BD and operations status, July 7 to July 28.
A three-week status review of business development and supporting operations, covering campaigns, accounts, reports, and content produced across the four AI systems in daily use. Reconstructed from the session logs of each system and cross-checked against the files those sessions wrote under C:\Users\moveb\mna.
Eight workstreams
Grouped by what shipped, not by which tool ran it. Several streams crossed two or three systems.
PA NCF email campaigns
The largest single push of the period. Two more campaigns built, scored, corrected, and pushed live through Mailcoach, taking active outreach to 247 facilities across five lists.
- Campaign D (Western PA) topped an existing 30-subscriber list up to 47 active, adding 20 new facilities
- Campaign E (South Central / Lehigh) created as a new list at 50 active subscribers
- Both passed all nine compliance hard gates on July 22, including the no-PRN and Joint Commission "pursuing" language checks
- Signature, logo, and facility-naming fixes applied across every sequence before send
- 205-facility recycled pool assembled from inactive campaigns, with opt-outs and the three live Q3 campaigns excluded
New account development
Five new account folders stood up from Pulse links and Diane's call notes, each with contacts, research, and a first-touch email.
- Rose Manor Health and Rehab
- TSALI Care Center – RSI (Cherokee)
- Pine Crest Health & Rehabilitation Center
- Fletcher Rehabilitation and Healthcare Center
- Universal Healthcare – Fletcher
Weekly BD state focus
The weekly research and outreach cycle ran three times and moved from manual to scheduled.
- Research and outreach plans for July 13, July 20, and July 27
- July 20 report built as branded HTML and deployed to Cloudflare Pages
- Codified as the
mna-branded-reportskill so the styling stops being rebuilt each week - Now runs as a scheduled task, first automated execution July 27
Vendor and risk evals
Four evaluations completed, each following the same house pattern: research brief, scored eval, branded HTML report.
- USA Lab – backed risk deck
- Growth Beacon HQ – vendor eval
- StarCall AI – eval, research brief, and Pulse integration notes
- SourceChannel – eval and visual report
Recruitment marketing
Job requisitions turned into social assets, plus a lead magnet built to feed ManyChat DM capture.
- Four job carousels: LPN Michigan, RN Vermont, LPN New Hampshire, RN Ohio nights
- Vermont Travel Nurse Bucket List PDF lead magnet and landing page
- Instagram and Facebook story hyperframes at 9:16
- LinkedIn content batches on market and travel nursing research
Entity and finance
Corporate groundwork that unblocks government and MSP channels.
- SAM.gov entity re-registration playbook, notarized Entity Administrator letter drafted
- Entity registration checklist published as a live visual
- Business financing research: LLC lending options for a minority-owned company
- MSP tracking log picked back up against the Obsidian MSP Leads base
Team enablement
Getting a second caller productive without a manual ramp.
- Multi-page BD onboarding site built for Diane's first week: what we sell, systems, call playbook
- One-pager template reworked into a blind-referral version with rate language removed
- Anonymized candidate bio and follow-up email templates saved to /mna/bd
- Custom GPT scaffolded to give Diane clinician-to-clinician outreach context
Agent infrastructure
Work on the machine that does the work. Not billable, but it is why the other six streams moved faster by week three.
- MNA brand rules pushed to all four systems: logo, colors, no-PRN constraint
- Dedicated Hermes profile for MNA work, Telegram channels rebuilt, model routing moved off extra-usage credits
- Shared skills installed across Claude, Codex, Gemini, and Hermes
- Mailcoach API wired directly into Claude Code for campaign sends
Pulse is the CRM of record. Every call, email, and outreach activity for MNA is logged in Pulse, and that is what the team works from. The Twenty CRM instance referenced in the campaign build notes is a personal working database used only to assemble and de-duplicate lead lists before they reach Mailcoach. It is not a team system and nothing is logged there.
How the three weeks moved
Content production, then account and report building, then campaign execution at volume.
Content and groundwork
Carousel pipeline exercised on four job reqs. Vermont lead magnet and landing page built and made live. BD team agent definitions written. First weekly state focus research run on July 13. Twenty CRM migration started from the Obsidian PA base.
Reports, evals, onboarding
Heaviest build week: USA Lab risk deck, social traffic report, financing report, entity registration checklist, Growth Beacon HQ eval, and Diane's onboarding site. The MNA report look was codified into a reusable skill on July 20 and the weekly BD report went live.
