The 27-Language Construction Platform Built to Keep Data in Its Place

A marketing site available in 27 languages would normally arrive with an equally long list of third-party connections. Analytics, advertising tags and consent tooling tend to accumulate quietly, until the page designed to explain a product is also reporting each visit elsewhere.

Gewerkton takes the opposite route. Its marketing site supports 27 content languages while using zero trackers, showing no cookie banner and running on a fully egress-free architecture. That combination makes the site more than a multilingual shop window. It is a compact privacy-engineering case study for the platform behind it: reduce unnecessary data movement, make regional choices explicit and treat custody as part of the product rather than an administrative detail added later.

This matters because Gewerkton is a voice-first construction documentation and defect management platform. Its job is to turn activity on a construction site into evidence: dictated records, defects, daywork reports, instructions, plans, models and operational coordination. The company’s marketing line puts the premise plainly: “On site, what counts is what’s proven.”

A platform built around proof cannot treat the location and handling of project data as secondary questions. The architecture has to address not only what gets recorded, but also where it resides, which providers process it and who retains operational control.

A multilingual site without the usual tracking layer

The most visible expression of that thinking is the marketing site. It serves content in 27 languages, contains a media bank of more than 51 self-produced clips and posters, and does so with zero trackers. There is no cookie banner because there is no tracking layer to negotiate, and the site is fully egress-free.

That is a useful inversion of the usual consent experience. Many sites begin with a web stack that creates external data flows and then place a banner in front of the visitor to manage them. Here, the architectural decision comes first: the site does not need a tracking request because it does not deploy trackers.

The distinction is important. A cookie banner is an interface. Privacy is a property of the underlying system. Removing the interface while retaining the same outbound behaviour would solve nothing. Gewerkton’s case is interesting because the absence of the banner accompanies zero trackers and an egress-free design.

Supporting 27 languages also makes the choice more consequential. This is not a small, single-market page with little content to manage. It is a global-facing site carrying a substantial bank of original media. Its privacy stance therefore sits alongside, rather than in place of, international reach.

There is an editorial lesson here for security and privacy teams. Data minimisation does not have to mean reducing the usefulness of a public service. A site can offer broad language coverage and rich media without making visitor tracking part of the bargain. Capability and collection are separate design choices.

From public website to project evidence

The marketing architecture establishes a principle, but the harder questions sit inside the product. Construction records can move between people, trades, sites, plans and operational systems. They may begin as speech in the field and end as a report, defect record, instruction or coordinated project task.

Gewerkton divides this work across three product lines under one brand. Gewerkton Field is the voice-first construction site app. It turns dictation into evidence and supports defects, daywork reports, takt and a portal. For housing and building construction, that can mean a defect recorded with a photo and deadline, a dictated daywork report, or a signature captured on the device available at handover.

Gewerkton Studio is the browser workspace for plans and models. Where no model exists, the site team can create one in the browser. That puts model creation closer to the people dealing with the physical work, rather than assuming every project begins with a complete model ready for use.

Gewerkton Cloud carries the wider architecture story. It handles operations and model and data coordination between Field, Studio and third parties. This is where questions of residency, regional processing and custody become operational rather than theoretical.

Gewerkton is in beta now, with a public beta planned for fall 2026. That status should be read plainly: the platform is not being presented as a finished, generally available system. The privacy and data choices are nevertheless already central to how the product is described.

Data residency should be a product choice

Gewerkton offers two stated routes for data residency: an EU cloud or the customer’s own infrastructure. The choice is not framed as a hidden deployment detail. It is part of the platform’s data stance.

For a system coordinating evidence between field activity, browser workspaces and third parties, that choice affects the basic relationship between the platform and its users. Some organisations may want an EU cloud. Others may want the system in their own house. Gewerkton does not force both groups into one location model.

“Your own infrastructure” also changes the custody discussion. It means an organisation can choose to keep the platform within infrastructure it controls, instead of accepting the EU cloud as the only route. The point is not that one option is universally correct. The point is that deployment and custody requirements differ, and the architecture acknowledges that difference.

This becomes particularly relevant across the sectors Gewerkton targets. Wind farms and renewable projects involve distributed sites, rotating crews, field acceptance and offline capture in dead zones. Data centres and industrial plants can have many trades working in parallel under tight deadlines, with meeting decisions becoming trade-sorted task lists. Infrastructure and tunnel projects run for long periods, accumulate many change orders and may need instructions backed by the original audio.

Those scenarios do not all share one organisational structure or one preferred location for data. Treating residency as a selectable property is therefore more coherent than presenting a single hosting arrangement as suitable for every project.

Regional AI choice without a single-provider dependency

Residency is one part of the picture. AI processing is another. Gewerkton’s bring-your-own-AI model supports 13 AI providers, with users bringing their own keys and selecting providers by region. The available regions include the EU, the US and Asia, including mainland China.

That is a notable architecture choice for a voice-first product. Speech and language processing are not peripheral embellishments here; they sit close to the path from spoken site activity to structured evidence. Provider selection therefore has implications beyond model preference.

Gewerkton’s design does not bind that path to one AI vendor. Organisations can choose among regional providers and use their own keys. The stated benefit is no vendor lock-in, but the privacy significance is just as clear: AI-provider choice becomes part of the deployment decision rather than an invisible dependency fixed for every user.

