The Workflow
It’s Monday morning and I’m scrolling back through last week’s calendar trying to remember what I actually did. Three training sessions? Four? There was a deck I sent to the sales DL, but which day? The Intranet post about the new compete battlecard, was that this week or last week?
Ninety minutes later I have a newsletter draft.
I do this every week. I have been doing this every week for over a year.
The trigger was realizing that the 90 minutes wasn’t writing the newsletter. It was archaeology. The writing took fifteen minutes once I had the inputs in front of me. The rest was hunting down my own work across Gmail, the LMS, the Intranet, and Slack.
So I stopped doing the archaeology and built a capture layer instead.
Step 1. Define the inputs: I told Claude what counts as newsletter-worthy activity. Meetings I host with more than ten people invited. Emails to the DL-Sales distribution list. Documents published to the Intranet in the last 7 days. Anything posted to a dedicated Slack channel called #enablement-edge-submissions.
Step 2. Build the capture channel: The Slack channel is the part that does the real work. Any time I publish a training, record a session, or ship a resource, I drop a one-line note in the channel. Three to seven entries a week from me. One or two from other people who’ve asked to contribute.
Step 3. Show it what good looks like: I shared a previous edition of the newsletter, the one I had formatted by hand in Google Docs over an hour, with Claude as the reference template. Sections, headers, tone, length per item.
Step 4. Run the weekly pull: Every Monday morning Claude scans the last 7 days across the defined inputs, consolidates the entries, drops them into the template, and hands me a draft.
Step 5. Edit and ship: I spend ten to fifteen minutes editing voice and order, then paste it into Gmail.
The newsletter goes out Tuesday. I get my Monday afternoon back.
What Broke
V1 of the AI-generated newsletter opened like a task list. A wall of bullets. Long run-on text under bolded headings that were trying very hard to look like sections but mostly looked like a status report somebody forgot to format.
It also had a bigger problem. The meetings and emails were all there. The training content I’d posted to the LMS and the new battlecard on the Intranet were not. Claude couldn’t see them because nothing in those systems exposed activity in a way it could pull.
I almost spent a weekend trying to wire up API access to the LMS.
Instead I asked Claude how it would capture that activity if it were me. It suggested a dedicated Slack channel I post to whenever I publish something. One line per entry. It would read the channel on Monday morning.
That fixed the capture problem. The formatting problem got fixed separately. I took a previous edition I had spent an hour hand-formatting in Google Docs, shared it with Claude, and added to the prompt: Match the section structure, header style, and per-item length of the attached reference document. Do not produce a bullet list.
The lesson: when AI can’t find your work, the problem is almost never the model. It’s that your work doesn’t exist anywhere structured enough to be found. The fix is a capture layer, not a better prompt.
Time Ledger
- Time saved: 90 minutes a week of calendar archaeology and cross-system hunting.
- Time added: Roughly 2 hours up front to define inputs, build the Slack channel, and feed Claude a reference edition. Plus about 90 seconds a day to log entries as I publish them.
- Net: Break-even by week two. Pure win from week three on.
The number that actually matters isn’t the 90 minutes. It’s that the newsletter now ships on a predictable Tuesday instead of whenever I finish the dig.
The Prompt File
You are my Enablement newsletter editor. Your job is to assemble the weekly draft of "Enablement Edge" from structured activity logs across the last 7 days and output it in the exact format of the reference edition I provide.
**Step 1. Ingest the inputs:** Pull from the following sources for the date range [START DATE] to [END DATE]:
- Slack channel `#enablement-edge-log`, every message posted in the window, including thread replies
- Calendar meetings I hosted with more than 10 invitees
- Sent emails addressed to `DL-Sales@[COMPANY].com`
- Documents I published to the Intranet folder `[INTRANET FOLDER PATH]`
**Step 2. Deduplicate:** If the same activity appears in more than one source (e.g. a training session logged in Slack and also on the calendar), keep one entry and use the Slack version as the source of truth for the description.
**Step 3. Categorize:** Sort entries into the sections used in the reference edition: `[REFERENCE EDITION SECTION LIST]`. If an entry does not fit, place it under "Also This Week" rather than inventing a new section.
**Step 4. Write each item:** 2 to 4 sentences per entry. Lead with what shipped or happened, not with "This week I." No adjectives like "exciting," "amazing," "powerful." If an entry has a link in the Slack post, preserve it.
**Step 5. Match the format:** Match the section structure, header style, and per-item length of the attached reference document: [PASTE REFERENCE EDITION]. Do not produce a bullet list. Do not produce a status report. Section headers are bold, one line. Items sit under their section as short paragraphs.
**Step 6. Output:** Return the full draft in Google Docs compatible formatting. Do not include commentary, a summary of what you did, or suggestions for next week. Draft only.
Constraints:
- If a source returns zero entries for the window, omit that section entirely. Do not write "Nothing this week."
- Do not infer activity that isn't in the inputs. If it isn't logged, it didn't happen.
- Total length: 600 to 900 words.Manager’s View
Ask your VP of Marketing: what activity in your function isn’t captured anywhere a system could read it?
Most Marketing leaders assume their work is well-documented because campaigns live in a project tool and assets live in a DAM. The gap is everything in between. The enablement session for a product launch. The competitive note shared in a Slack thread. The exec briefing that never made it into a deck.
If you ever want AI to summarize, report on, or repurpose your function’s output, the bottleneck won’t be the model. It will be that half of what your team actually ships is invisible to any system trying to read it.
The cheap version of the fix is a single Slack channel and a one-line posting habit. No new tools, no integration project, no governance review. Just a place where work gets logged the moment it ships.
Ask the question once. Then watch how long it takes anyone to answer.
Field Notes
The distribution list started at 60 people in Sales and SE.
It now sits at around 150 across Sales, SE, Customer Success, HR, Legal, and Marketing.
Over half the company reads a newsletter I built for one function.
Legal asked to be added after seeing a forwarded copy. Legal. Reading Sales Enablement updates. Voluntarily.
The Slack channel was supposed to be a private capture tool for me. Two people asked for access in the first month. I now get one or two contributor entries a week without prompting anyone.