AppenjiStart a conversation
Insights

What we areworking through.

Research notes, architecture perspectives, governance guidance, and case studies — written for the people who have to act on them.

Governance

What an open AI stack is actually worth — to a university, and to a software vendor

A reference architecture is only interesting if it changes what an organization is allowed to do. Two translations, and the security work that decides whether either happens.

Sep 2026 · 14 min read
Architecture

A reference walkthrough: document backlog under data-residency constraints

An illustrative pattern, not a client engagement: how an open-weight stack is assembled when the data cannot leave the jurisdiction and the decisions are appealable.

Sep 2026 · 11 min read
Architecture

Building on open weights: a reference architecture

Open-weight models have closed the capability gap far enough that the deciding question is no longer whether they are good enough. It is who controls the roadmap, and where the data goes.

Sep 2026 · 13 min read
Governance

Governance is a decision accelerator, or it is theatre

Most AI governance is designed to prevent bad decisions. The good kind is designed to make decisions fast, by settling in advance who is allowed to say yes.

Aug 2026 · 10 min read
Architecture

Sequencing decides when the benefit arrives, not whether

Two modernization programmes with identical scope and budget can differ by years in when they deliver anything, and by the entire duplicate estate in what they carry while they run. The variable is the order of the tranches.

Jul 2026 · 11 min read
Security

Retrieval-time authorization is what lets you index the useful corpus

Assistants get pointed at the harmless documents because nobody trusts the permissions. Solve authorization properly and the corpus with the actual answers becomes available.

Jun 2026 · 10 min read
Data

Most decision latency is an argument about whose number is right

The meeting is not slow because the analysis is hard. It is slow because two teams brought different figures and nobody in the room can say which one governs — and both teams are paying to produce theirs.

May 2026 · 9 min read
Research

Seven questions that shorten an AI procurement from months to weeks

Procurement teams are being asked to evaluate AI capability without machine-learning specialists. These seven questions separate a capability from a demonstration, and can be asked by anyone.

Apr 2026 · 10 min read

Tell us what you are trying to change.

Send a note and we will come back with the questions we would ask first.