AI coding agents produce working software at remarkable speed, but they know nothing about your organisation: which systems exist, how they are named and which rules apply. Enterprise Architecture provides exactly this context in written form.
Agents without architectural context produce software that works on its own but does not fit the landscape around it. Correcting that afterwards is expensive.
01
The meta model defines the vocabulary
Agents must speak the language of your organisation. The meta model defines capabilities, applications, processes and data objects, and the relationships between them. Together with agreed definitions, these concepts form the shared ontology that every agent uses.
02
The repository grounds the meaning
The repository connects this ontology to your enterprise: what each capability and data object means, which systems support it, who owns it and how everything relates. Agents use governed organisational knowledge instead of guessing.
03
Principles are the guardrails
Architecture principles, standards and process models translate directly into constraints and acceptance criteria that agents can be measured against and reviewers can check.
04
Governance is the control plane
Review paths, decision records and evidence requirements remain just as relevant when a machine writes the code. They are what allows you to take responsibility for the result.
How we work
Engineering discipline, at agent speed.
We develop software with AI agents in production use, including our own business platform. This is the approach we arrived at, and the one we establish with our customers.
01Specification-driven delivery
Specification before implementation.
Prompting and accepting the result is acceptable for prototypes, but not for software that people rely on. For substantial features we work specification-first: use cases, functional and non-functional requirements, edge cases and acceptance criteria are documented before implementation begins. The agents build against this specification and derive the tests from the requirements rather than from the generated code, so the tests describe the intended behaviour.
What it gets youSustainable speed, because the specification remains the reference when the feature is extended later.
Use cases & requirements
Acceptance criteria
Tests from requirements
Spec-driven tooling
Definition of readyGate
Before any agent starts.
Use cases written and agreed
Non-functional requirements explicit
Edge cases enumerated
Tests derived from requirements, not code
“The prompt sounds about right”
Required
02Architecture as context
Grounded in your actual landscape.
An agent that does not know your landscape will invent a plausible one, for example a service that was decommissioned last year or an integration pattern you have moved away from. We provide the architecture repository as working context, so agents build against the applications, interfaces and data objects that actually exist. The meta model serves both purposes at once: as the model your architects maintain and as the shared vocabulary your agents follow.
What it gets youResults that fit the existing landscape, and less review effort spent on correcting invented assumptions.
Repository as context
Shared vocabulary
Landscape-aware output
Fewer hallucinations
03Guardrails
Defined boundaries for agents.
Architecture decisions have to be binding, and with agents they have to be enforceable through the delivery pipeline. We translate stack decisions, integration patterns, security requirements and process models into explicit constraints: which patterns are mandatory, which components may not be touched and which changes require a human decision first. Without these boundaries, every session introduces its own patterns and the codebase gradually loses its structure.
What it gets youA codebase that retains one consistent architecture, even after hundreds of agent contributions.
Binding stack decisions
Mandatory patterns
Process-derived criteria
Human decision points
Working agreementAgents
What agents may change.
Approved patterns and stack only
Architecture decisions are binding
Schema and API changes require human approval
New dependencies require human approval
Access control and audit code are not changed
Enforced
04Review & evidence
Automated review, documented evidence.
Generated code must be reviewed. At the volume agents produce, that review must be automated. We set up automated code reviews that run on every commit and flag typical risks in generated code: certain logic errors as well as weaknesses in access control and other cross-cutting concerns. At the same time, we maintain the required evidence, because compliance requirements apply regardless of who wrote the code. As a TISAX-certified company, we must demonstrate a controlled development process ourselves and apply the same standard in customer projects.
What it gets youDefects are caught before the merge, and the review and test history is available for audits.
Automated review per commit
Access-control checks
Traceable change history
Audit-ready evidence
Review ledgerThis week
Every change accounted for.
Change 481Tests ✓ · review ✓ · merged
Change 482Returned: access‑control gap
Change 482 rev. 2Tests ✓ · review ✓ · merged
Change 483Tests ✓ · review ✓ · merged
The complete decision and test history remains available.
Audit‑ready
05Where it fits
The right scope for agentic development.
Agentic development is particularly well suited when the scope is manageable and you can define the requirements clearly: internal applications, platform extensions, custom reports, integrations and migration tooling. It becomes more demanding when an application turns into a shared digital backbone and every feature has to account for permissions, notifications, audit logging and existing modules. We help you assess your initiative realistically, including when licensing a product is the better solution.
What it gets youA well-founded build-or-buy decision before the budget is committed.
Internal applications
Extensions & reports
Integrations & tooling
Build-or-buy advice
AssessmentBuild or buy
Internal asset management tool.
Scope bounded
Requirements ours to define
Cross-cutting obligations
Standard product available
RecommendationBuild, agent‑based
Proven on our own platform
Our own platform, written by agents.
ITM InMind is our internal business platform. Every line of it was written by AI agents, while our developers set the direction, review the results and test each module. It supports our daily operations, so we know from experience where this approach works and where its limits are.
Most teams produce a working prototype with agents within days and then encounter very similar obstacles. In our experience, the cause is rarely the model, and it is rarely solved by more elaborate prompting.
