From Shadow AI to Governed Intelligence, How a Swiss Construction Group Implemented AI Governance in Just Two Months

The Starting Point: Shadow AI in the Construction Industry

A mid-sized construction group from Zürich Oberland, over 300 employees, multiple entities, decades of industry expertise, found itself facing a challenge that many Swiss companies are quietly dealing with today: Shadow AI.

Employees had discovered the power of AI tools like ChatGPT. They were using them daily, drafting emails, summarizing documents, comparing quotes. The problem? They were doing it on private accounts, with company data flowing into platforms the organization had zero control over.

As the Head of IT put it in an internal assessment:

"We are not implementing something new. We are governing something that is already happening, uncontrolled."

That single sentence changed the conversation. This was no longer a technology initiative. It was a risk management priority.

Why Governance First, Technology Second

The decision to act came quickly. But the approach was deliberate. Rather than rushing to deploy yet another tool, leadership chose to flip the typical implementation script:

Governance and policy first. Technical setup second. Testing third.

This sequence turned out to be one of the most critical success factors. By establishing clear rules, responsibilities, and boundaries before anyone touched the platform, the organization avoided the chaos that often accompanies digital transformation projects.

Why Hegyon.ai

The Head of IT evaluated several AI platforms before presenting his recommendation. Three factors made the difference:

• Swiss-based platform, data residency and compliance aligned with the company's requirements
• Multi-model flexibility, access to multiple large language models, not locked into a single provider
• Deep integration capability, connections to email, Microsoft 365, SharePoint, and the roadmap to integrate ERP systems like BauBit Pro and A3

The pitch to leadership was efficient: a 30-minute live demo to the CEO and CFO, paired with a short, actionable implementation roadmap. The reaction was immediate buy-in. When decision-makers can see the value and the path forward is clear and realistic, approvals move fast.

The Two-Month Implementation

Month 1: Building the Foundation

The first step was assembling an AI Force Team on the client side. This internal task force was structured with clear responsibilities:

• Internal process owners, mapping workflows and identifying AI-ready tasks
• Knowledge management leads, curating and preparing company documents for the AI knowledge base
• Testing team, validating outputs and flagging issues before broader rollout

In parallel, we worked closely with the AI Force Team to establish governance policies, define access controls, and configure the platform architecture. We provided ready-made templates and frameworks drawn from previous implementations, so the team didn't have to start from zero.

Month one passed quickly. It felt like groundwork. But it was essential. A strong foundation meant everything that followed could move faster and with confidence.

Month 2: From Configuration to First Users

The second month was hands-on. Technical setup, integrations, and platform configuration moved into the foreground. By week six, the inner circle of the AI Force Team was already actively testing the platform in real work scenarios.

This overlap, where testing begins before the official rollout, proved valuable. It meant that by the time the first departments were onboarded, the platform had already been stress-tested by internal champions who could speak to its value from first-hand experience.

Addressing the Elephant in the Room: Employee Fear

Let's be honest. Not everyone was excited.

In a workforce spanning ages 20 to 60, reactions varied. Some employees were curious. Others were skeptical. And some were genuinely afraid. Two concerns surfaced repeatedly:

"Now they are going to learn how we work, so they can replace us later."

"The company can read our messages."

These fears are not irrational. They are human. And ignoring them is the fastest way to sabotage an AI rollout.

Here is what we did: during every rollout workshop, we addressed these concerns head-on. We showed employees the admin interface, the actual backend, proving that no one was reading their conversations. Transparency dismantled the rumours faster than any memo could.

And for the fear of replacement, we consistently reinforced one message:

AI is not here to replace you. AI is your copilot. You are the pilot. You decide what gets delegated.

This framing resonated. It shifted the narrative from threat to empowerment.

The Rollout: Department by Department

We deliberately chose a partitioned rollout rather than a big-bang approach. Department by department, in groups of at least four employees, each team attended an intensive hands-on workshop.

Why small groups? Because in a workshop with four people, you can feel where the question marks are. You can answer concerns in an almost one-on-one setup. You can observe body language and adjust your explanations in real time.

The workshops were not technical deep-dives. People don't care about architecture diagrams. They care about what this thing can actually do for them. So every exercise was hands-on, task-based, and relevant to their daily work.

The Aha Moments

The first reactions during workshops were consistently powerful:

• Email phishing detection, employees were visibly impressed when the platform flagged suspicious emails and explained why they were potentially dangerous
• Research and document analysis, comparing two multi-page supplier quotes in seconds instead of hours
• Calendar and workflow automation, creating appointments, generating graphs directly from SharePoint documents
• SharePoint integration, accessing and analyzing company documents without leaving the AI interface

The adoption curve followed a pattern we've seen before: initial logins were tentative, one message, then logout. But over the following days, usage increased steadily. Employees began discovering their own use cases. The platform stopped being "the new IT tool" and became a daily work companion.

The key insight: give people time. Don't force adoption. Let them discover that the platform genuinely helps with the boring, repetitive tasks they've always disliked. That's when adoption becomes organic.

Three Lessons for IT Decision-Makers in Construction

Based on this implementation, here are three lessons we'd share with any Head of IT considering AI governance:

1. Know the Business of Every Department

You cannot do a meaningful AI rollout if you don't understand what each department actually does every day. Before the workshops, we mapped at least four daily tasks per department, so we could demonstrate, hands-on, how AI could support their specific work. Generic demos don't create buy-in. Relevant demos do.

2. In a Diverse Workforce, More Information Helps

When your employees range from 20 to 60 years old, don't assume everyone has the same baseline understanding of AI. Communicate broadly and frequently: What is AI? What are we implementing? What can it do? What does "automation" actually mean here? Yes, two months is tight for an implementation. But investing time in clear, honest communication generates more positive attention and less resistance.

3. Workshops, Workshops, and More Workshops

This is a digital transformation. You are changing how people work. That doesn't happen with a PDF manual and a login link. It happens through repeated, hands-on teaching. Plan for ongoing workshops, not just during rollout, but in the months that follow.

What Comes Next

The implementation is complete, but the journey continues. The current focus areas include:

• Ongoing workshops and advanced training for power users and new employees
• Enhancing business process documentation within the AI platform
• Expanding the company knowledge base to increase the platform's contextual intelligence
• Integration with BauBit Pro, the modular construction management software by ABBF Bausoft AG, designed specifically for the Swiss construction industry
• Connection to A3 ERP software, bringing financial and operational data into the governed AI environment

Each of these steps deepens the value of the platform and moves the organization further from Shadow AI toward a fully governed, integrated AI ecosystem.

The Bigger Picture

Two months. That's all it took to move from uncontrolled AI usage to a governed, secure, and productive AI platform, integrated with email, SharePoint, and Microsoft 365, with ERP connections on the near-term roadmap.

The Swiss construction industry is not typically seen as a frontrunner in digital transformation. But this project proves that with the right approach, governance first, practical implementation second, and relentless focus on people, even traditionally conservative industries can adopt AI responsibly and effectively.

The question for IT decision-makers is no longer "Should we implement AI governance?"

It's "Can we afford not to?"

Hegyon.ai is a Swiss AI governance platform that helps organizations implement AI securely, compliantly, and effectively. If your company is facing similar challenges, we'd love to hear from you.