Inside the 14-Person IT Team: Why One Food Company Leader Treats AI Like a Risk Test
A compact corporate tech group is turning explainability into an operational habit, using it to keep speed, governance, and AI-assisted work from drifting out of balance.
When an IT organization is small, every decision carries more weight. In Calbee’s 14-person information systems team, that reality shapes both leadership and daily operations: choices are not just judged by whether they work, but by whether the decision-maker can explain them cleanly to himself first.
Fast Facts
- Calbee’s information systems organization has 14 people.
- IT leader 井原史晶 says his main decision rule is whether he can explain the choice to himself.
- He uses generative AI as a “sparring partner” for learning, troubleshooting, and early design work.
- He has worked in SIer roles, an audit firm, and joined Calbee in 2017.
- He frames IT leadership around fairness, business contribution, and control.
The security value in that mindset is easy to miss. Explainability is not only a management trait; it is also a practical filter against rushed automation. In a small team, generative AI can accelerate log review, idea testing, and proof-of-concept drafting, but it can also compress the time available for skepticism. If an answer cannot be justified in plain language, it is more likely to hide a mistaken assumption, a bad requirement, or an unreviewed risk.
That is why his emphasis on fairness matters. He describes weighing the needs of business teams and SIer partners without treating either side as automatically right or wrong. From a defensive perspective, that is close to a security review mindset: compare constraints, document tradeoffs, and decide only after the operational impact is understood. In environments where low-code tools and AI make it easier for non-specialists to build faster, that kind of judgment becomes a control, not just a leadership style.
The broader lesson is that speed and governance do not have to be opposites. A small IT team can move quickly, but only if it has a consistent way to validate outputs, assign responsibility, and challenge convenient answers before they become production decisions. That matters whether the work is a system rollout, a troubleshooting session, or a new AI-assisted workflow.
He also points to industry collaboration around EDI and cross-company information sharing as part of the job. That reflects another reality of modern enterprise IT: business systems rarely sit in isolation. Identity, integrations, data exchange, and partner processes can create risk far beyond the internal team if they are not handled with discipline.
The clearest takeaway is not that AI should slow IT down. It is that AI should raise the standard for human review. In a 14-person team, the real strength is not doing everything alone. It is building a habit of decisions that can survive scrutiny, explanation, and operational pressure.
TECHCROOK
hardware security key: A compact hardware security key adds a physical second factor for account logins and admin access. For small IT teams, it can help keep privileged accounts tied to a specific device and user, making everyday access control more disciplined without adding much friction.
WIKICROOK
- Explainability: the ability to clearly justify a decision or output.
- Generative AI: software that creates text, code, or other content from prompts.
- PoC: proof of concept, a small test to check whether an idea is viable.
- Governance: the oversight and rules that keep technology use controlled and accountable.
- EDI: electronic data interchange, a structured way for businesses to exchange documents.
