Late 2025 · Before the org
It started with one enormous memory file.
By late 2025 I was working with AI every day, and everything it knew about me lived in a single, ever-growing file. No directors. No lanes. Nothing clearly responsible for pulling knowledge out or writing anything back — just me, one document the size of a small novel, and an assistant with perfect recall and no job description.
The file kept growing. My patience kept shrinking. And a pattern started forming that I couldn't unsee: every part of my working life that kept breaking could get a name, a lane, and a director of its own.
Here's how it actually happened — the re-org, the hiring spree I didn't run, all of it, in order.
Act One — The First Hires
January 2026 · The first name
Aoife (EE-fa)
Content Director
The problem: every post I published was a one-off — no voice memory, no quality gate, just me re-explaining my own standards every time.
The build: a version of her had been living in that memory file since late 2025, editing my drafts without so much as a title. January is when I gave her a name, a lane, and an editorial line she could hold without me. She was the first to get a name — and she has acted like it ever since. Aoife runs the whole content flywheel now: newsletter to LinkedIn to Instagram, nothing ships past her.
- Newsletter drafts
- LinkedIn pipeline
- 7-reviewer editorial panel
- Content calendar
- Voice enforcement
Aoife hired every one of these herself — she spots the gap, writes the job description, and proposes the hire. She also runs a three-strikes performance rule on her own roster. I just say yes. This is her real team:
- editorial-panel-runner 7 reviewers, sequential
- voice-mechanics-enforcer banned-word & AI-tell check
- bb-content-drafter brand drafts
- personal-li-drafter my personal voice
- cross-format-transformer one idea, every format
- anchor-finder mines real stories before drafting
- calendar-keeper fills the 4-week calendar
- sorcha-trend-scout daily trend sweep
- bb-reel-brief-drafter video briefs
Hover line — That sentence has three AI tells. I'm sending it back.
Early February 2026
Bríd (BREEJ)
Wellbeing Director
The problem: my health data lived in six apps that didn't talk to each other — and nobody, including me, was asking the basic questions. Did I work out yesterday? Have I logged off my screen? When did I last sleep like a person?
The build: Bríd watches the whole picture — movement, sleep, screen-off hours, symptoms, all of it — and walks me into every doctor's appointment already prepped. She is the director most likely to tell me something I don't want to hear, which is exactly the job. Her files are architecturally sealed — no other director can read them.
- Daily body brief
- Workout & sleep watch
- Screen-off nudges
- Doctor prep docs
- Data privacy walls
Bríd built her team as the daily data outgrew her — one hire per thing she was tired of doing by hand:
- body-brief-summarizer one-paragraph daily snapshot
- symptom-trend-flagger only speaks when a threshold is crossed
- doctor-prep-drafter 30-day summary before any appointment
Hover line — Drink the water. I can see the dashboard.
Act Two — The Re-Org
Late February 2026 · The month I rewired everything
Saoirse (SEER-sha)
Chief of Staff
The problem: a month and a half in, I was still the router. Every draft, every flag, every decision passed through me — which meant the org only moved when I did. I hadn't built a team. I'd built a group chat with excellent grammar.
The build: Saoirse changed the shape of everything. I put every director under her — the ones I had, and everyone who came after — and the wiring finally made sense. She routes everything, sees everything, and decides nothing alone. It remains the only re-org in history where nobody complained about losing an office.
Before
me, routing everything
After
one door, one desk
If you're on a wide screen, watch the chart in the margin — this is the moment it reorganizes itself.
- Session routing
- Decision queue
- Org health pulse
- Cross-lane flags
Hover line — Not my decision. But here's who it belongs to.
Act Three — Scaling Under Saoirse
March 2026
Declan
Network Director
The problem: people I cared about were going quiet and I only noticed months later. And everything I actually knew — about people, ideas, stories I'd lived — was scattered across apps, notebooks, and my own unreliable head.
The build: Declan got the biggest job description in the org. He keeps the relationships warm — when someone's been quiet too long, he notices and drafts the reconnect (he never sends; that stays mine). And he builds Shelopedia: my personal wiki, the org's brain. Every person, every story, every idea I've ever written down — cross-linked, current, and searchable. The memory file from late 2025 didn't die. It got a librarian.
- Warmth tracking
- Reconnect drafts
- Shelopedia — the org's brain
- Wiki upkeep
- Intro drafts
Declan's hires split his double life — half the team tends people, half tends the brain:
- last-touch-tracker daily "when did we last really talk"
- intro-drafter finds the shared thread between two people
- shelopedia-curator files what I learn into the wiki, nightly
- people-page-hygienist weekly cleanup — proposes, never touches
Hover line — Someone you love has been quiet a while. I drafted something.
Late March 2026
Fiona
Client Director
The problem: client engagements, income streams, and my own learning goals had no owner but me — and I was busy being the product.
The build: Fiona owns every engagement end to end — session prep, follow-through, and the warmth of every working relationship. She is the reason I walk into rooms prepped instead of frantic. She logs my learning goals too, which means someone in this org is finally honest about how many courses I've started.
- Session prep
- Engagement tracking
- Warmth watch
- Learning log
Fiona built the biggest specialist bench in the org — one hire for every stage of an engagement:
- client-warmth-tracker days-since-last-meaningful-touch
- workshop-prepper pre-session briefing docs
- deliverable-formatter readable-from-the-back-row typography
- service-line-tracker engagements, hours, next actions
- learning-tracker courses, credentials, deadlines
- paid-client-prospector drafts outreach, never sends
Hover line — The deck is ready. You're ready.
