Guiding Progress

Data and AI decisions cut across business, IT and security. Lasting solutions take shape when these perspectives work together.

We bridge the gaps and stay with open questions until they are resolved.

Shaping Data and AI Solutions

How a platform is structured, how components interact and how data flows sets the direction for years.

• Architecture and platform design. A shared reference architecture keeps teams aligned. It connects the platform, its components and the systems already in place.

• Build or buy. Building in-house or adopting a vendor product sounds like a technical decision, but hinges on maturity, data sovereignty and long-term dependencies.

• Data architecture for AI. Data architecture sets the ceiling for AI. How data is captured, stored and served determines which initiatives are realistic.

Organisational Structures and Operating Models

Clear responsibilities, decision paths and interfaces between business, IT and security carry an initiative from pilot into production.

• Operating models for AI. After go-live, an AI product needs clear ownership of operations, evolution and quality. The interplay between business and IT that this requires rarely emerges on its own.

• Platform organisation. Platform teams with product ownership make the teams that build on them faster and more autonomous.

Integrated Management Systems and Governance

A management system connects strategic intent with operational reality. ISO 27001 and ISO 42001 deliver value when they are anchored in day-to-day operations, not just in audit cycles.

• ISO 42001. Building an AI management system or integrating AI-specific requirements into an existing ISO 27001 control environment. Regulatory drivers like the EU AI Act and the Swiss DPA become part of the system instead of producing parallel documentation.

When does external advisory on data and AI make sense for Swiss organisations?

External advisory on data and AI makes sense for Swiss organisations when a decision spans business, IT and security, and no internal role brings all three perspectives together. We take on that bridging role and work toward the organisation owning it independently over time.

What is the difference between advisory and coaching for data and AI?

Advisory for data and AI works actively on the solution and contributes recommendations. Coaching strengthens the ability to make decisions independently. In practice, both formats often blend: we advise where needed and enable where possible.

What matters in build-or-buy decisions for AI?

Build-or-buy decisions for AI differ from traditional software procurement because the organisation’s own data is part of the solution, whether built in-house or sourced externally. Choosing a vendor product requires clarity on how training data, model outputs and feedback loops remain under the organisation’s control. Building in-house, on the other hand, often leads to underestimating the ongoing effort for model maintenance, monitoring and retraining. The right answer rarely sits at one extreme but at the boundary between internal control and external capability.

Why is a traditional BI architecture often insufficient for AI?

AI places different demands on data than traditional reporting: higher granularity, fresher data and end-to-end traceability. Many organisations have data that works for dashboards but falls short for training or evaluating models. Common gaps include missing provenance documentation (data lineage), quality issues that surface only during model training, and access models that do not reflect data protection requirements at sufficient granularity. Identifying these gaps early saves months downstream.

Why do AI products need their own operating model?

AI products bring requirements absent from traditional software: models degrade as underlying data shifts (model drift), prediction quality requires continuous monitoring, and retraining demands access to current, quality-assured data. Organisationally, a gap frequently opens between the team that develops a model and the IT function expected to run it. A working operating model closes this gap with clear roles, defined handovers and a shared understanding of product quality.

How can ISO 42001 and ISO 27001 be combined in one management system?

Integrating ISO 42001 into an existing ISO 27001 ISMS succeeds when AI-specific controls are systematically mapped onto the existing control structure. In practice, this means extending risk assessments with AI-specific scenarios, broadening existing change management and internal audit processes to cover AI systems, and introducing new controls only where the current control environment does not already address them. Organisations that take this path avoid duplicate documentation and keep the maintenance burden for the overall system manageable.