A Company Brain answers questions from approved organisational sources. The choice of platform depends not only on development time but on data sources, identities, publication channel, operations and residency.
Updated on 8 September 2026 against Microsoft and Replit primary documentation.
Copilot Studio
Often fits when knowledge lives in Microsoft 365 and the agent is used internally in Teams, Microsoft 365 or SharePoint.
Strengths: low-code tools, supported Microsoft 365 knowledge sources, Microsoft identity and several publishing channels.
Check: Copilot Credits or licence rights, premium connectors, external users, data movement and feature limits. The billing rates include certain B2E use for authenticated Microsoft 365 Copilot licence holders; that does not make a B2C portal free.
Microsoft Foundry and Azure AI Search
Often fits when development teams need targeted control of retrieval, models, tools, evaluation and interfaces.
Strengths: flexible development components, regional Azure resources, your own security and operating architecture.
Check: model and tool availability by region, deployment type, Azure costs and self-implemented permission filters. According to Microsoft's Azure OpenAI data privacy documentation, regional Standard deployments stay within the selected geography; Global and Data Zone deployments follow different processing rules. “Azure in Switzerland” must therefore not be equated across the board with “all data always stays in Switzerland”.
Replit
Often fits when a custom web application is to be built quickly on an integrated development and publishing platform.
Strengths: agent-assisted development, a freely designed web interface and integrated publishing.
Check: According to Replit's publishing documentation, published apps run on Replit infrastructure based on Google Cloud; they cannot be moved to an arbitrary other cloud provider as a Replit Deployment. Primary hosting is in the US; Enterprise customers can request EU hosting. Enterprise pricing is custom (Replit Pro and Enterprise).
Microsoft 365 permissions, Teams channels and access to business systems must be integrated for each application. Replit hosting should therefore not be ticked off as meeting Swiss data residency; a separate architecture and contractual review are needed.
Internal canvas or external portal?
Answer this question before choosing a platform:
- Internal/B2E: authenticated employees with existing Entra ID and Microsoft 365 rights; Copilot Studio is often the obvious choice.
- Partner/B2B: guest identities and explicitly shared data; check each service's licensing and access.
- Customer/B2C or anonymous: plan your own portal, abuse protection, consent, support and consumption costs. An internal canvas cannot simply be treated as a licence-free public portal.
Decision matrix
| Criterion | Copilot Studio | Microsoft Foundry | Replit |
|---|---|---|---|
| Internal M365 knowledge | strong, with supported configuration | possible, more custom work | possible, integration required |
| Custom orchestration | limited to extensible | high | high at application level |
| External branded portal | check channel/custom canvas | fully custom | fully custom |
| Swiss processing | service- and feature-dependent | region- and deployment-dependent | no documented Swiss hosting |
| Cost model | credits/licences | Azure consumption | plan plus app and model usage |
| Operating responsibility | shared with Microsoft | mostly with the customer | application with the customer, platform with Replit |
Approach
- 1Define the audience and identity.
- 2Record sources, data classes and permissions.
- 3Define the residency requirement for each data flow.
- 4Run a representative pilot with quality and cost measurement.
- 5Only then choose the platform and contract.
Conclusion
Copilot Studio is not automatically the cheapest choice, Foundry is not automatically “Swiss”, and Replit is neither provider-neutral nor operable in Swiss data centres. A sound decision comes from a documented data flow and current offers.
Further reading: Architecture and data sources
