I'm the founder of EMGT Labs. I learn complicated domains from the inside, make sense of how the work actually happens, and turn that understanding into reliable products and robust, AI-enabled operational workflows.
I've spent 10+ years across product, marketing, e-learning, sales and operations. That hands-on experience helps me design systems that are stress-tested in real-life business situations.
Four examples of turning complicated work into products, business systems and processes that people can actually use.
Mapping how a multi-channel media business actually works, from pre-production all the way to distribution, then designing clearer responsibilities, project management systems, and a clear separation between old-school software, AI-enabled reasoning, and human taste and judgment.
Designing a macOS tool around the reality of media production, taking as much load off the humans involved in the process as possible: take footage from multiple cameras and microphones, back it up safely in multiple locations, then generate proxies and multicam sync files, and deliver a ready-made stringout to the editor so they lose zero time on busywork and can focus entirely on their creative process.
Building a language-learning media business, producing hundreds of YouTube videos that reached millions of language learners, all with a small founder team of two and a remote team of editors and teachers.
I learned copywriting and digital marketing while we were trying to sell our own language courses, selling well over $1m in the process. I went on to work with other educators on offers, campaigns and multiple 6-figure product launches.
I start with the outcome, then learn how people actually work. That often means finding unnecessary steps, undocumented knowledge and the situations that make a process frustrating. From there, I design a cleaner model of the data, decisions and workflows, and try to break it with real-life scenarios.
Only then do I decide where straightforward software, AI reasoning and human judgment each belong. The goal is less mental load and outcomes people can rely on.
What does success look like, and what do people actually need?
Where do people get stuck, repeat tasks or keep essential knowledge in their heads?
Design the data, stages and handoffs. Then test the awkward, real-life cases.
Decide where humans, reliable automation and AI each make the work easier.
Understand the outcome and the real work
I start by learning the domain from the people who know it best. What are they trying to achieve? What does a successful result look like? How do they currently get there?
I follow the work from beginning to end, looking at where people get frustrated, where information gets lost, and which decisions or tasks depend on someone remembering something that's never been written down.
Coming in with an outsider's perspective helps me question things that may have become so normal that nobody thinks to question them anymore.
The goal is to understand what people actually need, so we don't end up automating unnecessary work or making a bad process run faster.
Design a cleaner process, then try to break it
Once I understand the outcome and the reality of the work, I start again from a clean slate.
I combine what I've learned with research into how others solve similar problems. Then I map out a better way to run the process: the information we need to track, the stages work moves through, who's responsible for what, how decisions get made and what counts as finished.
Before building anything, I try to break the model.
What happens when someone misses a step? When two people change the same information? When a project gets handed over halfway through? When a drive disconnects during a backup?
I'd much rather discover those problems in a model than after ten people have started depending on the software.
The result should be a process people can understand and trust, including when things don't go according to plan.
Let people, software and AI do what each does best
Only once the process makes sense do I decide how to implement it.
Some tasks can be handled by simple software or deterministic automations. Other tasks involve interpreting messy information, where AI reasoning can help. And some decisions still need a human, particularly where taste, relationships, money or consequential judgment are involved.
I work out where each belongs, put the appropriate checks in place, and then build and test the system in real use.
What I'm ultimately looking for is pretty simple: the person doing the work should know what's finished, what needs their attention and what happens next, without having to keep the whole process in their head.
Less chasing updates, fewer things falling through the cracks, and more time to focus on the work that actually needs them.
Questions I keep returning to.
A small space for ideas beyond my work, and for an occasional longer essay when I have one to share.