25d • Plug-and-Play
Design your own digital co-workers
Stop trying to become a developer.
That sentence alone has probably saved me more time, energy, and sanity than any AI hack or automation trick I’ve ever used.
Because here’s the quiet pressure a lot of us feel right now:
> “If I want to stay relevant, I guess I need to learn to code…”
> “If I want to use AI ‘properly,’ I probably need to become some kind of prompt engineer…”
> “If I want leverage, I should be building apps.”
I don’t buy that.
My whole mission with Citizen Developer is simpler and way more practical:
> You don’t need to become a programmer.
> You need to learn how to design your own digital co-workers.
Not “AI toys.”
Not another dashboard no one opens.
Actual digital workers that sit alongside you, take a slice of your workload, and report back to you with receipts.
In this post, I want to show you how that starts.
And it doesn’t start with “learn Python.” It starts with a Personal OS and one question:
> “If I had a tiny team of AI co-workers, what would I actually have them do for me this week?”
---
Stage 1: Personal OS – AI-First, Not Codemonkey
Before we talk about browser agents, automations, or NotebookLM, I want to strip this all the way down.
If you’re going to build a digital workforce, you need a lightweight operating system for how you use AI day to day.
Not a complicated stack.
Not a Notion template with 97 properties.
Just four pieces:
1. One conversational model
ChatGPT, Claude, whatever you like.
This is your thinking partner: your writer, explainer, rubber duck, and first draft machine.
You talk to it the way you wish you could talk to a patient colleague who never gets tired of your “one more question.”
2. One browser agent
Fellou, Delegat8, or any computer-use / browser agent you’re comfortable with.
This is your hands:
* It clicks around websites.
* It scrolls, copies, pastes, downloads.
* It pulls out the boring data you don’t want to babysit.
When I say “digital co-worker,” a browser agent is a big part of that picture. It’s someone at the computer, doing the tedious clicks, following your instructions.
3. One automation hub
Make, Zapier, n8n – pick one.
This is your nervous system:
* It watches for triggers (new emails, form submissions, calendar events).
* It runs your steps in order.
* It passes information between tools without you as the middleman.
You’re not “coding.” You’re dragging little blocks into a flow that says:
> “When *this* happens, send *this* info *there*, then do *these* three things.”
4. One personal knowledge base
Notion, Obsidian, Google Drive – doesn’t matter.
This is your memory:
* SOPs
* Logs
* Templates
* Meeting notes
* “Things I don’t want to figure out twice”
That’s your starter kit.
You don’t need twelve tools. You don’t need “the perfect stack.” You just need:
* a place to think,
* a pair of digital hands,
* a wiring board,
* and somewhere to store what happened.
If you’re reading this as:
* a teacher trying to survive planning, marking, and parent communication,
* a non-technical founder wearing six hats at once,
* or a working professional who’s tired of living inside their inbox,
this four-piece Personal OS is how you stop treating AI like a novelty and start treating it like staff.
---
The Habit Shift: From “Do My Homework” to “Here’s the Spec”
Here’s the second part of your Personal OS: professional habits.
The people who get the most out of AI aren’t the ones with the fanciest prompts. They’re the ones who talk to AI like a project manager, not a panicked student.
Instead of:
> “Write me a lesson plan on photosynthesis.”
> “Make me an email funnel for my agency.”
> “Do the homework for this chapter.”
You start sending:
* A clear goal
* The inputs you actually have
* A short step list (what you’d ask a junior staff member to do)
* The output format
It looks like this:
> “You are my lesson planning assistant.
> Goal: By the end of this lesson, students should be able to explain X in their own words and answer Y and Z.
> Inputs: Ontario curriculum strand A.2, my rough notes below, and yesterday’s exit tickets.
> Steps:
> 1. Create a 3-part lesson outline (minds-on, action, consolidation).
> 2. Suggest 3 formative checks that are AI-resistant.
> 3. Draft a simple parent summary in plain language.
> Output: A Google-Doc-ready outline with headings, bullets, and clear timing.
That’s a spec, short for specification (which is the "secret sauce" you're being sold everywhere).
And any time AI helps you with something important, you do two things:
1. You write down what AI did vs. what you did.
* "AI created draft 1 of the parent email.”
* “I edited tone and added specific student examples.”
2. You log time saved, errors caught, and outcomes.
* “Estimated 20 minutes saved.”
* “AI suggested a date; I corrected it before sending.”
* “Parent replied positively and said the explanation was clear.”
It sounds boring.
It’s actually your career insurance.
That’s the difference between:
> “I play with AI sometimes.”
and
> “I can show you, with receipts, exactly how my digital co-workers save me hours and improve my work.”
That’s also the difference between fluff on a resume and real bullet points that hold up when somebody asks, “Walk me through what you actually did.”
---
The Real Question: What Workload Do *You* Actually Have?
Tools are the easy part.