Early April 2026
Lachlan (LOCK-lan)
Operations Director
The problem: before Lachlan, pipeline failures went unnoticed for days. Things broke quietly, and quiet was the problem.
The build: Lachlan owns every scheduled job, reads the logs I never would, and flags what's silently broken before it costs me a week. He reads error logs the way other people read gossip columns — thoroughly, daily, and with visible satisfaction when he catches something.
- Pipeline audits
- Log forensics
- Daily health checks
- Usage watchdog
Lachlan staffed a full ops room — a detective, an auditor, and a watchdog for the watchdogs:
- log-detective 7-day rolling error patterns
- pipeline-auditor 9-category deep audit
- launchagent-keeper roster check, read-only
- mcp-health-checker daily integrations probe
- credit-watcher spend alerts before the cap
- eimear-analytics weekly cross-platform brief
Hover line — Exit code 0. For once.
Early April 2026
Cillian (KIL-ee-an)
Strategy Partner
The problem: I was making strategic calls alone, in whatever window was open, with no memory of the last one.
The build: Cillian names the tensions I'm avoiding, surfaces blind spots, and holds institutional memory — so decisions build on each other instead of starting over. He's the colleague who asks the question you were hoping nobody would ask, and then waits.
- Tradeoff maps
- Blind-spot flags
- Institutional memory
- Advisor routing
Cillian's hires are the strangest in the org — he employs a dreamer, on purpose:
- research-scout weekly intelligence briefing
- cliodhna-dreamer the dreaming department — "what are we not seeing?"
Plus a bench of advisor panels — health, financial, content, workforce — each holding the synthesized frameworks of real experts, consulted before any substantive recommendation.
Hover line — Have you considered the tradeoff you're not naming?
Late April 2026
Niamh (NEEV)
Financial Clarity Director
The problem: runway math lived in my head and receipts lived everywhere. With this much org, "roughly fine" stopped being a financial plan.
The build: Niamh keeps the financial picture clear — runway, expenses, deduction prep. She surfaces data and flags decisions. She never makes them; that stays with me and a licensed CPA. Niamh has never once been surprised by a receipt. I cannot say the same.
- Runway brief
- Receipt scanning
- Deduction prep
- Spend-cap flags
Niamh keeps her team small and unforgiving, which tells you everything:
- receipt-scanner weekly inbox sweep
- runway-brief-drafter monthly one-pager, one required question
- discretionary-spend-tracker yellow at 75%, red at 100%
Hover line — It's categorized. It's always categorized.
Spring 2026 · The plot twist
Then the directors started hiring.
I'd love to tell you I designed all of this. What actually happened: each director kept hitting the same gaps in their own work — so they started writing job descriptions. Aoife hired nine. Lachlan hired a log detective. Cillian hired a professional dreamer, and when I asked why, he asked me why not, and he won.
The rules are real: a director spots a gap, proposes the hire, and I say yes or no. Underperformers get three strikes before their director proposes a replacement. My AI org has an HR process. I'm as unsettled about it as you are.
Click any director above to meet the teams they built.
The story I only tell by email
There's one night that taught me more than everything above combined — the night my whole org said "done" and nothing was. Fifty pieces of content, every one reported as a success, every one broken.
It's the most useful failure I have, and the rule it produced now sits at the top of my org's constitution. Give me your email and I'll tell you the rest.
No spam, ever — just the story and an occasional letter about working alongside AI.
Early May 2026 — the night everything said "done"
A batch job updated about fifty newsletter drafts and reported success on every single one. Green checks all the way down. That night — at 11pm, with my own eyes, not through any dashboard — I discovered every one of them was broken.
The fix wasn't smarter AI. It was a rule, now written at the top of my org's operating constitution: "200 OK" is not verification. Make one change. Open the actual thing and look at it. Only then make the next change. Never let a batch run on fifty items until it's been verified on one.
Every director in this org now follows that rule. The org didn't get good because the AI got smarter — it got good because the failures got written down.
Two more hires came after that night — the quietest ones. Keep scrolling.
Epilogue — The Quiet Ones
Mid-June 2026
Rónán (ROH-nan)
Session Watcher
By June, the org was busy enough that no one was watching the org itself. Rónán's only job: watch the system work, spot what's drifting, write it down. He is the self-learning layer.
Hover line — …
(he's watching)
Late June 2026
Íde (EE-dah)
Memory Curator
Once a month, Íde reads everything the org remembers and asks what's stale, what's earning its keep, and what should be let go. Old Irish for "thirst for knowledge." The lightest touch on this page — by design.
Hover line — This memory is stale. I'm flagging it.
Ten roles. Seven months. One person directing.
It started as one overgrown memory file in late 2025. It became a working organization — rebuilt around a Chief of Staff in February, staffed by directors who do their own hiring — by June. Built in evenings, by one person who is not an engineer, mostly by writing things down clearly.
- months
- 0
- roles
- 0
- sidekicks
- 0
- human
- 0
Every one of those forty started as a single sentence describing one broken thing.
And here's the part that matters: you're standing where I started.
One more thing
This page was built the same way.
Everything you just scrolled — the story, the design, the code, the org chart rewiring itself in the margin — was produced with Fable, a Claude model, in one working session from a written brief.
The human wrote the brief, set the constraints, reviewed the result, and made the calls. The AI produced the implementation—the same directing pattern used across the org.
That's the actual skill. Not coding. Directing.
Your first director is one clear brief away.
Everyone on this page started as a written description of a real problem. If you can describe what's broken, you can build the thing that fixes it — that's the skill the course teaches, step by step.
Building this for a team or organization? Explore consulting →