AI, SAP and the Slow Loss of System Thinking

Notes from the edge of modern transformation

Something feels strange in many current ERP and IT transformation programs.

Not technically strange. Structurally strange.

**Free-link for non-members -> **If you like this kind of work => imagine to become a member

After 30 years in SAP environments, I increasingly observe a paradox:

Systems become more distributed, more connected, more API-driven, more “agile” — while at the same time deep system understanding seems to decline.

Not everywhere. Not for everyone. But enough to become visible.

And perhaps critical.

In earlier SAP years, modifications were expensive. Difficult. Risky.

A “Z” object was usually a conscious decision.

Not because old systems were simpler — they absolutely were not — but because complexity was respected differently.

There was often an implicit architectural instinct:

“Do not touch the core lightly.”

Today many organizations are technically more modern, but organizationally less flexible.

Why?

Because standardization is rarely just technical.

It affects:

• local autonomy
• historical compromises
• responsibilities
• reporting structures
• power distribution
• organizational identity

So instead of simplifying processes, companies often externalize complexity:

• into extensions
• side-by-side services
• APIs
• workflows
• cloud layers
• governance abstractions

The complexity does not disappear.

It moves.

This becomes especially interesting when AI enters the picture.

Large Language Models are astonishingly capable at:

• summarizing
• connecting concepts
• generating architecture patterns
• producing convincing explanations

Which creates a dangerous temptation:

delegated judgment.

Not because AI is “bad”.

But because plausible synthesis can slowly replace human depth of understanding.

The real danger may not be:

“AI replaces architects.”

But rather:

“Humans stop thinking architecturally because AI sounds convincing enough.”

ERP systems are not social media apps.

They:

• book money
• govern logistics
• produce tax-relevant data
• support compliance
• connect real-world operations

Their failures are not cosmetic.

They are operational.

Which means: deep system thinking still matters.

Perhaps more than ever.

This is why I increasingly believe:

Good AI will not replace good system architects.

It will amplify them.

At the same time, AI may also amplify weak governance, shallow understanding and fragmented responsibility — if organizations lose the ability to think in systems.

And maybe this is the real transformation challenge ahead:

Not cloud. Not SAP. Not AI.

But maintaining human judgment inside increasingly abstract technological environments.

The future probably does not belong to:

• “ECC forever”

or

• “everything agile-cloud-native”

The future likely belongs to organizations that relearn:

• what must remain stable

• what may become adaptive

• where standardization is useful

• where individuality is essential

• and where complexity must be consciously owned instead of hidden.

Because complex systems eventually reveal whether they are truly understood.

Usually under stress.

And always eventually.

The iTegrity Field Field Notes on Systems, Structure and Collective Leadership