Dr. Pascal Giessler · AI Principal · Freiburg

The architect against
Cognitive Debt

AI-native transformation is making judgment cheap to skip. Organisations adopt agentic systems they cannot explain, and quietly lose the ability to assess their own decisions. I build sovereign AI for regulated and future-critical DACH organisations, where the harness keeps the judgment inside the house: auditable, exit-capable, resilient.

PhD · KITMIT · Innovation ManagementMBAAzure Solutions Architect ExpertFormer CTO · DACH consultancy
Dr. Pascal Giessler

The interest comes due later

Cognitive Debt is the slow erosion of an organisation's judgment when AI systems produce decisions no one can any longer explain or check. Like technical debt it stays invisible, until the interest falls due.

Liability

In a regulated context you have to be able to justify every output. From August 2026 the EU AI Act makes traceability an obligation, not a comfort.

Dependence

Hand your judgment to a black box and you cannot get it back. The vendor owns the model, the logic, and eventually the decision.

Fragility

One provider outage, one price change, one deprecated model, and the organisation's ability to decide goes with it.

The framework, and how to measure it →

Three levels of the same move. Judgment is not a policy you write, it is a property you build in.

01

AI-Native Transformation

Rebuild how the organisation works around AI, rather than bolting AI onto how it already works. The tell is where the judgment sits: a bolted-on rollout buys tools and hopes for output. A native transformation redesigns the decision paths and the review that keeps those decisions explainable.

decision pathsgovernanceenablementreview

Definition →

02

Agentic Engineering & the Harness

An agent is a model in a loop calling tools. The loop is the trivial part. The reliability, and the judgment, live in the harness: the tools you expose, the context you assemble, and the condition on which the loop stops. That is where the engineering belongs, not in the prompt.

harnesstool usecontext engineeringevalstermination

Definition →

03

Sovereign AI

Architecture the organisation actually owns: auditable, so it holds up in front of a board and an auditor. Independent, so no vendor owns your exit. Resilient, so an outage or a deprecation does not take your decision-making with it. Not on-premise for its own sake, but control over model, data, logic, and liability.

auditableexit-capableresilientEU AI Act

Definition →

Instruments, not demos

Two tools I built and use. Each is the same thesis in running code: valuable AI forces a decision instead of deferring it.

"Ever tried. Ever failed. No matter. Try again. Fail again. Fail better."

Samuel Beckett, Worstward Ho (1983)