The week it clicked.
One real story of the system earning its keep — being picked and written up.
Six Sigma's DMAIC is the spine: Define, Measure, Analyze, Improve, Control. My ten build phases sit inside it. Each stage below names the goal, the phases, the classic tools, and what AI adds. Pick a stage; open only what you need.
Stage 1 of 5 · Define
From my houseMy whole team is basically my SIPOC, with names. And every rock on my 90-day plan names the theme it serves.
Every system optimizes for something. If you don't name it, it optimizes for busy. Write three short things — in pencil. You'll revisit them, and your AI will read them before it helps you with anything.
ShapeA few sentences, present tense. Where you are, who's around, what you're spending your hours on — and what you're not.
ShapeOne sentence you could say out loud without cringing. Seasons change; so can this.
ShapeThree to five words, each with a one-line “this means…”. These become your filters — the rules the system checks against.
From my houseTwo of my filters show up everywhere in this system: love as the filter — if it crowds the people I love, it doesn't ship — and Sundays for being present, away from the tech. Those two lines shaped more of the design than any tool choice.
Somewhere private and quiet (mine live in a local notes vault), plus a short, shareable version your AI can read as context. The long version is for you. The short version is for the system.
Vision is too big to act on, and a to-do list is too small to steer by. A macro plan sits in between: three or four year-level themes (big rocks). The 90-day plan pulls from those, and the week pulls from the 90 days. Lean teams call this a hoshin cascade. I call it “every small thing points up.”
From my houseEvery rock on my 90-day plan names the theme it serves. If it can't name one, it waits for next quarter.
Print the blank →The catchball part: each level checks with the one above. If a rock can't name the theme it serves, it goes back up for a rethink.
Borrowed from continuous improvement: five columns that show where your life's work comes from and who it's for. Fill it in once, on one page. It's the map everything else hangs on.
Why this comes earlyBefore AI, the SIPOC tells you what's worth automating. After AI, it tells you what a specialist is for. My whole team is basically this page, with names.
The C column tells you who. This grid tells you what each of them actually needs from you, how often, and how you'll keep them in the loop. It's where “be more present” turns into something you can plan around.
| Who | What they need from me | How often | How I keep them in the loop |
|---|---|---|---|
| Family | Presence, a calm plan, me showing up for what matters | Daily | Shared family calendar and a real check-in at dinner |
| Spouse | The same picture of the week, and no surprises | Weekly, plus as needed | A week-ahead sync and one shared list |
| Work team | Clear priorities, follow-through, quick answers | Workdays | A calendar that's true, short status notes, drafts held for my yes |
| Close friends | Being remembered, and time together | Every few weeks | Gentle reach-out reminders; voice notes, not form letters |
| Guests we host | A warm welcome, fast answers, a smooth stay | Every stay | Messages drafted in my voice and approved before they go |
| Myself | Rest, movement, time to think | Daily | A morning page, and a day kept for being present |
From my houseThis grid is also a spec for the bots. The “how I keep them in the loop” column is basically a list of jobs a specialist can draft for me.
Print the blank →Next: turn each need into something you can check. Voice of the Customer → CTQ ↓
Six Sigma calls this the Voice of the Customer. You write down what people would actually say, find the need underneath, and turn it into a CTQ (critical to quality): a plain, checkable standard for “good.” It's how “be more present” becomes something the system can help with.
| Who | What they'd say | What they really need | CTQ: how I'll know it's good |
|---|---|---|---|
| Family | “Put the phone down at dinner.” | Presence at meals | Dinners without screens, most nights |
| Spouse | “I didn't know that was this week.” | The same picture of the week | Week-ahead shared by Saturday evening |
| Work team | “Did you see my note?” | Follow-through | Replies within one workday; prep ready before key meetings |
| Close friends | “We never see you!” | Being remembered | A reach-out every few weeks, in my voice |
| Guests we host | “How do we get in?” | Fast, warm answers | Replies within the hour during the day; five-star stays |
| Myself | “I need a minute to think.” | Rest and thinking time | One day a week kept free; quiet minutes most mornings |
GuardrailThese customers are people I love. AI can suggest a CTQ; I decide what “good” means.
Print the blank →From my houseShine is the step I used to skip. Once it became a standing nudge, Downloads stopped being a junk drawer.
One cloud source of truth, three to five lanes inside it, and a rule for the junk drawer. AI is only as good as what it can find.
The Downloads rule is the one people skip and regret: nothing lives in Downloads.
On a Lean shop floor, 5S is why every tool has an outline painted on the wall. The virtual version does the same for files, inboxes and notes, and it's the reason an AI can find anything at all.
Delete or archive what you don't need.
Old screenshots, duplicate downloads, apps you never open.
One home for each kind of information.
Documents → cloud lanes. Ideas → capture inbox. Dates → calendar.
A weekly clean-up of Downloads, desktop and inbox.
Fifteen minutes, same time every week.
Names and places anyone (and any bot) can predict.
Date · lane · what it is.