The hard part is answering one brutally honest question:
> “What is my actual weekly workload?”
Most people skip that step. They search “best AI tools 2025” and bolt random assistants onto a life they haven’t mapped.
I’d rather you do this instead.
Step 1 – Map your goals
Look at the next 90 days and write down your top 3 goals. Examples:
* “Get a remote role that pays at least 75k.”
* “Survive this semester without burning out.”
* “Get my agency to $10k/month without working every weekend.”
Now your AI use has a direction. You’re not just poking at tools; you’re aiming at something.
Step 2 – Map your week
Describe a typical week like you’re explaining it to a stranger:
* How many classes or meetings?
* How much time in inboxes?
* How much time creating materials, proposals, or reports?
* Where do you constantly feel behind?
Actually write it down. “Monday: 3 classes, 1 staff meeting, 2 hours in email, 1 hour marking,” etc.
Step 3 – Find the repeat offenders
Circle the tasks you do 3+ times per week that feel like busywork:
* Copying info from emails into a spreadsheet.
* Renaming and filing student submissions.
* Rewriting the same explanation for parents or clients.
* Manually moving notes from your notebook into your system.
Now we’re getting somewhere.
Because your digital workforce should not start from:
> “Cool things AI can do.”
It should start from:
> “What am I sick of doing that still matters?”
That’s the difference between a gimmick and a genuinely useful co-worker.
---
The Digital Workforce Planner: A System Prompt, Not an App
Here’s where we get into the “no-code, plug-and-play” part.
You do not need a custom SaaS to figure out what digital co-workers you need.
You can start with a single system prompt – I call it a Digital Workforce Planner – inside ChatGPT or Claude.
You tell it, in plain language:
> “Your job is to help me design 1–3 digital co-workers that match my goals, my weekly reality, and my risk level.”
Then, you let it interview you.
It asks things like:
* What are your top 3 goals for the next 90 days?
* What does your week actually look like?
* What tasks do you repeat that feel like busywork?
* What tools do you already use?
* What are you okay automating, and what must stay under human approval?
* What is absolutely off-limits for automation right now?
From those answers, it gives you something priceless:
A short, ranked list of digital co-workers to “hire” first.
Not vague “use AI for everything.”
Not a random 200-tool directory.
Actual candidates, like:
* Inbox Clerk – drafts replies for routine student/parent/client emails.
* Lesson Packager – takes standards + notes and turns them into a lesson skeleton.
* Proposal Drafter – turns bullet points into a first-pass contract or proposal.
Each one should come with:
* Trigger – when this agent should wake up.
* Inputs – what it needs to see.
* Steps – what it will do, in order.
* Output – what “done” looks like.
* Stack – which of your four core tools it will use.
* HIL gate – where you must approve before anything goes external.
You’ve just moved from “AI is overwhelming” to:
> “I have a shortlist of three hires for my digital workforce.”
No code. Just a good system prompt and honest answers.
---
NotebookLM: The Mirror That Shows You Hidden Work
Now let’s add one more layer: NotebookLM.
If the Digital Workforce Planner is your recruiter, NotebookLM is your mirror.
You create a notebook called something like:
> `My Digital Workforce OS`
Then you start feeding it real artifacts from your life:
* Your resume or CV
* Screenshots or exports of your calendar
* A few example email threads (scrub out sensitive data)
* Call or meeting notes
* A doc where you’ve described your goals and bottlenecks
Then you ask NotebookLM very targeted questions, like:
* “Group my recurring tasks into categories: email, docs, research, data entry, scheduling.”
* “Which of these tasks could a digital co-worker do at least 80% of, with me approving the final 20%?”
* “Based on my real week, what are 3 agents that would save me the most time without touching sensitive stuff?”
At that point, NotebookLM isn’t just a note tool.
It becomes a real time diagnostic tool for your workload.
Instead of guessing what agents you need, you have a system that looks at how you already work and says things like:
* “You reply to parents about the same 5 topics.”
* “You spend a lot of time formatting and filing documents.”
* “You manually rebuild the same kind of report every Friday.”
Every one of those is a **job posting** for a future digital co-worker.
---
Designing Your First Digital Co-Worker (Agent Card)
Once the Planner and NotebookLM agree on a top candidate, you don’t build a giant system.
You write a one-page Agent Card.
Something like:
> Agent Name: Inbox Clerk
> Goal: Draft replies for routine emails so I only have to review and personalize.
> Trigger: New email from students/clients with common topics (payments, deadlines, scheduling, basic questions).
> Inputs: Sender, subject, email body, any related info from a Sheet/CRM.
> Steps:
> 1. Classify the email (topic, urgency).
> 2. Pull the right template or past answer.
> 3. Draft a reply in my voice.
> 4. Log draft + classification into a Google Sheet.
> 5. Send me the draft for approval (no auto-send).
