Insights
Thoughts on enterprise architecture
Analysis, field experience and points of view on EA, integration and AI-augmented transformation.
Getting off the legacy core banking system: big bang, phased migration or hollowing out the core
Replacing a bank's core system is almost always framed as choosing a software package, yet the decision that commits the bank most is how it gets off the old one. Three strategies — switch everything at once, migrate in waves, empty the core of its functions before replacing it — and what each really demands: a change freeze, two systems of record side by side, accounting reconciliation.
Nobody remembers why: recording architecture decisions so they can be reopened
Diagrams describe what was built, never what could have been built instead. Once the reasons for a choice are lost, an organisation is left with two attitudes: treat what exists as sacred, or challenge it every time someone new arrives. Why decision registers so often fail to prevent this, what a decision must record to be reopenable, and why it is the assumptions, more than the decisions, that need to be tracked.
The project portfolio is the real target architecture: what budget decisions draw without saying so
The IT master plan describes one target; the list of projects funded each year builds another, without anyone deciding to change course. Why the target and the portfolio drift apart, how to read a portfolio as an architecture trajectory, the three signals that reveal the gap, and how to make consistency with the target part of the funding decision rather than an opinion delivered afterwards.
An architecture principle nobody could ever break is useless: writing principles that decide
Most lists of architecture principles are approved unanimously and then never settle a single decision. Not because they are badly applied, but because they are written to be approved. The test that separates a principle from a statement, what a principle must contain to arbitrate, why principles contradict each other, and how to test a list against decisions already taken.
When the business decides without architecture: it is not a question of power, it is a question of mandate
A business unit chooses a tool, signs the contract, then asks architecture to connect it. Asking for a veto changes nothing in that scenario. Why the decision is rational from the business side, why it is made in circuits where architecture is absent, and what an explicit mandate — criteria, a moment, a signed exception — changes in place of a power that will never be granted.
Opening your APIs to partners: an external interface is a commitment, not a connector
An API opened to third parties loses everything that made an internal interface governable: a common arbiter, knowledge of how it is used, the freedom to retire it. What changes when the consumer no longer belongs to the organisation, why the evolution policy must be published before the first partner, and how to protect the information system from those it has invited to connect.
Access rights are a model of the business: why entitlements describe the organisation of ten years ago
The annual review certifies a stock it is in no position to judge, while the mechanism that produced it keeps running. Why rights only ever accumulate in one direction, what the three constantly conflated kinds of right actually are, and where the authority that translates a situation into access should live.
The tender document has already chosen the architecture: what procurement decides before the architect arrives
The tender pack settles the split into lots, the shape of the requirements, the contracting model and the award criteria months before the first design meeting — and none of those trade-offs is examined as an architectural decision. What a tender closes off without saying so, what belongs in it, and how to evaluate a bid on what it makes possible afterwards.
A target architecture is worth nothing without a trajectory: designing the intermediate states
Roadmaps almost never describe the state an organisation will actually find itself in, because transformation always stops somewhere between the starting point and the target. What an intermediate state must satisfy to be liveable, how to design a coexistence regime, and why the exit from the temporary has to be written the day it opens.
What never gets switched off: decommissioning is an architectural decision, not an end-of-project task
Application portfolios almost never shrink — not through negligence, but because switching a system off is the only lifecycle activity with no sponsor, no budget line and no visible benefit to whoever would fund it. The four locks that really prevent retirement, and why the terms of a system's exit belong in its commissioning.
Buy or build: the trade-off everyone treats as financial, and which is structural
The choice between a packaged product and bespoke development is almost always framed as a cost comparison, when it actually decides what the organisation will still be able to change. A grid based on business capabilities, and the anatomy of the most expensive drift of all: configuration that quietly becomes bespoke.
Integration is not plumbing: the interface contract is an architectural asset
Integration disorder almost never comes from the bus you chose, but from interfaces with no owner, no definition and no rule for change. What an interface contract actually contains, how to choose between synchronous and event-driven, and why an organisation's ability to change is written in its interfaces.
Designing for failure: resilience is an architectural property, not a recovery plan
Continuity plans answer the clean disaster; real incidents are partial, slow and contagious. How to translate continuity objectives into design decisions — bulkheads, degraded modes, control of state — and reserve the right level of resilience for the business capabilities that justify it.
Technical debt is not a technical problem: it is a funding decision
As long as debt stays an engineering topic, it never gets a budget. How to qualify it, tie it to the business capabilities it weakens, and make it arbitrable at the investment board — including the portion you deliberately choose never to repay.
Choosing the cloud without getting locked in: trajectory, reversibility and sovereignty
The real risk of the cloud isn't the cloud — it's irreversibility. Why vendor dependence is an architecture property decided upfront, what Europe's new regulatory framework changes, and how to keep an exit open without giving up the benefits.
Total cost of ownership: why an application's real price is decided in the architecture
TCO isn't a budget line, it's a design property. What total cost of ownership really covers, why it has become vital, how to calculate it — and why most of it is settled at the moment of architectural choice, long before the first invoice.
Architecture governance: guide without gating
When the architecture review board becomes a bottleneck, teams route around governance instead of following it. How to shift from up-front control to a framework that guides — guardrails, recorded decisions, automated assurance — without losing coherence.
Data governance: who owns what, and why it's an architecture question before it's a tooling one
Most data governance programmes fail not for lack of a platform, but from ambiguity over ownership, meaning and lifecycle. Why governance is decided in the architecture — and how to set a model that holds.
Security by design: build the threat into the architecture, not after it
Treated as a layer bolted on at the end of a project, security is expensive and protects poorly. How to embed threat modelling, segmentation and sovereignty requirements into your enterprise architecture views from the start.
Business capability mapping: the most underrated architecture artifact
The business capability map is the most powerful lens for prioritising IT investment and aligning technology with strategy. Why an application inventory isn't enough — and how to build yours.
The augmented architect: what AI really changes about the craft
Beyond the buzz, how AI concretely transforms an enterprise architect's day-to-day — and where their value concentrates.
TOGAF isn't a religion: how to adapt it to your context
How to extract value from TOGAF without falling into dogmatism — a pragmatic guide for CIOs who want outcomes, not a full ADM cycle.
Modernizing a banking IS in Africa: 5 realities frameworks ignore
Bank Al-Maghrib, BCEAO, available talent, telecom constraints: why a transformation program in Casablanca or Abidjan isn't run like one in Frankfurt.
Native AI: the real differentiator in enterprise architecture
Why bolting AI onto an existing platform isn't enough, and what changes when it's integrated natively into every module.
TCO: the hidden cost of legacy EA platforms
The license price is just the tip of the iceberg. A breakdown of the eight layers of total cost of ownership of an enterprise architecture platform — and a concrete framework to estimate it.