Automations
The repeat work that quietly eats a week. Mapped, bounded, and handed to something that runs it without being watched.
Work
Four things I do. The detail lives in the conversation.
AI chatbots
Assistants that answer from your own data, inside the tools people already have open. Permissions live outside the model.
Websites
Built by hand. Fast, quiet, and no heavier than it needs to be.
Enablement
Sit down with owners and map how the work actually moves. Often the answer is a better process, and the honest version of that conversation is worth more than a model.
Approach
How the work gets bounded before a model is invited in.
Start with the work
Before a model, ask whether the task should exist, whether a process change is enough, and whether deterministic automation is safer. Do we even need AI?
Permissions live outside the model
The model reasons. The system decides. Identity, tool scope, and secrets never sit in the prompt.
Cost per successful task
A cheap model that fires five warehouse calls can cost more than a stronger one that solves the task once.
Pilot, evaluate, scale or stop
A compelling demo should never silently become operational software. Measure, then decide.
The model reasons. The system decides.
I am not replacing Data Platform or Cybersecurity expertise. I create a reliable interface between that expertise and business demand.
About
I spent more than five years at SumUp, starting close to the customer and moving through operations into analytics. The last stretch has been practical AI enablement: scoping the work, deciding whether a model belongs in it, and turning that into something people can depend on.
I am strongest at the join between business demand, data, and security. I build enough to know the risks. I know when to stop and ask a specialist.
- SQL
- Snowflake
- dbt
- Python
- Tableau
- Langdock
- Cursor
- Git