Concept card

Deep Work

As of 2026-10-10

Edited by Cédric Mary, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is Deep Work?

C. N.'s deep-work ceiling is about four focused hours a day — most consultants actually deliver one to two, and that gap, not the tooling, is usually the real constraint on delivery speed.

What it is

C. N.'s concept — cognitively demanding work performed in distraction-free concentration. 4 hours/day is the realistic ceiling; most knowledge workers achieve 1-2 and wonder why output is low.

Example: In a real Category M · Productivity & AI engagement, Deep Work is used to align business ownership, data contracts, and delivery sequencing before solution build starts.

Counterexample: Teams jumping directly into dashboards without clarifying Deep Work usually create rework, semantic drift, and sponsor distrust.

Decision tree: If the business question is stable and recurring, prioritize canonical modeling; if volatile, start with a constrained pilot and explicit assumptions; if data quality is low, run remediation before scale-out.

KPI exercise: define baseline and 90-day target for freshness, trust score, adoption rate, and decision latency; then attribute variance to one change at a time.

Case prompt: design a 2-sprint plan using Deep Work for a cross-country SAP analytics program in Productivity & AI, with trade-offs, governance checkpoints, and executive narrative.

Why it matters

The realistic daily ceiling for distraction-free cognitive work is ~4 hours — planning a delivery sprint on more than that overstates real capacity.
Most knowledge workers land at 1-2 hours of true deep work, so protecting even a modest block is a bigger lever on output than adding headcount.
Skipping deep work on aligning business ownership, data contracts and delivery sequencing before build starts is what produces rework and sponsor distrust later in the engagement.

Key points

C. N.'s concept — cognitively demanding work performed in distraction-free concentration.
4 hours/day is the realistic ceiling; most knowledge workers achieve 1-2 and wonder why output is low.
On an SAP analytics engagement, deep work concentrates almost entirely in build weeks (data modeling, CDS view logic, calculation view tuning) — workshop weeks are structurally shallow and should be scoped as such, not blended into the same calendar rhythm.
Calendar fragmentation, not tooling, is the most common deep-work killer on a delivery engagement — a client defaulting to daily status calls during a build phase caps available focus time at whatever gaps survive between meetings, often under an hour.
Generative AI tools (Joule Studio's intent-based generation, an AI-first draft workflow) compress the shallow, mechanical part of a task — scaffolding, boilerplate, a first-pass prompt template — without compressing the judgment layer that decides whether the output is actually right.
Reviewing AI-generated output is a harder form of deep work than producing original work from a blank page, because an AI failure is more often silently plausible than obviously broken — skimming a review in the gaps between meetings is exactly the wrong calendar shape for that task.
A senior consultant negotiates meeting cadence explicitly at kickoff — fixed office-hours windows for ad hoc questions, the rest of the day protected — and revisits it phase by phase rather than defaulting to the client's own internal rhythm.

Terms used on this page

Judgment layer: The part of the work that AI cannot do — prioritisation, trade-offs, client-context reading.
AI-first draft: Workflow where AI produces the first pass (code, memo, slide) and the consultant edits rather than writes from scratch.
Agent reliability: The consistency, cost, safety, and policy compliance of an agent across repeated runs.
Calendar fragmentation: A meeting cadence (e.g. daily status calls) that caps available deep-work time at whatever gaps survive between calls, often under an hour — the most common deep-work killer on a delivery engagement.
Attention residue: The lingering cognitive cost of a context switch — part of why back-to-back client calls during a build week degrade complex modeling work even in the gaps between them.
Silent failure (AI review): An AI output that is plausible-sounding but wrong rather than obviously broken — the reason reviewing AI-generated work needs deep-work attention, not a shallow-work skim.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks · the facts worth quoting.

Open in the app →