.jpeg){kind=avatar}
The Opportunity
Sonae Sierra operates across dozens of markets with a globally distributed workforce, where consistent access to information and decision-support tools is an infrastructure challenge in itself. Employees were experimenting with consumer AI tools on their own, while leadership wanted a single, governed way to bring AI into daily work at scale.
Fragmented, consumer-first AI adoption
Employees who wanted AI support had to rely on tools like ChatGPT or Copilot, with no integration into internal systems and no way to share knowledge or workflows across teams. Others went without entirely, creating an uneven baseline of productivity and access to information.
No governed workspace for AI
Leadership wanted AI deployed across the organisation, but not as a patchwork of tools with their own identities, permissions, and data policies. The requirement was clear: one governed workspace, aligned with existing Entra ID groups and single sign-on, where access, permissions, and audit trails followed the same rules as the rest of the infrastructure.
Costs tied to headcount, not usage
Per‑user licensing made it expensive to roll out AI beyond a small group of early adopters. Scaling from hundreds to thousands of employees under a seat-based model would have multiplied costs without guaranteeing real usage or measurable impact.
The Solution
We deployed SierraGIA, Sonae Sierra’s general assistant, as the foundation for the organisation’s AI capability.
Employees access a single conversational workspace where they can ask questions, summarise documents, draft content, and run specialist agents configured for their functions. Answers are grounded in the organisation’s own knowledge - operational procedures, legal documentation, HR policies, commercial guidelines - and every response cites its exact source, so users can verify and trace the underlying documentation.
Access is managed via the organisation’s existing Entra ID groups and single sign-on. Governance and permissions are inherited directly from the identity infrastructure already in place, ensuring that only the right people see the right content, and that usage is auditable by design.
On top of this foundation, specialist agents have been introduced and continue to expand:
- An operations agent that supports procedures governing shopping centre operations.
- A legal agent that answers questions grounded in contracts and policy documentation.
- An MS Office 365 agent that helps employees manage calendars and prioritise their inbox.
Underneath every agent is the GenOS Platform. Usage is visible across teams and functions: which agents are being used, which knowledge sources drive the most queries, and where coverage is missing. As new agents and knowledge sources are added, they inherit the same governance model, controls, and observability.
Crucially, the platform follows a usage-based pricing model. Instead of paying per user, the organisation shares a common message pool across thousands of employees. The system is model‑agnostic, avoiding vendor lock‑in and allowing best‑model‑per‑use‑case selection as needs evolve.
The Impact
Sonae Sierra went from no governed AI capability to a platform used by around 550 employees, generating roughly 150k messages in three months, without a forced rollout. Adoption spans HR, Finance, Legal, Operations, and commercial teams, who now share a common workspace instead of experimenting in isolation.
By shifting from per‑user pricing to a shared, usage‑based model, the company unlocked AI access for far more employees at a sustainable cost, while maintaining full control over data, security, and observability.
SierraGIA now serves as the organisation’s entry point into a broader AI ecosystem. New agents can be added on the same governed foundation, reducing friction for future use cases and turning AI from a set of isolated pilots into an operational capability embedded in daily work.
