Understand what’s really there.
Map the workflow, the data, the systems and the gaps. If the underlying information is unreliable, adding another AI product usually just gives you a more sophisticated way to use unreliable information.
WORK / APPROACH
Start with the
actual problem.
The technology is rarely the first question. I’m interested in what the business is trying to achieve, what the data can support and what people will actually use.
Fix the foundation — and how it gets created.
Cleaning data once isn’t enough. The more useful problem is understanding why quality deteriorates and making good data easier to capture as part of the work itself.
Make the stack earn its place.
Understand usage, overlap, integration and cost before buying more. Test products against real data and real workflows rather than the perfect dataset in a vendor demo.
Shipping isn’t adoption.
Context, training, evidence, leadership sponsorship and reinforcement matter. I learned this partly by doing it badly: showing people a good tool and assuming they’d obviously use it.
Close the loop.
Did it save time? Surface an opportunity? Improve a decision? Generate revenue? Reduce spend? A tool being available isn’t an outcome.
ABOUT THE WORK HERE
I work with real commercial systems and real company data. Public case studies will focus on the problem, approach and lessons without publishing employer IP, confidential data or pretending every experiment worked.