A routine keeps it tidy, so willpower doesn't have to.
A weekly nudge from the chief of staff: “Shine time?”
Date first so it sorts. Lane second so it files itself. Plain words last so search finds it.
From my houseShine is the step I used to skip. Once it became a standing nudge, Downloads stopped being a junk drawer.
Print the blank →One low-friction inbox for thoughts: a wearable, a voice note, a single notes app. Voice beats typing. Whatever you capture should sync somewhere searchable. In CI terms, the capture inbox is your data collection point: if it isn't captured, it can't be triaged, remembered or measured.
It transcribes and syncs in the background, so capture costs a sentence, not a session. Captures land in the memory archive, and the morning sweep gives each one a lane.
A baseline is two ordinary weeks, measured honestly, before you fix anything. The data collection plan says exactly what counts (an operational definition), where the number comes from, and who collects it. Callie's time audit is my baseline for where the hours go.
| Measure | Operational definition: what counts | Source | Who collects | Baseline |
|---|---|---|---|---|
| Time where I meant to | Share of scheduled hours tagged to a stated priority | Calendar tags | Callie | Example: 58% |
| Inbox at a calm level | “Needs you” pile is five or fewer at 5 pm | Inbox counts | Emmy | Example: 3 of 7 days |
| Decisions waiting on me | Items in “Waiting on you” older than 24 hours | The status board | Ford | Example: 4 |
| Routines on time | Routines that ran within 15 minutes of their time | Heartbeat log | Ford | Example: 90% |
GuardrailWrite the definition before you look at the number. Otherwise the number quietly changes meaning.
Print the blank →Two numbers Ari tracks for work actions are a nice first taste: cycle time (an action opened → closed) and carryover (items rolled into next week, the personal version of work-in-progress). They measure how smoothly work moves, not how busy anyone looks. AI timestamps every action, so there's no stopwatch.
Try thisWalk one real week before you automate anything. You can't fix a step you've only imagined.
Lean says go to the gemba: look at the real work, not your memory of it. Here that means a real week of screenshots, inbox and calendar. Walk the daily loop as a value stream, one step at a time, and tag anything that doesn't serve a customer from your SIPOC.
One low-friction inbox for thoughts. A wearable, a voice note, a single notes app.
What is this? Trash, someday, or real — and which lane?
Yes, no, not now, or hand it to someone. The human step.
Do it, delegate it, or schedule it. Drafts come ready to approve.
Back to its home. Review weekly so nothing rots.
| Waste | What it looks like at home | The fix (and where AI helps) |
|---|---|---|
| D · Defects | A draft that has to be rewritten; a reminder set for the wrong day | The PDCA loop learns from every edit |
| O · Overproduction | Briefs and reports nobody reads | Retire a routine nobody opens |
| W · Waiting | Decisions stuck on me; a reply waiting on one detail | “Waiting on you,” cleared within 24 hours |
| N · Non-utilized talent | Me doing sorting a routine could do; ideas I never capture | Hand the sort to a specialist; capture by voice |
| T · Transportation | Copying details between apps; forwarding things to myself | One home, quiet sync |
| I · Inventory | Unread newsletters, a full Downloads folder, open loops | Unsubscribe pile; the Downloads rule |
| M · Motion | Hunting for a file; checking five apps for status | One glance page |
| E · Extra processing | Re-reading the same email three times; polishing a quick yes | Four piles; decide once |
GuardrailLook at real data, redacted. Private details stay out of anything shared.
Print the blank →A Pareto chart is a tally sorted biggest first. Usually a few causes make most of the mess. AI does the tally across a month of email in seconds, so you fix the top few instead of everything.
The readTwo kinds of email were about two-thirds of the pile. Emmy's Trash and Unsubscribe piles handle both, and I approve them in one tap.
Print the blank →When something slips, the Toyota habit is to ask “why?” until you reach a cause you can actually fix, instead of promising to try harder. A fishbone helps when there are many possible causes: sort them by category first.
AI can suggest causes and gather the evidence for each. I choose the root cause and the fix.
The last question before you improve anything: which steps are clock work, which are one-offs, and who decides? This table is the output of the analysis, and the spec for everything in Improve.
Things that happen on a schedule, every time, without anyone asking: the morning brief, the inbox pass, the week-ahead. These are where AI shines — reliable and boring on purpose.
A trip, a big rewrite, a new project. These get a brief, a deep-work session, and a finish line — then they're done. Don't turn a one-off into a routine just because you can.
From my houseDraft-and-hold is a little slower than full autopilot. That's the point: the tap is where my judgment lives.
Here's how the pieces actually work together — including the unglamorous parts. I talk to one chief of staff. It routes to specialists, who draft and report back. Nothing leaves the house without me.
One chat you actually talk to. It briefs you, routes asks, and speaks human so specialists don't scatter noise.
Narrow lanes, deep context. Start with one — usually email. Add the next only when the first is boring.
One page for the week, what's waiting on you, and the links. The calm glance layer — not another inbox.
