Written between the jobs.
From the journal
Platform and AI for e-commerce.
Insights on operational design for e-commerce platforms and agents: how the command centre replaces the dashboard, why most AI projects fail, and how tailored development serves your business better than off-the-shelf software.
Tokens make code: what a business can do with AI
For thirty years software was configured to fit, and the business changed to fit the software. Code is now cheap enough to turn that around. A glossary, a process description, dashboards, command centers and the limits.
When the agent leaves the sandbox
Inside a sandbox an agent can only draft. Outside it can act, for better and for worse. The documented incidents, and the least-access rule we work by.
The mood: fear, politics and data centres
Fear for work, rules that keep moving, and the argument about data centres and the grid. The mood around AI in autumn 2026, with every figure linked to its source.
The new equation: build your own, don't rent SaaS
E-commerce has grown up on a decade of SaaS. That model was rational when it took shape. Then it flipped. A story about why custom software is now both faster and cheaper, and how emberloom.ai makes it possible with minimum technical debt.
Stop piloting. Start operating.
78% of enterprises are running AI agent pilots. 14% have reached production. The problem is not the technology.
The command centre: why e-commerce needs an operational layer on top of SaaS
E-commerce companies do not always need more systems. Often they need an operational layer that connects SaaS, data and decisions.