with Leach. We want to keep asking whoever is best placed to provide the right solution.

Mr. Nagai: For manufacturing executives facing similar issues, I'd say: bring in outside expertise. For companies that share the common pain of "hiring is hard" and "we have no IT department," an external technical advisor is a realistic option.