AI & legacy modernisation
How to cut admin effort by up to 40% when your staff fear the new software
Replacing five COBOL and FileMaker applications looked like a database project. In reality it was a project about fear, and that is exactly what AI rollouts face today.
By Christian Liebich · · 2 min read · Executive Briefing
A few years ago I was tasked with replacing five legacy applications for the Swiss Federal Customs Administration. They were written in COBOL and FileMaker and handled up to 1.5 million records. On paper it looked like a classic database and web-architecture project. In reality it was a project about human psychology.
The real bottleneck
Specialists for the old administration could no longer be found on the market. And the long-serving employees, 15 people who kept the operation running every day, were genuinely afraid that they would not be able to operate the new web applications.
People who are afraid of looking incompetent quietly slow down every digital transformation. We are seeing the same fear today in AI rollouts.
What we did instead of a 100-page specification
- Prototypes first: people cannot trust what they cannot see. I sat with the end users and built visual, clickable prototypes early, so that they could feel the new workflow before a single migration happened.
- Respect for domain knowledge: we treated the long-serving staff as co-architects of the process, not as relics to be replaced.
- One win after the other: all five applications were built in parallel on an internal Linux cloud with three environments, then signed off, migrated and put into production one by one.
The result
All five applications went live on schedule and to the full satisfaction of the end users. The administrative effort of the 15 employees fell by up to 40%.
Whether you are migrating off 30-year-old COBOL or introducing AI agents today: technology only moves at the speed of human trust. Remove the fear first, and the return follows.
