Across visitor management, legal AI, enterprise integration, internal operations tools and websites, the projects look unrelated. The process underneath them is consistent enough to describe.
Software written to a specification usually solves the problem as stated, which is often not the problem that exists. The gap between what a client asks for and what would actually help them is where most project failure lives, and it is invisible if discovery is treated as a formality.
I spend real time on discovery, design against the workflow that exists rather than the one described, build the spine first, deploy early enough to be corrected, and treat the first months of use as part of the project rather than as support.
A consistent process: understand the real workflow, model it, build the spine, deploy while correction is still cheap, and stay close enough afterwards to fix what discovery missed — because discovery always misses something.
There is no framework here and no diagram worth putting on a slide. It is a description of what has repeatedly worked for me, including the parts that are uncomfortable — deploying earlier than feels safe, spending longer on discovery than the budget nominally allows, and accepting that the first month of use will reveal something I missed.
What it runs on