ARCHIVE · May 14, 2026 · Issue 5 · Real Results

What a Week Looks Like When Claude Runs Your Workflow

By Cynthia Schomp · Fractional CTO · May 14, 2026

What a Week Looks Like When Claude Runs Your Workflow

By Cynthia Schomp · Fractional CTO


I tracked a full work week with Claude running as a workflow layer. Here's what actually happened.


Monday: Client onboarding intake

A new client kicked off Monday morning. I had an intake form response, a 45-minute call recording transcript, and a scattered thread of emails. Claude turned all three into a structured brief: client goals, technical constraints, open questions ranked by urgency, and a first-week action plan.

That used to take me 90 minutes. It took 12. The output needed one round of corrections. Claude flagged a deadline as Q3 when the client had said end of June. That's not a failure. That's exactly the kind of thing I want caught before it becomes a problem.


Tuesday: Proposal draft

The brief became a proposal. First draft was 80% there. The problem: Claude led with a capabilities overview instead of the client's problem. I redirected it. The second draft was right. Two passes total, 35 minutes start to finish. The version I would have written solo would have taken two hours.


Wednesday: Content day

I had five pieces queued: a LinkedIn post, two newsletter sections, a speaker bio update, and a short technical explainer for a client's internal team. Claude handled the drafts for all five. Four were usable with light edits. The LinkedIn post was flat. Correct voice, wrong energy. I rewrote the opening line myself and Claude tightened the rest.

Total content production time across all five pieces: 1 hour 40 minutes. My previous baseline for the same output was around 4.5 hours.


Thursday: Technical documentation

A client needed internal documentation for a new AI tool they were rolling out to their team. Claude built the full doc: section headers, step-by-step instructions, a troubleshooting section, and a one-page quick reference.

The troubleshooting section was better than what I would have written. Claude anticipated questions I hadn't thought to include. That doesn't happen every time. Thursday was a good day.


Friday: The problem

End-of-week review: I asked Claude to summarize the week's decisions. What we'd committed to, what was still open, what needed a follow-up.

It couldn't do it. Not because it failed. Because it didn't have access to the previous sessions. Monday through Thursday each existed in their own context window. I had to paste in notes from each day manually and reconstruct the summary myself.

That worked. But it was the friction point that sat with me all weekend.


The Numbers

Estimated time without Claude: 22 hours. Actual time with Claude: 9 hours. Time spent correcting or redirecting Claude: about 90 minutes of that 9.

The 90 minutes of corrections isn't overhead. It's the job. I'm still the one making the calls. Claude is handling the drafting, structuring, and formatting so I can spend my time on judgment instead of production.


The Thing That Bothered Me

The week worked. The week was also a dead end. Each session was its own closed system, and the compounding that should have happened between them didn't.

Every Monday it started over. No memory of the decisions we made Tuesday. No awareness that the client in the proposal preferred plain language. No record of what got corrected.

I sat with that for a while. Then I decided to do something about it.


If you missed any emails in this series, the full archive is waiting: hellocyn.com/archive -- every issue, in order, yours to read whenever.


-- Cynthia Fractional CTO · Schomp.ai · The CTO's Desk


Reply to this if you're building something similar. I read every reply.

Get the next one.

One email per day. Real deployments. No theory.