04 How I work
A close working relationship
From understanding the business to helping people use what we make. Your knowledge and your team’s participation shape the decisions all the way through.
Four steps
How we work
1
Listen
I learn the business, the people and their journey. What are they trying to accomplish? What do they need to understand first? What stops them? What should this help them decide or do next? I also question the request itself, because the first framing of a problem isn’t always the right one.
2
Shape
We agree on what success means. We define the tasks the work has to support, how it fits the systems you already have, and the limits on scope, budget and upkeep. Those become the criteria we judge every option against. This is what discovery produces: the design brief and the plan.
3
Build and refine
I build working prototypes so we can try different approaches, not just imagine them. We compare them with the people who will use the result, see where each works and where it creates friction, and combine what works. Generated work gets checked, and specialists review wherever mistakes would be costly. Exploring ends in a deliberate choice, made against the criteria we agreed on.
4
Hand off or continue
We agree on documentation, training and administration. I train your team to run what we’ve built. If we continue, it’s through an agreed support or development arrangement, shaped to what you need.
Judgment
Why judgment comes first
AI gives intentions and assumptions more reach. A flawed premise can produce a polished result that still fails the people it was meant to help.
So before speeding anything up, I want to know that we’ve framed the problem well, whose needs we’re meeting, and what the result should do.
Imagination opens possibilities. Judgment chooses among them, notices when something needs to change, and checks whether the result does what it was meant to do. That judgment comes from experience, user feedback and the right specialist review.
Five questions
Five questions I ask of every piece of work
The answers combine. A single workflow can mix automation, AI assistance, defined hand-offs to people and specialist review.
What can run on its own?
Routine steps and automatic checks. Tests that run on every change belong here, and they aren’t AI.
Where does AI do the first pass, and who decides?
Drafting, research and code. AI produces a first version, and a person reviews it and decides.
What conditions bring a person in?
Defined triggers while the system is running.
Who has the expertise to approve it?
Specialist review wherever errors are costly. On the RF platform, the company’s engineers reviewed every tool, its logic and the customer workflows.
What must a person own?
Decisions, relationships and accountability. On the on-premises AI product, a person switches features on, and the customer keeps control of their own system.
AI
What AI changes, and what it doesn’t
I’ve been designing and building websites, applications and learning systems for years. AI has made that building more direct. I can explore more alternatives, get to a working version sooner, and spend more of the project with the people who will use it. What to build, and whether it’s right, are still human decisions.