I Gave My Work to an AI Agent
Delegating research and modeling tasks to agents changed my thinking on control.
The first time I handed over part of my workflow to a GPT-powered agent, I expected chaos.
Not because I didn’t trust the model. No, I must admit that I’m a bit of a control freak at times. I didn’t trust myself to let go of my unique skills and hand them over to a machine. Even one that I myself had built up.
Like many analysts and founders, I had internalized a belief that doing everything myself was a mark of rigor.
I’m fine delegating tasks to people that I know are exceptional at their work, better at their specialty that I’ll ever be. That’s okay for me.
However, delegating to a piece of software that I’d written in barely half a day felt like giving up control in a big and risky way.
But as I watched the agent summarize dense reports, suggest feature ideas, and organize my notes in minutes, I realized: I wasn’t replacing myself. I was creating space for myself to think better.
The Control Problem in Sustainability Work
I have this theory that most of us in technical work develop an obsession with control. In fact, I’ve seen it be proved right time and time again.
If you’ve spent years writing models, debugging data pipelines, or crafting client deliverables—you know how hard it is to let it go.
Especially when the output reflects your reputation. Especially when stakes are high, and your name could get another star, or get tarnished forever.
So when I handed a research task over to an AI agent—one of the first real ones I’d wired up for sustainability work—I expected to be disappointed. Either the output would be sloppy and I’d have to redo everything, or worse: it would look almost right but contain just enough nuance errors to make it untrustworthy.
I was prepared to fix it.
What I wasn’t at all prepared for was for it to free me.
What Agents Actually Do Well
Here’s what surprised me: it didn’t replace my judgment. It just gave me a head start.
It parsed long PDFs and summarized them in crisp bullet points. It drafted outlines for sustainability scenarios that I could reshape instantly. It structured scraped web data into something resembling usable inputs for a modeling pipeline.
The best part was that I could throw any piece of information at it, and it would decide itself what to do with it and how the result fit into the overall workflow. To me, that’s the true beauty of agentic AI.
Was my agent perfect? Nope. Not even close.
But that wasn’t the point. The point was that I wasn’t stuck at zero.
I had moved from “blank page paralysis” to “momentum.”
That alone made the work better. And faster. And lighter. And flow better.
From Analyst to Orchestrator
As the week went on, I noticed something else shifting. I wasn’t spending my energy on tedious tasks. I wasn’t cleaning CSVs at midnight (guilty). I wasn’t copy-pasting from PDFs or searching one mess of a database (that I’d created, guilty again) for missing metadata.
Instead, I had space — to think, to test, to ask better questions. I had more time to cross-check assumptions, rerun forecasts, and dive deeper into edge cases that normally get dropped.
I was doing less execution — and more design.
In some ways, I had stopped being just an analyst and started being an orchestrator.
The agent wasn’t replacing me. It was expanding me.
That’s what true leverage looks like.
What It Still Couldn’t Do
As of now, there are limits.
I don’t trust agents to make investment recommendations or interpret regulatory nuance — not yet. They can’t understand all minutiae of stakeholder dynamics, political context, or the social undercurrents in sustainability data.
The worst part is when they start hallucinating in areas they’re just not good at. As of now, this does require me to manually check where appropriate.
Instead of seeing those gaps as failures, though, I started seeing them as boundaries — helpful indicators of where I was still needed most. Where my expertise mattered.
The agent didn’t steal my work. It showed me where my judgment mattered.
Letting Go Without Losing Integrity
I used to think delegation meant dilution. That handing something off, especially to a machine, meant lowering the quality bar.
Now I think that delegation means direction.
If I can automate 40% of the work that used to clog up my week, I can redirect that time into strategic thinking, deeper modeling, or simply enjoying some more free time with friends or in nature. (Most of us need more of that, I suppose?)
The Bottom Line: When You Let Go, You See More
We don’t scale judgment by doing more. We scale it by knowing when to step back — and when to let the agent take the first pass.
The biggest shift wasn’t technical. It was emotional.
Letting go felt risky. But what I got in return wasn’t just speed — it was space and clarity.
I could see my own thinking more clearly by watching how the agent processed data like I taught it to do. I became a better modeler not by working harder, but by directing better.
It’s just like how we become better technicians by directing junior staff.
Giving my work to an AI agent didn’t make me less rigorous — it made me more strategic.
Wangari’s Curated Reads
Breaking Down the New B Corp Standards—this was a necessary read for me personally, so thanks to
and for penning it. Certification is evolving to meet rising expectations around sustainability, equity, and governance. From new foundational requirements that bar harmful industries to action across key ESG areas like climate, labor, and justice, the changes aim to make certification more meaningful and globally accessible.- has a beautiful piece for us: On kimono, style, and sustainability uses a kimono rental in Japan to question our fast-paced, disposable fashion culture. Through her experience, she reveals how garments like the kimono—unchanged for centuries—offer timeless comfort, dignity, and sustainability. For Wangari Digest readers, it’s a powerful reminder that true style doesn’t chase trends; it lives in garments we love, repeat, and respect—much like the slow fashion ethos we’re trying to reclaim.
This piece from our friends at
spotlights how the founder of Levels, a health tech startup, defied typical efficiency advice by personally appearing on over 100 podcasts to promote his company's mission. The “unproductive” growth hack was this: Instead of outsourcing marketing, he chose to tell the story himself—banking on the authenticity and emotional resonance of long-form audio to build trust and awareness. The gamble paid off: Levels became the go-to name in metabolic health, showing that in a noisy market, founder-led storytelling can be the ultimate growth hack.