Home/Insights
Thinking on enterprise systems.
Not a blog. These are the questions that come up repeatedly in scoping conversations, answered at the length they deserve, with the reasoning shown rather than the conclusion asserted.
Articles
How to evaluate an AI use case in a business workflow
The question is not whether a model can perform the task. It usually can. The question is whether the workflow around it can survive the model being confidently wrong, and that is answerable before you build anything.
Workflow automation is not process ownership
Automating a process and owning a process are different things, and organizations that confuse them end up with software that executes a workflow nobody is accountable for. The distinction becomes expensive at exactly the wrong moment.
Connecting an AI assistant to an enterprise platform's API
An enterprise platform can expose several hundred API operations. Handing that surface to an AI assistant is less a question of integration than of where the boundary is drawn, and the read-only bridge we built to draw it is open source.
Why auditability matters in automation
An automated process that cannot explain what it did is a process the business has stopped owning. Auditability is not a compliance overhead bolted on at the end; it is what makes an automation trustworthy enough to leave running.
How to plan a legacy modernization without a risky rewrite
Most modernization programmes are cancelled rather than completed. The reason is almost always structural: the benefit was deferred to a cutover that never arrived, while the risk was carried from the first week.
How to design a handover-friendly system
Handover is not a phase at the end of a project. It is a constraint on every decision made along the way, and a system designed without it in mind cannot be retrofitted into one that somebody else can operate.
When low-code is the right answer in an enterprise system
Low-code is neither the future of software nor a toy. It is a tool with a specific envelope, and most of the disappointment surrounding it comes from projects that were outside that envelope from the first day.
Why internal tools fail after launch
Internal tools rarely fail at go-live. They fail in the eighteen months afterwards, and they fail in a small number of predictable ways that have almost nothing to do with the quality of the original build.
Build versus buy for internal platforms
The decision is usually made on licence cost and delivery time, which are the two least informative inputs available. A more useful question is which of the two options you can still change your mind about in three years.
Facing one of these questions on a real system?
A short advisory engagement often settles it faster and more cheaply than a project.