My Week with Agentic AI: The productivity boom (and nervous system bust)
- Jan 25
- 7 min read
Updated: Jan 27
TL;DR
I spent a week replicating my company's workflows using an "AI-First" agentic workforce.
The Good: It is exhilarating, fun, and compresses weeks of work into minutes.
The Bad: It pushed my nervous system into "cognitive bankruptcy."
The Takeaway: The tech is ready, but our biology isn't. We need to prepare for a "wired but tired" future.
I feel like I'm still recovering from a week-long experiment we ran in late December. We’ve coined it the Byron Experiment because, well, we did it in Byron. And we did it there, not for any other reason other than we needed to pull ourselves away from the day-to-day to avoid being dragged back into the grind. Plus, it adds a level of commitment by going to a place to do it.
The experiment was to take an AI-first approach to building software and an entire organisation. How far could we take the tooling? How far had it come?
And while I have many answers to those questions, my biggest takeaway wasn’t where the tooling was; it was the unexpected emotional rollercoaster that I went on over that week.
Hype Cycle of Emotions
Gartner’s hype cycle visualises a technology’s lifecycle relative to our expectations for the value it will deliver. It follows 5 phases that, if you look at the chart, take you on what one might call a rollercoaster. And those phases were exactly the emotional rollercoaster I went through in a few short days as I entered the first phase of understanding where the technology stood.
Gartner's 5 phases are:
1. Technology trigger
2. Peak of inflated expectations
3. Trough of disillusionment
4. Slope of enlightenment
5. Plateau of productivity
My Journey Through the Hype Cycle of Emotions
Knowing what it was capable of, I went straight to the peak of inflated expectations. Having heard so much about what is possible from the tech bros, I was not only hopeful but excited.
I quickly fell into the trough of disillusionment. I was feeling flat and tired by it all. How can there be so much out there to what is possible with agentic AI, but not have that feeling when I was in it? I was expecting it to be easy, quick, and magical.
What I found in my initial stages was that the tooling was predominantly geared towards development, with too much complexity or a lack of capability in those designed for businesspeople. And some demand a sales meeting before giving you a sandbox to play in, if they will even do that.
Once I found my toolset, hope started to creep in.
I was still waiting for the “aha” moment, where it all clicked. The type of research and strategy work I tend to do isn’t as simple as software development in terms of structure, process, and repeatability.
Automating an article-writing engine, compared to ideating on product strategy and design, requires different roles, tasks, and workflows. Each workflow needed an entirely different thought process. On top of that variability, I was learning how best to architect my crew, their bounds and context, what tooling they have access to, temperature for creativity, how to stop them hallucinating, as well as what was the best model for these different roles.
As the hours unfolded, I started to move up the slope of enlightenment. There was a light at the end of the tunnel. Things were starting to work, and I entered a state of flow. The possibilities felt endless. I was back to the excitement I had entering the experiment, but with a newfound hunger to see how far I could take it.
I hit the plateau of productivity, finally finding the ‘vibe’ in vibe coding. I felt like I was off the emotional rollercoaster, or so I thought. This led me to my next discovery.
Agentic AI is More Addictive Than Sugar
Agentic AI (or the AI minions, as we call them) are very, very addictive. If Fortnite, pokies, and sugar had a baby, it would be more addictive than that.
It operates on the same psychological trigger as gambling: the 'Variable Reward Schedule.' Sometimes the agent produces magic, sometimes it hallucinates. That unpredictability kept me pulling the lever, hoping for the next jackpot.
I had so many experts in my crew, sitting on my shoulders, guiding me through every idea I set my mind to. I automated integrations into Jira, Confluence, Slack, Miro, and YouTube within 30 minutes. The number of possibilities was widening ever wider. For someone with my type of brain, the ideas kept on flowing, and I was automating multiple flows at once.
We knew that to learn this as quickly as possible, we needed to devote every waking minute this week to it. But what I didn't expect was having to pry myself away to do life things, like eat and sleep. As soon as I woke up, I’d go straight to them, and by nighttime, I had to force myself to go to bed. I left the house once in 7 days, and pretty sure I forgot to shower a few times.
If you give yourself enough time to go deep on this, you realise just how powerful it is, and how much you can do. It was definitely a fun addiction, though, albeit a depleting one.