For one-offs: a rewrite, a plan, a page like this one. Claude, in my case. Deep work, not daily ops.
Standard work is the current best way to do a job, written down: when, what, and how you know it's done. It's what makes the loop reliable on a tired Tuesday. It's also what lets a routine run a step for you, because you can't automate a step you haven't defined.
From my houseMy morning brief started as a hand-written card like this. Once the steps stopped changing, it became clock work.
Print the blank →Design the process so the mistake can't happen. Specialists produce drafts, not sends. Nothing sends, spends, books or deletes without my yes, so a wrong draft costs a tap, not an apology.
Toyota's looms stopped themselves when a thread broke. Mine do the same: when a specialist is unsure, it stops and puts the item in “Needs you.” When a routine breaks, it fails loud.
Kanban makes work visible by status and caps how much can be in progress. My “Waiting on you” card is a kanban column with a small limit: when it's full, specialists hold new drafts and Ford tells me. Nothing new lands on me until I clear something.
Why a limitWithout one, drafts pile up and the tap stops being a real decision. A small limit keeps my yes meaningful.
Print the blank →Heijunka means spreading work evenly instead of in spikes. For a week, that's deep work, admin, people time and home tasks spread across the days, so no single day breaks and Sunday stays free. Callie proposes where things fit and flags an overloaded day; she never moves people without my yes.
GuardrailSundays are for being present, away from screens. The level week is built around that, not the other way round.
Print the blank →Narrow lanes, deep context. Start with one specialist (usually email) and add the next only when the first is boring. Each lane gets a RACI row before it gets a single task: specialists can own the work, never the yes.
Everyone has inbox pain, so this is the one to copy first. The goal isn't inbox zero for its own sake. It's a calm inbox — without AI freelancing your relationships.
A one-off, start to finish: how a trip goes from “we should” to booked and briefed, and which moments stay human. Being written up.
Lean's A3 puts a whole problem on one page: background, current state, goal, plan, check. The mind map is how I fill it in fast.
Before a big one-off goes to Claude, I spend ten minutes mapping it: the topic in the middle, the questions around it. The branches become the brief's headings, and a good brief is why the session ends with a finish line instead of a new routine.
From my houseI hand Claude the map, not a blank prompt. The “keep out” branch saves the most rework.
Print the blank →From my houseThe goal is a mostly green page. Red is rare on purpose, so when it shows up I actually look.
Control is where most improvement efforts quietly die: the new way slides back to the old way. Classic CI holds the gains with visual management, a control plan and a steady routine of check-ins. AI is what makes that affordable at home, because it does the checking.
A Six Sigma control plan lists each key measure with its normal range, who checks it, the signal that means “act,” and the reaction. It turns the Dashboard from pretty tiles into a system that responds.
| Measure | Normal range | Checked by | Signal: act when… | Reaction |
|---|---|---|---|---|
| Routines on time | 90–100% | Ford · daily | Below 90% two weeks running | Heartbeat check; fix the routine or retire it |
| Decisions waiting on me | 0–3 open | Ford · daily | Anything older than 24 hours | Moves to the top of the morning brief |
| Inbox at a calm level | 5+ of 7 days | Emmy · daily | Three rough days in a row | Re-run the Pareto; propose a new filter |
| Time where I meant to | 65% or more | Callie · weekly | Below range three weeks in a row | Monthly look-back picks one small change |
| Money follows the plan | KPI being set | Penny · weekly | Monthly money review not done yet | Chat flag; we refine the plan |
| Sundays kept free | 4 of 4 | Callie · monthly | One missed Sunday | Ask why; protect the next one first |
GuardrailReactions that leave the house still wait for a yes.
Print the blank →Plan, Do, Check, Act: the oldest loop in continuous improvement. Here it runs on the draft-and-hold flow. When I review a draft, what I change, what I cut and what I reject doesn't just vanish. It gets captured and saved as style notes, rules or memory, so the next draft needs less from me.
The draft or proposal starts from my style notes, rules and memory, not a blank page.
Yes, edit or no. The tap where my judgment lives.
It compares what it drafted with what I sent: tone I softened, lines I cut, drafts I turned down.
The lesson is saved as a style note, a rule, or a remembered preference.
| What I said (or did) | What it learns | Saved as |
|---|---|---|
| “Warmer, shorter.” | Drafts to friends run too formal; keep them to three sentences | Style note |
| “Never say ‘circle back.’” | A phrase that's banned in anything with my name on it | Rule |
| “Always cc my spouse on house stuff.” | Who to loop in, by topic | Preference |
| Rejected a draft: “this one's mine to write.” | Some messages stay fully human; flag them, don't draft them | Rule |
| Swapped the sign-off | “Warmly” for people I know, “Thanks” for everyone else | Style note |
From my houseThis is the “feed it so it learns you” principle, made mechanical. Two weeks of honest edits did more for my drafts than any prompt I wrote.
Print the blank →Kaizen is the same loop at the scale of the system: one small improvement at a time. New routines get a two-week pilot (boring and reliable) before they become clock work.