共有:

Claude · Workflow Compression

Claude Compresses Two Months of Data Work Into One Week

"Two months of data cleanup done in one week by Claude. Nobody on the team missed those seven weeks." — Anthropic co-founder Kaplan's real-world benchmark is landing hard.

AI Navigate Editorial·2026.06.27·6 min read

Before: 2 months (8 weeks) manual data cleanup Claude 1 week done 7 weeks nobody missed
01
The Claim

Kaplan's Benchmark

Anthropic co-founder Kaplan publicly stated that two months of data cleanup work was compressed to one week using Claude — and not a single team member wished they'd done it the old way. While many organizations remain stuck in proof-of-concept mode out of quality concerns, this is a first-party, publicly disclosed operational result.

The number matters less than the phrase: "nobody missed those seven weeks." After recovering the time, the team implicitly asked whether it had ever been necessary — and concluded it wasn't. This is about surfacing unnecessary process, not just accelerating necessary work.

02

The Numbers

8wk→1wk
timeline compression
87.5%
effort reduction
0 people
wanted to revert
03
Your Org's Turn

How to Find Tasks to Compress

01

List every manual task done more than once a month

Reports, data aggregation, template emails, meeting notes, spreadsheet updates. The tasks that have become "just how we do it" are usually the highest-compression-ratio candidates.

02

Pilot the highest-effort task with Claude first

Aim for "80% quality in 20% of the time." If time spent checking the AI's output is less than the time saved, adopt it. Don't wait for perfect before shipping.

03

Decide in advance where the recovered time goes

Without a plan for the freed capacity, work expands to fill the time. Pre-assign the hours to strategic work before the compression happens — otherwise you've optimized a process without capturing the value.


Claude Compresses Two Months of Data Work Into One Week — Team Never Looked Back | AI Navigate