What is at stake in software engineering for a large organisation
In a large company or a public administration, software engineering is not just about writing code. It concerns an estate: applications of different ages, built by successive teams, linked by exchanges that no one any longer knows in their entirety. Every business request builds on this existing estate, and it is this that sets the real cost and lead time of change.
Three tensions recur in many IT departments.
- The cost of change is rising. A change that looks simple requires modifying several applications, coordinating several teams and rerunning manual tests. The budget goes on maintaining what exists, and the capacity to launch anything new shrinks.
- Knowledge is becoming concentrated. Business rules live in the code and in the memory of a few people. A departure, the end of a third-party maintenance contract or the end of support for a technology become operational risks.
- Requirements are rising. Security, service continuity, protection of personal data under the GDPR (Regulation (EU) 2016/679), traceability: these constraints apply to the software itself, and not only to the infrastructure that hosts it.
The subject is therefore not technical in the narrow sense. It involves the organisation’s ability to deliver its projects, to absorb a new obligation or to open a digital service without jeopardising what already works.
Our reading: what makes enterprise software succeed or fail
Failures attributed to technology often have a cause that predates the first line of code. Four mechanisms recur in degraded situations.
Architecture decisions taken by default. How components are broken down, the integration mode or the technical foundation commit the organisation for a long time. When they are neither discussed nor written down, they are settled as things go along by the team of the moment, and no one can say any longer why the system has this shape.
Project logic where product logic is needed. An application delivered ‘finished’ then handed to a team that did not build it degrades quickly. Software that supports a living business process needs a stable team, an identified owner and a budget for evolution, not just a budget for construction.
Quality checked at the end. When testing comes after development, under schedule pressure, it is cut short. Defects are discovered in acceptance testing or in production, at the point where fixing them costs the most.
Invisible debt. Every shortcut taken to meet a date creates technical debt whose interest is paid in slowness and incidents. It is not a problem as long as it is known, measured and repaid by choice. It becomes one when it appears in no trade-off.
Conversely, organisations that deliver regularly share a few traits: application scopes with clear boundaries, an automated delivery pipeline, tests that run on every change, and teams that operate what they build or work in direct contact with those who operate it.
How we work
Our approach follows four phases: Frame, Prove, Industrialise, Operate. Framing means establishing the problem, the constraints and the options before committing a build budget. Proving means removing the major uncertainties on a reduced scope, under real conditions. Industrialising means extending what has proved itself with the expected quality and security guarantees. Operating means keeping the software alive with your teams, to whom our consultants pass on what they need to maintain it.
This approach applies to four families of work, which correspond to the most common situations.
Building what does not exist off the shelf. When the need draws on know-how specific to your organisation, our consultants strengthen your teams for custom application development: product framing, the build-or-buy decision, incremental production releases.
Evolving an ageing estate. When an application becomes too costly or too risky to maintain, application modernisation means choosing, application by application, the right takeover strategy and carrying it out without service disruption.
Providing overall coherence. When projects collide or exchanges between applications become unmanageable, the work focuses on IT architecture: target, trajectory, integration rules and an arbitration body.
Making deliveries reliable. When every production release is a gamble, our consultants help structure software quality and testing: test strategy, automation, controls built into the delivery pipeline.
These kinds of work combine. A modernisation often begins with a safety net of tests, custom development fits into a target architecture, and an architecture trajectory only has value if teams put it into practice. In every case, we work with your teams and not in their place: the code, documentation and decisions are produced in your tools and stay with you, readable by those who will take over.
Software engineering in contact with the cloud and AI
Software does not live alone. Its design depends on the platform that hosts and deploys it: continuous integration and deployment pipeline, on-demand environments, observability. These subjects fall under our cloud and platforms expertise, with which we align application choices. An application redesigned without taking its runtime platform into account shifts the problem onto operations.
Artificial intelligence is also changing the profession, in two ways. On the one hand, applications incorporate AI components, language models or agents, which call for specific design and testing safeguards. On the other, coding assistants are changing the way code and tests are produced. We address both aspects with our artificial intelligence expertise, without losing sight of a simple rule: generated code is still code to be reviewed, tested and maintained.
Where to start?
The right entry point depends on the question you are asking yourself today.
- You need to decide the fate of one or more applications. Start with a diagnostic: state of the code, test coverage, dependencies, business criticality, available skills. It should produce compared options, not a single recommendation.
- You are launching a new digital product. Start with a short framing exercise that settles three points: what is built first, what is bought, and how the first version will be put into the hands of real users.
- Your production releases are too rare or too risky. Start with the delivery pipeline and the tests of the most critical application. Progress there is visible and serves as a model for the other teams.
- Your projects contradict one another. Start with a map of applications and flows for the scope concerned, then with an architecture body capable of deciding.
In each of these cases, the first step is limited in scope and produces a result your teams can use, whether what follows is done with us or without us. To discuss it, describe your situation to us via the contact page.