Deploying Data Infrastructure Around Legacy Systems

You keep what works, and give its data somewhere modern to go.

Why legacy is the reality — not a blocker

Manufacturing runs on long asset lives. Equipment and systems stay in service for decades because they're expensive, validated, and — crucially — they work. That's not a problem to apologize for; it's normal. The trap is the belief that modernizing your data means first replacing all of it. That belief keeps manufacturers stuck, waiting for a massive rip-and-replace that never comes (and shouldn't). The reality: you can build a modern foundation that connects your legacy systems in, and reach the same connected, AI-ready state without tearing out the floor.

Why you don't (and can't) just rip and replace

Replacing working legacy systems wholesale is the wrong move for concrete reasons:

• Cost. Wholesale replacement is enormous capital expense, often for little functional gain.
• Risk. Swapping out a system that runs a live line risks exactly the downtime you can least afford — and re-validation, in regulated settings.
• They work. A system doing its job reliably doesn't need replacing just because it's old.
• Asset lifecycles. Manufacturing equipment is meant to run for decades; forcing replacement on a data project's timeline fights that economics.

"Modernize everything first" is the instinct to resist. Build around instead.

How to deploy around legacy

The approach, in practice:

• Connect, don't replace. Extract data from legacy systems through the right interface — a standard like OPC-UA for older PLCs and SCADA, APIs or connectors for legacy ERP and MES, and direct database or file extraction for the oldest systems with no modern interface.
• Bridge with middleware. Use adapters and middleware to translate old, proprietary formats into something the modern foundation can use.
• Build the modern foundation on top. Land the extracted legacy data in a modern warehouse or lakehouse, cloud or hybrid — the connected foundation sits above the legacy systems, fed by them.
• Layer analytics and AI on the foundation, not the legacy systems. Your BI and AI run on the modern foundation, so they're never limited by what the old systems can do directly.
• Modernize selectively, over time. Replace individual pieces when there's a real reason to — not all at once on a forced schedule.

The honest challenge

Working with legacy isn't free of difficulty — that's worth being straight about. Old systems can be hard to extract from: proprietary formats, missing documentation, no APIs, and quirks only a long-tenured engineer remembers. Getting clean data out of a twenty-year-old system is real engineering work. But it's routine engineering work — solved on manufacturing floors every day — not an impossible barrier. The difficulty is a reason to do it carefully, not a reason to either rip everything out or give up on modern data.

How this gets you to Connected

This is the practical answer for the many manufacturers running older systems: yes, you can have a modern, AI-ready foundation without replacing your floor. Legacy systems and a modern foundation coexist — the old systems keep running production, the foundation connects their data and makes it usable. It's how a manufacturer with decades-old equipment still reaches Connected on the Data Maturity Model. It's core data engineering and deployment work, and it pairs naturally with phased migration when some modernization is warranted.