The breadth of language support gives that regional approach a practical context. Gewerkton is intended for global markets and supports cross-border teams in the EU, US and APAC working on the same project, each in their own language while the evidence original stays unambiguous. On projects in Asia, Chinese, Korean and Vietnamese crews can work multilingually from capture through to report, with data residency chosen by the organisation.

Regional selection does not by itself answer every possible governance question. What it does provide is a concrete control: users are not limited to a single provider or a single geographic provider category. Combined with bring-your-own keys, it makes the AI dependency visible and selectable.

Why custody belongs next to evidence

Evidence platforms carry an unusual responsibility. Their value comes from preserving the relationship between an event and its record: what was said, what was observed, what task followed and which supporting material belongs with it. If the platform treats custody as an afterthought, it weakens the user’s ability to decide how that record is handled.

Gewerkton’s deployment examples make the connection concrete. In a tunnel project, an instruction can be backed by its original audio. In housing, a defect can carry a photo and deadline. During handover, a signature can be captured on a device. In a data-centre meeting, decisions can become task lists sorted by trade. Across borders, people can work in their own languages while keeping the evidence original unambiguous.

These are not generic office notes. They are records tied to changing physical work, multiple participants and continuing operational decisions. Their usefulness depends on access and coordination, but that does not mean custody should default to one provider, one region or one deployment model.

The combination of EU cloud or customer infrastructure, regional AI-provider selection and bring-your-own keys gives organisations several distinct choices:

• They can select where the platform resides: in an EU cloud or on their own infrastructure.
• They can choose AI providers by region across the EU, US and Asia, including mainland China.
• They can bring their own AI-provider keys.
• They can avoid dependency on a single AI vendor.

None of those controls needs fear-based marketing. They are ordinary architectural responses to a platform operating across regions and handling project evidence. The case for them is strongest when stated without drama: different organisations and projects have different requirements, so the system should expose meaningful choices.

Global reach without erasing local depth

Gewerkton was born in the German market and has its deepest commercial integration there, including GAEB, REB, XRechnung and DATEV. At the same time, it is being built for global markets, with 27 content languages and AI-provider options spanning the EU, US and Asia.

That pairing avoids a common problem in international software. Global ambition can flatten local requirements, while deep local integration can make expansion difficult. Gewerkton’s stated structure keeps the German commercial foundation while adding multilingual capture, reporting and regional provider choice for wider use.

The deployment fields reflect that range. Distributed renewable sites have different operational conditions from housing handovers. Industrial plants with many parallel trades differ from long-running tunnels with numerous change orders. Cross-border projects add language and regional data questions to all of them. The platform’s answer is not a separate product family for every setting, but Field, Studio and Cloud working as one branded house.

An agent-built platform with an evidence-minded test story

There is another unusual element in the build. Gewerkton is being developed by a solo founder directing a fleet of coding agents using Codex and Claude. In one night, that fleet shipped 21 software packages, verified with negative controls and mutation tests.

The number is striking, but the verification detail matters more to this architecture story. An evidence product cannot rely solely on the appearance of rapid output. Negative controls and mutation tests were used to challenge whether the software behaved as expected, bringing an adversarial element to verification.

That does not turn development speed into a substitute for product maturity. Again, Gewerkton remains in beta, and the public beta is planned for fall 2026. It does show how the project’s emphasis on proof extends into its development account: not just packages produced, but packages subjected to specific forms of testing.

Cloud as the coordination and custody layer

Field provides the voice-first capture point. Studio provides the browser workspace for plans and models, including browser-based model creation where one does not already exist. Cloud coordinates operations, models and data between those products and third parties.

That makes Cloud the natural centre of the privacy-engineering story. Coordination is where boundaries meet: site and office, speech and report, plan and field observation, internal tools and third parties. It is also where a platform must make clear whether its customers have choices about infrastructure and AI processing.

Gewerkton’s answer is a set of explicit architectural positions. The marketing site collects no tracking data and has no outbound egress. The product can reside in an EU cloud or on infrastructure chosen and operated in the customer’s own house. AI providers can be selected by region, users can bring their own keys, and the platform supports 13 providers instead of requiring one.

The result is not a claim that architecture eliminates every privacy or security decision. It is a design that puts several of the most consequential decisions in view. For a platform whose purpose is construction documentation and defect management, that visibility is part of the product’s credibility.

Privacy as structure, not decoration

The most convincing privacy choices are often quiet. A tracker that was never added does not need a preference panel. An outbound connection that does not exist does not need to be explained away. A regional provider choice built into the platform does not have to be negotiated after adoption. A deployment option available from the outset does not require every organisation to surrender the same degree of custody.

Gewerkton’s 27-language marketing site is a small but unusually clear demonstration of this approach. It reaches across languages without attaching trackers to that reach. The same philosophy becomes more substantial in Cloud, where data residency and provider selection meet the operational reality of Field, Studio and third-party coordination.

For construction teams, the practical promise remains straightforward: capture what happened and connect it to usable evidence. For privacy and security readers, the more interesting point is how the platform frames the surrounding architecture. If what counts on site is what is proven, then where that proof resides, which provider processes it and who controls the infrastructure cannot be footnotes. They are first-class product decisions.