Campaigns live and automated
Two 50-facility PA NCF campaigns scored and sent through Mailcoach, three more accounts opened, the recycled facility pool built, and the weekly BD report handed to a scheduled task. Diane's custom GPT scaffolded on July 27.
Shipped and live
Deliverables from this window that have a URL, and the design system each one uses. Everything else lives in the file tree.
Design styles in use
Six distinct visual systems are live across those ten properties. Only the first is codified as a reusable skill; the rest were one-offs built to fit their content.
| Style | Type and color | Used by | Status |
|---|---|---|---|
| MNA branded reportThe house system | Inter / system sans, oversized tight-tracked h1, MNA blue accent, auto light and dark via light-dark() tokens, lucide icons |
Grants report, weekly BD report, entity registration checklist, Growth Beacon HQ eval, this status report | Codified as a skill |
| Security dossier deckScroll-driven slides | Space Grotesk display, DM Sans body, Fira Code mono, palette sampled from the logo PNG, red reserved as a semantic danger cue only | USA Lab backed risk deck | One-off |
| Office documentReads like a Word file | Aptos, MNA blue and navy, scorecard tables, no dark mode | StarCall AI eval, SourceChannel eval | One-off, reused twice |
| Editorial magazineAnalytics writeup | Barlow Condensed headlines over Work Sans body, full-bleed featured image | Social media traffic report | One-off |
| Terminal technicalDark diagram | JetBrains Mono throughout, slate background with cyan and emerald accents, SVG workflow map | BD dashboard | One-off |
| Presentation deckBuilt to read across a room | Segoe UI at large sizes, MNA blue on white, multi-page nav, sized for 1920x1080 | BD onboarding site | One-off |
Six systems for ten pages is drift. The branded report style now covers half of them and is the only one that survives as a skill, so anything new defaults to it. The deck, dashboard, and onboarding styles each earn their difference; the Aptos vendor-eval style and the editorial report style overlap with the house system and are the two candidates to retire.
Who did what
Four systems, four distinct roles. The split was not planned, it emerged from what each tool was good at and where its logs show the work landing.
| # | System | Prompts | Sessions | Active days | Role it actually played |
|---|---|---|---|---|---|
| 1 | Claude CodeCLI + desktop | 149 | 43 | 15 | The execution engine. Owned the PA NCF Mailcoach campaigns end to end, all five new account folders, the vendor evals, the weekly BD report build and deploy, and the branded HTML report system itself. |
| 2 | HermesAgent gateway | 27 | 10 | 11 | The intake and reminder layer. Forwarded emails turned into tasks, calendar and callback reminders, state licensing lookups, template retrieval, and the Nemotron lead-list prompt handoff. |
| 3 | Gemini / Antigravityagy CLI + IDE | 15 | 11 | 6 | Recruitment marketing. Job PDF intake to carousel, the Vermont travel nurse concept, LinkedIn content batches, grant searching, and staffing services copy. |
| 4 | Codexcodex-cli | 10 | 6 | 3 | Corporate and financial lane. MSP tracking log review, the SAM.gov re-registration playbook, business financing research, and the Diane custom GPT scaffold. |
One Claude Code prompt can run an hour of agent work, and a Hermes email intake is a single line. Read the column as "how often you steered this system," and read the workstreams below for what each one actually produced.
Method and limits
What this report is built from, and what it cannot see.
A session counts as MNA work if it ran inside an MNA directory or its prompt matched MNA terms: the brand name, travel nursing, Mailcoach, Attio, NCF, carousel, recruiter, or per diem. MNA work phrased without any of those terms is undercounted. Duplicate and system-injected prompts were removed before counting.
The Antigravity CLI stores full conversations in per-session binary databases that are not readable as text. The 15 Gemini entries come from its prompt history file and conversation summary index, so its real volume is higher than the table shows, particularly for the carousel work in week one.
Most sessions ran from the home directory rather than inside /mna, so directory alone could not be used to attribute work. File modification times under /mna were used as the corroborating evidence for every workstream listed above.
~\.claude\projects\*\*.jsonl
~\.codex\sessions\2026\07\*\rollout-*.jsonl
~\.gemini\antigravity-cli\history.jsonl
~\.gemini\antigravity-cli\conversation_summaries.db
~\AppData\Local\hermes\state.db
C:\Users\moveb\mna file modification times