Carousel — From Noise to Signal

Douâa Bergheul

B. Ing | Strategy & Technology Leadership

Episode 01 — Learning AI in the real world

"Use AI more."
Everyone says it.
But what does it mean ?

So I spent a day building a framework with Claude.

From noise → to signal.

02 / 05

First, let us not forget what AI actually is.

An LLM is a statistical prediction engine. It predicts the most likely next word based on patterns in training data. It is, by design, backward-looking and imitative it cannot reason about what has never existed.

AI is most powerful when your task has clear patterns in its training data and most dangerous when it doesn't.

03 / 05

I built a first framework... but it was not quite right.

High stakes

Low stakes

Fully humanLeadership & people management
Culture & change
Crisis management

Human leadsComplex project planning
Stakeholder strategy
Risk assessments

AugmentLearning new topics fast
Brainstorming
Research synthesis

Delegate to AIEmail templates
Meeting summaries
Data formatting & cleanup

Low pattern

High pattern

The crack I found

This framework does not decompose the problem the right way. It's not about keeping some subjects off limits, it's more about what layer of work you are doing.

You can totally write communications about culture or change with AI, if you provide it with the content you selected.

04 / 05

The real framework: four layers, one question.

Layer 1

SENSE

Room dynamicsWhat's unsaidPolitics

AI cannot intervene

Layer 2

JUDGE

What to doWhich riskThe call

AI simulates, you verify

Layer 3

BUILD

DraftStructureSummarizeFormat

AI genuinely accelerates

Layer 4

EXECUTE

PresenceAccountabilityAdapt live

AI cannot intervene

Ask before every task: am I handing over the judgment, or just the execution?

05 / 05

What good AI use actually looks like.

1

Understand what it is. A prediction engine that sounds confident to encourage users to continue engaging.

2

Know your layer. Sensing and judging are yours. Building is where AI earns its place.

3

Push back. If something feels off, it probably is. Your discomfort is data.

4

Never outsource the judgment. Speed at Layer 3 means nothing if Layer 2 was skipped.

Episode 01 of learning AI in the real world.
Follow for what I discover next, including when I get it wrong.

Douâa Bergheul

From noise to signal