Bratha

Part 3.0 — The Workflow Engine

10 min read

This is Part 3 of 7 in the Productivity Enhancement Series

Table of Contents

The loop that runs everything

With the machine and the room solved (Part 2.0), the next question is what goes on those screens, and the answer is a workflow: the loop that catches every commitment and moves it to done. This is the classic capture–clarify–organise–reflect–engage loop from David Allen’s Getting Things Done,1 with one rename: the last step is Execute, because “engage” is vague and the whole point is that you actually do the work. The loop's only job is to make sure nothing lives in your head and nothing falls through the cracks. That’s the Forgetfulness and Mindlessness defence from Part 1.1, turned into a routine. Practical-first as always: here’s how to run each step, then where AI multiplies it.

Step 1: Capture

Capture means: the instant a commitment, idea, or task appears, it leaves your head and lands in a trusted inbox. Not later, not “I’ll remember it,” not as an unread email kept unread as a reminder. Immediately, and into one place you trust completely.

The rule is ruthless: if it isn't captured, it doesn't exist, and your brain knows it doesn't exist, which is why it keeps anxiously reminding you. Capturing relieves that background load (the open-loop tax from Part 1.1).

  • Pick one inbox. A single notes app, a single task app, even a single text file. The failure mode is many inboxes (some in your head, some in email flags, some on sticky notes), because then no place is trusted and you have to check all of them.
  • Make capture faster than the urge to skip it. A global hotkey, a phone widget, a voice memo. If capturing takes more than a few seconds, you won’t do it under pressure, which is exactly when it matters.
  • Capture raw. Don’t organise at capture time. “Email Faiz re: invoice” is enough; sorting comes later. Mixing capture and organising kills both.

The AI multiplier on Capture

The frontier here is capturing by talking. A voice memo to an AI assistant that transcribes and drops the item into your inbox means you can capture while walking, driving, or mid-conversation. Even better: forward a messy email or a long voice note and have the assistant extract the actual commitments into clean inbox items. The fundamental (get it out of your head, now) is unchanged; the friction just dropped to near zero.

Step 2: Clarify

Clarify means: for each captured item, decide what it actually is and what the very next physical action is. This is the step everyone skips, and skipping it is why inboxes rot into a list of vague nouns (“taxes,” “Mum,” “website”) that you reread guiltily and never act on.

For each item, ask in order:

  1. Is it actionable? If no: trash it, file it as reference (it goes to the Knowledge System), or park it as a someday/maybe.
  2. Is it one action or a project? If it takes more than one step, it’s a project (handled in Part 6.0) and it needs a defined next action.
  3. What is the next physical action? Concrete and physical: not “website,” but “draft the homepage headline.” Not “taxes,” but “find the 2025 receipts folder.” A next action you could start in two minutes if you sat down; that's the test.
  4. Two-minute rule: if the action takes under two minutes, do it now. It costs more to track than to finish.

This is the step that kills Mindlessness, because it forces a moment of intent between capturing a thing and acting on it. You decide what matters before you touch it, instead of drifting to whatever’s loudest.

The AI multiplier on Clarify

Hand the assistant a raw inbox item and ask it to propose the next physical action, flag whether it’s really a project, and draft the two-minute version if there is one. You stay the decider (it proposes, you approve), but the cognitive work of turning “website” into a concrete first step gets drafted for you. This is the first place the self-improving workflow kicks in: teach it once how you like next actions phrased, and it keeps doing it your way.

Step 3: Organise

Organise means: put each clarified item where it belongs so the right things surface at the right time. Clarify decided what each thing is; organise decides where it lives.

  • Next actions go on lists, ideally split by context: what you can only do at your computer, what needs a phone call, what needs you to be out running errands. Context lists mean that when you have ten phone-free minutes, you look at the one list that fits the moment instead of scanning everything.
  • Calendar is for time-specific commitments only. Appointments and hard deadlines. Don’t pad your calendar with aspirational task-blocks you’ll just drag forward; that trains you to ignore your own calendar.
  • Projects get a one-line outcome and a single visible next action each (more in Part 6.0).
  • Reference material leaves the workflow entirely and goes to the Knowledge System (PARA). The workflow is for things you’ll do; the knowledge system is for things you’ll consult. Keeping them separate is what stops either from becoming a swamp.

The AI multiplier on Organise

An assistant with access to your task system can route items for you: tag by context, attach the project, set the due date you mentioned in the capture. Pair it with AI-assisted calendaring that proposes where a task-block fits around your existing commitments. Again, the fundamental (one correct home per item) is old; the sorting labour is what gets offloaded.