> Output: A ready-to-edit draft email + a clean log of what happened.
> Tools: ChatGPT/Claude + Fellou (if browser is needed) + Make/Zapier + Google Sheets.
> HIL Gate: I must approve every draft before it leaves my outbox.
> Minutes Saved per Run (est): 3–5 minutes.
If you can write that, you can build it.
Because in Make/Zapier/n8n, you’re not “coding an AI.” You’re connecting:
* Trigger: “New email comes in with these conditions.”
* Actions: “Send content to AI with this prompt → save to Sheet → send me a notification.”
That’s it.
Your first digital co-worker is born.
---
Governance Without the Overwhelm
The moment we say “agents” and “automation,” a second fear shows up:
> “What if this thing sends something wrong?”
> “What if I break a policy?”
> “What if a parent, client, or manager gets a weird AI email?”
Totally valid.
That’s why governance has to be baked in from day one.
For your first few agents, I recommend three simple rules:
1. Draft-only mode
No agent sends anything to the outside world without your approval. Ever.
If something external is happening (email, calendar, document share), you are the last step.
2. Run-log or it didn’t happen
Every time the agent runs, it logs:
* Timestamp
* What it did
* Where the output lives
* Whether you approved or changed it
This is your black box recorder and your reference sheet in one. This should be done on its own AI agent to reduce token bloat and quality of output.
3. Budget and boundaries
You put hard walls around the agent:
* How many times per day it can run
* Which folders, inboxes, or calendars it can even see
* What categories of actions are off-limits (money, grades, HR decisions, etc.)
You don’t need a 50-page policy PDF to start. You just need guardrails you can explain in 60 seconds.
Over time, if you’re in a school or company setting, those simple guardrails make it way easier to have a serious conversation with leadership:
* You can show logs.
* You can show how much time was saved.
* You can show that AI never made final decisions on its own.
That’s what I mean when I talk about “100 hours of certs in 10 minutes.”
Most people don’t need a wall of badges. They need a way to show:
> “I use AI seriously, safely, and with proof.”
---
Concrete Scenario: How This Looks for a Teacher
Let’s zoom in on one real situation.
You’re a K–12 or private school teacher. Your week is a swirl of lesson planning, grading, and parent communication.
Your Personal OS might center on:
* Using AI to rewrite curriculum docs into student-friendly language.
* Using NotebookLM as a notebook for your units, standards, and assessments.
* Building a **Lesson Packager** digital co-worker that takes standards + rough notes and outputs a lesson skeleton you can refine.
Then you add a second co-worker: an Assessment Sorter that renames, files, and logs student submissions so you’re not stuck in Google Drive purgatory at 10pm.
You haven’t “become a developer.”
You’ve simply designed two small digital co-workers that give you back a few hours and a lot of sanity every week.
---
Why This Matters More Than “Learning AI Tools”
You’ll see a lot of content like:
* “Top 100 AI tools this month.”
* “Prompt pack for every situation.”
* “20 ways to use ChatGPT as a [insert job title].”
Some of it is useful. Most of it is noise.
Because the real leverage isn’t:
> “I know 50 apps.”
It’s:
> “I know how to look at my own workload, decide what to delegate, and design digital co-workers that do that work with guardrails.”
That’s what Citizen Developer means to me.
Not turning you into a part-time engineer.
Turning you into someone who:
* Designs your own digital co-workers.
* Owns your infrastructure instead of renting it blindly.
* Turns professionals into **problem-solvers**, not ticket-writers.
* Wraps everything in documentation and simple governance.
* Starts to build a new professional identity where systems do **real, functioning work** on your behalf.
---
## Where I’m Taking This (and How You Can Join In)
Everything I’m building – from this Skool community to the 30-Day Roadmap and Resource Hub – is aimed at one simple outcome:
> Help you stop working nights and weekends by building your own digital workforce, without writing code.
We do it with:
* System prompts you can copy-paste.
* NotebookLM notebooks you can clone and adapt.
* Agent Cards you can fill out in 10–15 minutes.
* No-code recipes for Fellou, Make, and the tools you already use.
* A culture of “show your run-logs, not your hot takes.”
If you’ve read this far, you probably don’t need another hypey “AI is the future” talk.
You need:
* a Personal OS,
* a short list of digital co-workers to hire,
* and a safe place to build them in public with other people who are done being overwhelmed.
That’s what we’re building through Citizen Developer and Plinko Solutions.
Return time where it matters most.
And start treating AI like what it should’ve been for you from day one:
Not a magic trick.
Not a threat to your job.
But your first, very real, digital co-worker.
0
0 comments
Design your own digital co-workers
powered by
skool.com/citizen-developer-1179
STOP Working Nights & Weekends!
Build your Digital AI Workforce for free. *PAID AFTER FIRST 500*
Access Everything: https://tinyurl.com/citizen-dev
Suggested communities
Powered by