Context is the Lever.

Not a day goes by without someone showing up at a Swiss SME with an AI tool under their arm, promising efficiency gains. In most cases, very little happens afterwards. The tool is tried out, writes usable emails, answers generic questions, and after three months, it ends up in the same folder as all the other pilot projects that never became productive.

The problem rarely lies with AI. It lies in the fact that without context, AI is nothing more than a very eloquent intern who has never worked for the company. It can formulate, but it doesn't know why a quote for a reseller looks different from one for an end customer. It doesn't know the cantonal specifics, the industry customs, and certainly not the unspoken rules that truly underpin the daily business of a fiduciary, a window manufacturer, or a logistics company.

The value is not created in the model.
It is created in the context.

A general language model like GPT, Claude, or similar is a commodity today. The models from major providers have become largely interchangeable in their capabilities for most use cases, and this will continue to converge over the next two years. Anyone who relies on the model is relying on a layer that everyone else also has.

The crucial leverage lies one level above: in the context with which the model operates. Specifically, this means three things.

1. Industry knowledge: how the business truly works, what cycles, what margins, what bottlenecks, what technical terms and linguistic nuances are used
2. Regulatory framework: FADP, VAT liability, industry standards, cantonal specificities, SUVA-relevant processes, Civil Code deadlines, Code of Obligations rules in daily contract practice
3. Customer-specific knowledge: what customer A calls the workpiece, how the bill of materials is structured, what supplier conditions apply, what internal processes have become established over twenty years.

This knowledge is the actual value creation. It is also why a generic SaaS solution from Silicon Valley rarely works in a locally rooted Swiss window manufacturing company, whereas a tailored process does.

A model without context provides usable answers. A model with context makes decisions that an employee with ten years of operational experience would make.

The question of scale has shifted.

Previously, the rule was: anyone who wanted to digitize processes end-to-end needed an IT team. A project manager, several developers, an architect, a test manager, a product owner. Quickly six to twelve people, several hundred thousand Swiss francs, durations of nine to eighteen months. For most Swiss SMEs, this was out of reach, both economically and organizationally.

Precisely this equation has shifted. What used to require a twelve-person IT team can now be implemented with three people, provided the context has been correctly modeled and the engineering is consistently AI-powered.

The Pod.

We call this constellation a Pod. A Pod consists of three people:

• A domain expert, who understands the industry, the regulatory landscape, and the specific realities of the customer, and who knows where the pitfalls lie.
• Two Software Engineers, who develop with AI support and translate the contextual information into robust, productive software.

In the background, the Vantikai Context Engine operates — it structurally stores domain, regulatory, and customer-specific knowledge and makes it available to the Pod and the applications running on it. It's why three people can deliver work that previously required twelve. Not because they type faster, but because they don't have to re-make every decision.

‍

An illustrative example: 
From initial contact to installed window.

A hypothetical scenario. A medium-sized Swiss window manufacturer, regionally established, facing the classic dilemma: high consulting quality, tight margins, long sales cycles, and a diverse funding landscape depending on the canton. Previously, a lot of manual work was involved between customer consultation, measurement, quoting, production release, logistics, installation, maintenance, and after-sales.

What can be built with a Pod and the right contextual foundation:

• Initial Contact: The customer enters their request online. The system identifies the region, cantonal funding programs, and building context, then immediately suggests suitable products to the customer, based on the company's existing product range logic, not on generic web responses.
• Measurement and Quote: The consultant arrives on site, the system has prepared the key parameters, the quote is out within minutes instead of days and can potentially be explained to the customer on the spot, in the correct language, with the right U-values, accurate VAT breakdown, and appropriate subsidy information.
• Production and Logistics: The order flows seamlessly into the bill of materials, appointments are coordinated with the installation team, the installation team is automatically scheduled based on their workload and holiday absences, and the customer can see the status of their order at any time.
• Installation and Handover: Digital acceptance protocol, warranty documents automatically generated, subsidy application prepared with the correct documentation.
• Maintenance and After-Sales: The customer is automatically informed about maintenance options and possible retrofit options (e.g., fly screens in summer).

The effect is not primarily a cost reduction, even though that exists. The effect is that the consultant can be a consultant again. He spends his time with the client, not on Excel spreadsheets and PDF templates. This is what can differentiate an SME from a provider in an industrial park, and that's precisely where the economic leverage lies.

Automation in SMEs is not an end in itself. It is the means to have more time for what makes the difference: the customer.

A new class of opportunities, especially for SMEs.

The honest observation: Swiss SMEs are at a point where, for the first time in decades, they can achieve true process sovereignty, without committing to an ERP provider or industry solution, without building an in-house IT team, without engaging a consulting firm that thinks in man-months.

What was previously reserved for corporations – end-to-end, automated processes highly tailored to their own operations – is suddenly feasible for businesses with twenty, fifty, or two hundred employees. At costs that fit within operational cash flow, not the investment budget for the next three years.

This is not a prediction for the future. This is happening now. And it's not happening through off-the-shelf tools, but through small teams (pods) that combine the contextual knowledge of an industry, regulations, and a specific operation with an engineering practice that uses AI as a lever, not as an end product.