Agentic AI Will Drive Human Burnout
There will be a bunch of humans this won’t affect because they will happily take the breaks as their agentic AI crews run off and do their thing, but if you are anything like me, you will turn those breaks into productivity opportunities to ramp up the next round of crews.
I would have 3-5 crews running at any one time on a bunch of different tasks. It might be that I am defining a crew from scratch in one, refining a workflow in another, or improving the quality of output of another. Things happen along the way as they get lost or confused, crashes occur, you're refining the process, with a thousand prompts to accept along the way.
AI can operate so fast that what used to take a week of thinking can now be done in 30 minutes. The amount of context I was absorbing and switching between was overwhelming. It felt like trying to run a Ferrari engine inside a Corolla chassis. My AI crews operate at the speed of silicon, but my nervous system operates at the speed of biology. There is a violent mismatch here.
It wasn't just multitasking; it was 'Attention Residue' on steroids. Every time I switched from a coding crew to a strategy crew, part of my brain was still stuck in the previous context. Doing that 50 times a day is a recipe for cognitive bankruptcy.
We are already in a world where multiple platforms on our devices are buying our attention, emails, Slack, Teams, meetings, and more. Now layer in another element of multiple agentic AI threads running at once. It's not just about the number of things you are doing at once; it’s the amount of information the human-in-the-loop now has to absorb to make a decision.
Decision Fatigue & Cognitive Bankruptcy
We only have a finite number of decisions we can make in a day, well, a finite number of good ones. Steve Jobs was a classic example of this, wearing the same black skivvy and pants every day to reduce cognitive load on unimportant tasks.
Every switch requires a micro-decision: Do I answer this now? Where did I leave this off? What is the priority?
Followed by a context refresh of: What was the last thing that the AI did? What was I trying to achieve in my last step? What did the AI "think" in its process?
We are moving from 'makers' to 'maker & manager' at hyperspeed. I wasn't exhausted from doing the work; I was exhausted from judging the work. Being the judge, jury, and executioner for five AI agents is a different kind of tired.
Our brains are going to be drained faster than ever before. Couple this with spiking cortisol and adrenaline... hello, our nervous systems are now in a chronic fight-or-flight mode. I call it the 'wired but tired' state, where your body is exhausted, but your brain is vibrating at a frequency that won't let you rest.
The Collective 'Fight, Flight, or Freeze'
Speaking of fight-or-flight modes, there is another consideration to layer on top of the potential burnout these types of workflows can trigger. After going deep on agentic AI and what LLMs are capable of, the lack of industry adoption leads me to one conclusion: we are in a collective 'fight, flight, or freeze' moment.
Underlying everything is fear and anxiety about what this all means for the future of the human race. The future for the individual, their friends, and their loved ones. There are reports of large companies downsizing, natural attrition silently happening, so what does that mean for our livelihood and, therefore, the very fabric of our society?
Full on, I know, but one of many potential realities could lead us down a path of change with a magnitude we've not experienced in our lifetime.
There are people digging their heads in the sand. They distrust it and claim it’s all just hype machines marketing their companies because it is in their best interest.
There are people who are playing around with AI on the surface, can see what it’s capable of, but telling their bosses that it’s “not where it should be” because they are afraid of automating themselves out of a job.
This is a very valid use case, as what happened at the Commonwealth Bank of Australia (CBA), where 45 people trained a chatbot that ultimately led to their redundancy. CBA has since rehired them, but the damage has already been done. How can anyone trust they'll still have their jobs off the back of an AI initiative now?
Apart from all that, there are people experimenting, learning, and figuring out what it all means with a sense of excitement. These are the people who will drive innovation and change forward, regardless of those frozen and fleeting in fear.
Fight, Flight, or Orchestrate?
The technology is ready for white-collar work, but are our nervous systems ready for the technology? The future of work will require us to orchestrate AI to achieve outcomes, but we must be careful not to orchestrate ourselves into burnout.
The challenge isn't just building the agents; it's building the human resilience to manage them.
Curious about the messiness of innovation? I’m exploring the intersection of psychology, design, and software every day. Here's how we can collide:
Connect on LinkedIn: I share more thoughts on the human side of the AI revolution. Let's Connect.
Book Me to Speak: Want me to talk to your team about AI orchestration, burnout, or the future of work? Reach out on LinkedIn.
What I Build:
Patient Zero: We build and design software that matters. https://www.pz.com.au/
10,000 Spoons: We are experimenting with AI-native business models & development workflows. https://10000spoons.com.au/
