About
I work between operations and software—understanding how a business actually runs, how information is ingested and communicated, and where manual work accumulates. Then I build systems that quietly handle it in the background.
- Building
- Paylit, Brix, BDR Desk, Prospecting Agent, Private Books, and Scout.
- Used by
- A Series A sales team, A101 Construction, Coeus Collective Ventures, a Managing Director at BNY Mellon, an Associate at Fortress Investment Group, and the reps on paylit.io.
- Before this
- All-American athlete, finance degree at a #7 ranked business school, and a few years inside revenue teams at high-growth software companies.
- The workflow comes first
- Automate what shouldn't need a person
- A visible gap beats a silent error
- AI extends judgment
- Reliability is the experience
- The system carries the work
Before building anything, I want to understand how the work actually happens: what people do, what information they use, what decisions they make, and where time gets lost. A process that looks simple from the outside is usually a collection of rules, exceptions, judgment calls, and workarounds. The job is to understand those pieces before deciding which ones software should own.
The workflow comes first. The tool follows.
Software is exceptionally good at work that is repetitive, structured, high-volume, and easy to define: moving information between systems, checking the same conditions thousands of times, keeping records up to date, producing the same output over and over. Those are hours people shouldn't have to spend.
The goal isn't to automate everything. It's to automate what shouldn't require a person.
The dangerous system isn't one that says I don't know. It's one that gives a confident answer when the evidence isn't there. When something is unclear, missing, or contradictory, the system should surface it and ask. A human can resolve a visible question. They can't easily catch an invisible mistake.
A visible gap is better than a silent error.
AI is remarkably good at defined work with clear constraints and a known output. It becomes much less reliable when the task depends on context that isn't written down, ambiguous tradeoffs, or years of experience.
That doesn't make AI less useful. It makes the boundary more interesting. The best systems I've built don't ask AI to replace judgment. They use it to handle everything around the judgment, so the person can spend more time making the decision.
Use AI to extend judgment, not pretend it doesn't exist.
A system isn't finished when the happy path works. Real users enter bad data. Requirements change. APIs fail. Processes get abandoned and restarted. Someone corrects an answer manually and expects that correction to stay corrected. So I care about what happens when things go wrong: work that can resume, data that doesn't silently disappear, failures that are visible, and systems that don't undo human decisions.
Reliability is part of the user experience.
And I don't want to build another piece of software people have to remember to maintain. The best systems fit into the work that already exists. They run when they can, surface something when they need a decision, remember what happened, and keep the user moving. Good software shouldn't create a new workflow just because it has a new interface.
The system should carry the work, not the person.
Email me at losiecki3@gmail.com, or use the contact page.