Step 4: Reflect

Reflect means: regularly step back and review the whole system so you trust it. Without this step the whole loop quietly dies, because an inbox you don’t review is an inbox you stop trusting, and a system you don’t trust sends you right back to keeping things in your head.

  • The Weekly Review is the keystone. Once a week: empty every inbox to zero, review every project for a valid next action, scan your calendar forward and back, and clear the stale stuff. An hour, protected, non-negotiable.
  • A 2-minute daily glance at today’s calendar and your shortlist keeps the day pointed in the right direction.

Reflect is the Mindlessness defence at the system scale: it’s the scheduled moment where you ask “am I working on the right things?” instead of just “am I working?” A workflow without a review is just a faster way to stay busy.

The AI multiplier on Reflect

This is one of the highest-value places for a scheduled AI task (a preview of Part 7.1). An assistant can run every Sunday: summarise what got done, list projects with no next action, surface items that have sat untouched for two weeks, and hand you a tight agenda for your review instead of a blank page. You still make the judgement calls; it just does the gathering.

Step 5: Execute

Execute means: actually do the work, choosing what to do by a simple, honest set of filters. All the previous steps exist to make this moment frictionless: when you sit down, you should be choosing from a small, trusted, clarified list, not negotiating with a pile of guilt.

Choose the next action by, in order: context (what can I even do right now, here, with this energy?), time available, energy level, and then priority. The reason context comes first is practical: there’s no point eyeing a high-priority task that needs a quiet office when you’re on a train with ten minutes. ==Pick the most important thing you can actually do right now, then do only that.==

And then single-task. The workflow’s entire job was to make it safe to ignore everything else, because everything else is captured and will be handled. That safety is what lets you give one thing your full attention, which loops straight back to the Distraction defences in Part 2.0.

The AI multiplier on Execute

Execution is where AI stops being an organiser and becomes a doer: drafting the email you clarified, writing the first version of the document, doing the research pass. Crucially, you hand it work that’s already been clarified and organised, which is why those steps came first. An AI given a vague task produces vague slop; an AI given a clarified next action produces a usable draft. How to hand off entire repeatable jobs (not just single drafts) is Parts 6 and 7.

The whole loop on one page

StepThe one-line jobKillsAI multiplier
CaptureGet it out of your head into one trusted inbox, instantlyForgetfulnessVoice capture; extract tasks from messy input
ClarifyDecide what it is + the next physical actionMindlessnessDrafts the next action, flags projects
OrganiseOne correct home per item (context lists, calendar, projects, reference)NoiseAuto-tags, routes, schedules
ReflectWeekly review so you trust the systemMindlessnessScheduled weekly digest
ExecuteDo the most important thing you actually can, single-taskedDistractionDrafts and does the clarified work

The loop runs continuously: capture all week, clarify and organise as you go, reflect weekly, execute daily. Get this turning reliably and you’ve built the engine. The next two layers (Knowledge and the projects on top) just feed it better fuel. And the moment any step in this loop is something you repeat, it’s a candidate to be taught to the system once and improved forever, which is Part 3.1.

Part 3 Takeaways

  • The workflow is one loop: Capture → Clarify → Organise → Reflect → Execute. Its only job is that nothing lives in your head and nothing falls through.
  • Capture to one trusted inbox, instantly, raw. If it isn’t captured, your brain won’t let it go.
  • Clarify is the step everyone skips: decide what each item is and its next physical action. This is the Mindlessness defence.
  • Organise gives each item one correct home; keep “things to do” and “things to consult” strictly separate.
  • Reflect weekly or the system dies. A workflow without a review is a faster way to stay busy.
  • Execute by context first, then time, energy, priority, and single-task because everything else is safely captured.
  • AI multiplies every step, but only after clarify and organise have been done. A vague task in gives slop out.

Your Workflow Task List

This week

  • Choose your one inbox and set up a sub-5-second capture method (hotkey, widget, or voice).
  • Do one full clarify pass: turn every vague noun in your current lists into a concrete next physical action.
  • Strip your calendar back to time-specific commitments only; move aspirational task-blocks to context lists.
  • Book a recurring Weekly Review (one protected hour) on your calendar now.
  • Pick one repeatable step and teach an AI assistant how you like it done, then read Part 3.1 to make that lesson stick.

Sources & references

Footnotes

  1. David Allen, Getting Things Done: The Art of Stress-Free Productivity (2001; rev. 2015). The five-stage workflow (capture, clarify, organise, reflect, engage), the two-minute rule, the next-action principle, context lists, and the weekly review all originate here. This article keeps the structure and renames the final stage “Execute.” The underlying reason capture relieves mental load is the open-loop / Zeigarnik effect referenced in Part 1.1. ↩

Graph View

Backlinks