From Operator to Orchestrator: Automating the work that keeps leaders from customers
The line I won’t cross
It seems like half the posts on LinkedIn today (at least in my feed) are debating how, where, and when to use Claude. Many of us are navigating this space and trying to find the right balance between the productivity boost and not sounding like a robot. I’ve developed my own boundaries that I thought I would share.
My goal is to automate the administration of my business, but never the conversations — which means you’ll never be hearing from me on LinkedIn with a DM that was written by Claude. And my reasoning for doing this in the first place isn’t just because automation is fun – though it is! It’s because I’m trying to free up my time so I can interact more with customers and prospects, which is the part of my work I love the most.
Why this way of thinking comes naturally to me
Some of my ability to do this comes from having spent a career in the enterprise setting, where — especially at SAP — every critical part of the business was a technical workflow with human checkpoints. It’s natural for me to think that way.
So when I look at my own business now, I ask the same kind of question I always asked about those workflows. How much can I automate? Where do humans need to check in? What can automation do to accelerate how I work, so I can focus on the places where I add unique value as a human — idea generation, creativity, understanding human behavior?
Asking that question repeatedly has given me clarity about what parts of my business I want to enable with AI. As I undertake more and more automation, I’ve seen that the work seems to move through four stages:
The four stages: capture, encode, delegate, orchestrate
I think of the sequence as a maturity model — a sure sign I spent too much time in management consulting!
Capture comes first. At the outset, work lives in my head or scattered in snippets of documents, emails, and post-its. None of it can be automated until it’s digitized in one place; capture is unglamorous, but everything else depends on it.
Encode is next, and it’s the stage many people skip past. Writing something down doesn’t make it usable. Encoding means structuring it so a model can read it — because a pile of notes can be immaculately organized for me and still be completely inaccessible to a model. I’ll come back to that gap in my next post.
Delegate is where the work starts to gain momentum. You give a task to a model and check it at points you choose. It does the work, and you stay responsible for whether it’s right.
Orchestrate is the far end, where many of those delegated pieces run together as one system, while keeping human decision-making at the center. The routine parts run on their own, and that enables me to turn my attention to the places where human judgement is truly needed.
I read this as an anthropologist
There’s a reason I observe this differently than a lot of people rushing into it.
I’m sure by now you’re convinced that I drank the Kool-Aid, and that’s completely fine. I am having a fabulous time right now, but I am also continuously reflecting and adapting as well. Because even though I’ve worked in the software industry for most of my career, I am also an anthropologist who seeks to understand and ease how humans use technology.
People ask how you’d explain your job to your grandmother. I like to say that I explain humans to engineers, and technology to humans. That’s the simplest expression of my core work identity — and it’s why, as I automate, I’m always asking what to hand to the machine and what to keep human.
Why this is worth doing now
In the first quarter of 2026, about 17.8% of the world’s working-age population had used generative AI at all. In the United States, where most of my clients are, it was 31.3% — so two out of three American workers still haven’t touched these tools. That’s the opportunity — anyone who starts now is early.
I’m figuring this out in public
I’m here to share how I’m flailing along, and how my work has gradually accelerated with the help of AI. As I get further along, my intention is to start hosting workshops where I can share the breadth of what I’ve automated and help others do the same.
Why I’m excited to bring people along
I spent most of my career in tech — not just on the business side, but hands-on — and I’m finding this an exciting evolution and time in my own work. Although I love running my own business, there are a lot of administrative aspects of the work that I don’t really enjoy, and I’ve been focused on using Claude to automate those for me, and to take care of things that were suffering from benign neglect as well.
What I’m hoping to do in this series is to make all of that real and within reach for others. I’ll be sharing the steps I’ve taken along the automation journey for my own business, along with some of the lessons I’ve learned, and exploring a little bit about what a progression from capture to encoding looks like. My hope is that people will enjoy learning how I’m taking advantage of these new tools — and maybe want to learn along with me.
Leave a Reply