We deliver faster.
With experienced developers,
supported by AI.
We use AI tools to speed up routine tasks, while our experienced developers handle system design, business decisions and go-live. Every project starts with an approved specification, and the AI works within that framework. This can shorten delivery times, make development effort more efficient, and you stay in control throughout.
AI support with spec-driven control
TitleAI-assisted enterprise software development is a way of working where the development team uses AI tools to speed up the repetitive, easily verifiable parts of the work, while professional decisions stay with experienced developers.
Webtown does this using the spec-driven development method. Spec-driven development is a development approach in which human validation happens at the level of the specification and the development plan, before any code is written. The AI works from the approved specification, and only code that meets the approved requirements moves forward.
Where AI speeds things up:
- Preparing boilerplate code and repetitive development tasks
- Preparing automated tests
- Drafting the first version of technical documentation
- Analysing existing code and supporting its clean-up
- Code search and faster review of larger codebases
- Preliminary summaries of development tasks and changes
Where we make the decisions:
- System architecture and technology direction
- Business logic and client-specific behaviour
- Security and data protection solutions
- Performance-critical and critical system components
- Code review, testing strategy and go-live
- Restricting or ruling out AI use for sensitive tasks
A developer checks every AI-assisted code change. Nothing goes into production without review.
Why is the control point at the specification?
Because that’s where most issues come to light. The specification defines what a feature does, what input it accepts, who can access it and what happens with invalid input. A permission gap, a missing validation, flawed data handling logic or a wrong integration assumption is already visible at this level. You don’t have to hunt for it in the finished code.
A senior developer can review a few pages of specification more thoroughly than several hundred lines of code changes. Traditional line-by-line review would also take away the speed that makes AI worth using in the first place.
How do we check AI-written code?
Testing also runs with AI support, but a human approves the benchmark. Every approved requirement has a test case that fails if the code deviates from what we validated in advance.
This builds trust the other way round compared to traditional code review. We don’t ask whether the code is good, we ask whether it matches what we approved. That’s faster and more reliable to verify, so speed doesn’t come at the expense of security.
WHAT YOU GAIN?
-

Faster delivery
Routine tasks take less time, so developers can spend more of their time on real design work, problem-solving and work that creates business value.
-

More efficient development effort
Working more efficiently can mean shorter delivery times, more development capacity or lower project effort. In every project, we set out the pricing model and the potential savings transparently.
-

Verifiable quality
The code is measured against the approved specification and the tests built from it. That way, quality doesn’t depend on gut feeling but on traceable requirements.
-

Transparent, and fits into your existing teams
At the start of the project, we agree together on which tasks use AI, with which tools and under what data handling rules. Our developers work in line with your tools, processes and security requirements.
-
Assessment
We get to know your system, your business goals, your technology environment and your security requirements.
-
Team & framework
We put together the right development team and agree on which tasks use AI, with which tools, under what restrictions and with what data handling rules.
-
Specification and approval
Every development task gets a specification and a development plan with acceptance criteria. No code is written until these are approved.
-
Development
The AI works from the approved specification. Changes are version-controlled and documented, and a developer reviews the critical components.
-
Testing and go-live
The tests measure against the approved requirements. Code only goes live once it meets every agreed condition.
Security and data protection
TitleBusiness-grade AI tools
We use business AI tools and configurations that fit the requirements of the project. When choosing tools, we look at data retention, access, the data processing region and whether client data can be used for model training.
Protecting client code and client data
We never put sensitive client data into AI tools. If needed, AI use can be ruled out for certain tasks, or development can be carried out within your own infrastructure.
Security at platform level
Liferay and Ibexa come with a mature, well-documented security model, from permissions and authentication to data access. When we build within this model, the project inherits what would otherwise have to be rebuilt and maintained every time.
Rules agreed in advance
At the start of the engagement, we agree on the scope of AI use, access, retention and restricted data categories. If the project requires it, we also cover these in a separate AI use and data processing annex.
Data protection requirements
We align the development and data handling process with the security and data protection requirements defined for the project. Specific compliance conditions are always determined by the system, the data involved, the infrastructure and the contract.
Why Webtown?
We have been building enterprise systems for more than two decades and have worked with our clients on more than 1,000 projects. We develop for banking, manufacturing, public sector and energy environments, where stability, security and long-term maintainability are baseline requirements.
We have been a Liferay partner since 2011 and are now a Multinational Platinum Partner. We have worked with Ibexa since 2016, today as a Gold Partner. As an IBM watsonx Silver Business Partner, AI governance is not a theoretical question for us.
We don’t treat AI as a standalone gimmick. With the spec-driven development method, we build it into our existing development, security and governance processes, so speed never comes at the expense of control.
We also have our own UX/UI team, so user experience is part of the work from the start of the project.
-
No, and that’s intentional. We put the control point in the specification and the development plan, because that’s where most issues come to light. The code has to pass tests built from the approved requirements, and a developer reviews the critical components.
-
It’s yours, just like any other code we develop for you.
-
No. We never put sensitive client data into AI tools, and the tools we use do not train on client data. We agree on the exact rules at the start of the engagement.
-
Yes. At the start of the engagement we agree on which tasks can use AI support and which ones we handle with traditional development methods only.
-
A development method where human validation happens at the level of the specification and the development plan, before any code is written. The AI works from the approved specification, and the tests measure against it. Webtown uses this method in every AI-assisted enterprise development project.
-
With us, the AI doesn’t guess. It works from an approved specification, and only code that meets the requirements agreed in advance moves forward. That’s the difference between controlled AI-assisted development and vibe coding.