Level 5: Workflow Augmentation - When AI Stops Being a Tool and Becomes a System
It All Begins Here
Published April 28, 2026 · The Human Question · By Rob Gonzales, CPA, PhD
You already use workflows. You just have not named them.
Wherever you are right now, on the train, at your desk, in line for coffee, in the elevator, reading this when you probably should be doing something else, you did not just appear there.
A lot of things had to happen first. You woke up. You brushed your teeth. You put on socks. You picked an outfit. You grabbed your bag. You walked out the door. You unlocked your phone. You scrolled past three things. You opened this.
That is a workflow.
Same thing with a professor walking into a classroom. From the outside it looks simple. They show up, they teach, they leave. But behind that moment is a long chain of preparation, reading, organizing, planning, second-guessing, and decision-making. Most of which you never see.
You already have workflows. You just have not designed them.
The thing you already do - but with intention
Think about your morning again. Some of what happens needs you. Deciding what to wear that matches the weather and the meeting. Reading the room before you say something hard. The actual thinking part of your job.
But a lot of what happens does not need you, in the strict sense. It just has to happen. The vacuuming, if you have a robot. The laundry, if you send it out. The dog walk, if you have a sitter that day. The grocery shopping, if you do delivery.
That is not laziness. That is intention. You are deciding, for each thing in your day, what actually needs me, and what just needs to be done.
That is exactly what we are doing with AI now.
From doing the work to designing the work
Levels 1 through 4 were about how to interact with AI. How to prompt. How to give context. How to break work into steps. How to deploy an agent.
Level 5 is different. It is the level where you stop thinking about each conversation, and you start thinking about your system.
The question shifts from “what can AI do for this task” to “how does AI become part of how I work, every week, without me redesigning it from scratch every time.”
That shift has a name. Workflow augmentation.
Workflow augmentation = letting AI handle the repeatable parts, so you can focus on the meaningful parts. You are not removing yourself. You are removing the friction around yourself.
What this actually looks like
Let me make it concrete. Take something most knowledge workers do every morning.
The old way
1. Open the browser.
2. Search “AI news” or whatever your topic is.
3. Click through five or six articles.
4. Read most of them.
5. Try to figure out what actually matters.
6. Maybe summarize for yourself.
Twenty minutes, on a good day. Sometimes more. And nine times out of ten you end up reading the same handful of sources you always end up at.
The augmented way
1. AI gathers the news for you.
2. AI filters for what matches the topics you care about.
3. AI summarizes in your voice, in the structure you always use.
4. You read and you think.
Same outcome. Less friction. More focus.
The most important concept in this whole lesson
If you take nothing else from these five levels, take this:
You should always have a context file for anything you care about.
A context file is not fancy. It is a plain document where you have written down, in one place, how you think, how you write, how you structure your work, what your standards are, what your rules are, and what good looks like.
Without a context file, every time you open a new AI session, you are starting over. You are explaining your project for the fifth time. You are pasting in the same examples. You are re-teaching the AI things it should already know about you.
With a context file, you drop one document at the start of a session, and the AI shows up like a colleague who has been on the team for six months.
And here is the bonus that nobody talks about: a good context file is portable. Move from Claude to ChatGPT, from ChatGPT to Gemini, from Gemini to whatever ships next. The context file goes with you. The thirty minutes of re-training does not.
Where this came from for me
I recently helped a tech firm and a learning company that were running into the same problem. They had a team of writers each working in slightly different voices, with the standards living inside the heads of the senior reviewers. Getting a new person up to speed took weeks of feedback rounds.
The fix was not better training videos. It was a context file.
One document, kept in the project folder, that covered everything: the mission, the folder layout, the voice, the editorial rules, the review pattern, the quality checklist, and the session-startup protocol that walked the AI through an intake before any real work began. Roughly two dozen sections. Versioned by date so a new edit never overwrites the prior version.
Distributed to every team member. Dropped into every new AI session. The same starting point every single time. Output stabilized. Onboarding time collapsed. The senior reviewers stopped being the bottleneck.
I do the same thing for my own projects. Every active engagement has a context file. Every new chat starts with that file. The amount of time I used to lose to re-explaining myself is gone, and the tradeoff — keeping the file maintained — is small. The human still drives. The file just makes sure the AI is up to speed.
The flagship example: an AI news agent for your class
Here is a real workflow you can build this week. It illustrates everything in this lesson.
The goal
Deliver a short, punchy AI news update to your students before every class, in your voice, with your framing, without spending an hour on it.
The pieces
1. Perplexity for the gathering. It is built for fresh information retrieval and citation. Set the date range to capture only what is new since your last class.
2. Claude for the thinking. Drop in your context file (your voice, your audience, your usual structure). Hand it the Perplexity output. Ask for a 200-word script and three discussion prompts.
3. HeyGen for the delivery. Paste the script. Pick the avatar — yourself, a stylized version of yourself, or honestly, a cartoon dog if that is the brand you want. Render. Download.
4. Your LMS or email. Post it the night before class. Watch it yourself once, so you know what your students saw before they walk in.
That is the whole workflow. Twenty minutes once you have the context file in place. Five minutes a week after that.
Notice what this does. It does not remove you. You are still the architect. You still review. You still teach. What it removes is the half-hour of clicking that used to come before the thinking, every Tuesday and Thursday.
One important rule: do not skip this
Not everything should be automated. And not everything should be exposed to AI.
Be careful with student information, personal data, anything covered by FERPA or HIPAA, confidential client files, or anything you would not want in a vendor's training dataset.
The simplest fix is also the most reliable one. Have separate folders. One that the AI can read freely. One that the AI never touches. Drag and drop accordingly. Do not use a single mega-folder for everything and rely on memory.
Build the wall before you need the wall. It is much easier than apologizing for a leak.
A small exercise
Pick one thing you do every week. Something repeatable. Something that takes you longer than it should and produces something fairly predictable.
Now ask two questions.
1. What part of this needs me?
2. What part is just repetition?
Whatever falls into the second bucket is a candidate for augmentation. Build a small context file. Define the steps. Pick the tool. Run it once. Refine. Run it again.
That is your first real workflow. The next ten will be easier.
Three rules to take with you
1. Build the context file first. Before you build any workflow, write down how you think and what good looks like. Without that document, the workflow is fragile.
2. Augment, do not abdicate. AI handles the repeatable. You handle the meaningful. The judgment never leaves your desk.
3. Wall off what is sensitive. Separate folders. Strict habits. Never rely on remembering not to upload the wrong thing.
The real lesson
Most people use AI like a shortcut. The ones who actually get value out of it use it to build systems.
And here is the line I keep coming back to.
You do not rise to the level of your tools. You fall to the level of your systems.
Levels 1 through 4 taught you how to think with AI. Level 5 is where you start to build with it. After this, the question is not what AI can do. The question is what you choose to design.
Coming next: Applications — AI in Teaching, AI in Research, AI in Industry, AI in Service