Business consulting · AI solutions on the processes you already run
Twenty years redesigning processes for professional services firms. I diagnose, I redesign what is broken, and only then do we build the solution that fits your stack and your budget.
The thesis
Most projects that fail did not fail on the tool. They failed on a process nobody had agreed on.
Nobody mapped who decides what, which system holds the source of truth, or what happens when the edge case shows up. Buy a tool on top of that and you have bought a faster version of the same confusion.
So the order is always the same: understand the business, fix the process, and only then choose the technology. Sometimes the honest answer is that a process should not be automated yet. That is also a deliverable.
Selected work
Accounting · Miami
The firm went from assembling every invoice by hand to flagging entities and triggering a batch.
~1,600entities administered by the firm
Specialty food wholesale · South Florida
Five workflows running daily, and the client can audit what the system is about to do by looking at his own spreadsheet.
5n8n workflows live, running every day
Children’s content channel
Ten episodes produced by a single operator, with cost control enforced by architecture rather than discipline.
7idempotent, resumable stages
What changed
Five years ago companies still bent their processes to fit whatever software the market offered, because building something custom was expensive.
Today we build solutions that adapt to the processes the company already runs — from the architecture up, and flexible enough to fit the budget.
How I work with you
01
Where the work actually goes: who decides what, which system is the source of truth, where the edge cases live. Then what to simplify, what to automate, and what to leave alone.
02
The design decision before the build: what the system has to do, where judgment is genuinely required, and what a rule or a schema resolves better and cheaper.
03
When nothing off the shelf fits the way your business actually works, the solution gets built: an internal app, a client-facing tool, or the piece that closes the gap between two systems that were never meant to talk.
04
The solution reads and writes in the systems you already pay for, not in a copy of them. Often the right integration is the cheap one the client will keep paying to run.
05
A proposal, a report, a client brief: the judgment written down once, so the output stops depending on who happened to build it that day.
06
The circuit running on its own, with a model only in the steps that need judgment. This is the line with the longest track record.
07
For teams producing volume with generative tools, where the real risk is budget: a human gate before anything expensive runs, and a dry run on every stage that spends.
After the diagnosis, we build the solution that fits.
Not the most impressive one, and not whatever is in fashion that month: the one your stack supports, your budget sustains and your team can keep running once I am gone. The diagnosis is mine; for the build I bring in solution architects, so the project never depends on one person’s calendar.
What decides it
Start here
A score, your three biggest gaps, and what to do first. It is a deliverable, not a sales call in disguise.
FAQ
Not on the model. On the process underneath it. Nobody mapped who decides what, which system is the source of truth, or what happens when the edge case shows up. So you end up with a chatbot nobody opens twice.
No. I build on no-code and low-code tools that do not require developers, and I handle setup and training. Every engagement closes with technical documentation and an operating guide, so your team runs the system without me. That is the point of the engagement, not a consolation prize.
Automating a broken process just makes broken things happen faster. That is why I start with a process audit: organize and simplify first, then automate. Strategy without execution is useless. Execution without strategy is expensive.
Everything is built against an isolated environment first — sandbox accounts, copies of the live data. Nothing touches production until it has been validated layer by layer: assembly logic, then creation, then delivery, then the full circuit end to end.
Professional services firms, agencies and remote teams of 2 to 50 people, with a document-heavy process that eats skilled hours. Not a fit: model training, infrastructure, or one-off prompt writing.
Let's talk
I will tell you whether it is worth automating before you spend anything. Thirty minutes, no commitment, no pitch deck.