Deployment is chosen around data, integration, operation and service requirements, not just user count. The offering distinguishes the managed cloud service (SMB Cloud) from Enterprise options, which are scoped per project: dedicated infrastructure, the customer’s cloud, on-premise or hybrid.
SMB Cloud is a managed service on shared infrastructure. Enterprise can include dedicated infrastructure, installation in the customer’s cloud account or on their own servers. A hybrid scenario combines private components and authorised external services. Availability, scope and responsibilities for Enterprise options are confirmed in the proposal.
Installing the platform on your own infrastructure does not determine where the AI model runs, nor does it by itself prevent calls to external services. Local inference requires suitable models, hardware and tools, alongside a review of data destinations, logs, backups and auxiliary services.
The privacy policy mentions OVHcloud as the infrastructure provider Akui uses to deliver the managed ConsulPoint service, with storage in European Union data centres. It does not state that this location applies to dedicated infrastructure, the customer’s cloud or on-premise deployments: in those cases the provider, location and sub-processors are agreed in the project. Nor does it determine where each AI model runs, which depends on the configured provider.
ConsulPoint supports multiple providers and OpenAI-compatible endpoints depending on the component. Custom models may have different billing treatment, but services such as transcription or speech synthesis can still consume credits. Multi-LLM does not mean optimal automatic routing.
The table summarises, for each option, where data is stored, where models run, which external services are involved, who operates and maintains the environment, what is required and what evidence the customer receives. Enterprise rows describe how each option is approached; final details are set in the project.
| Option | Managed SaaS (SMB Cloud)Standard offering | Dedicated infrastructure (Enterprise Cloud)Enterprise option, scoped per project | Customer cloud (VPC)Enterprise option, scoped per project | On-premiseEnterprise option, scoped per project | Local inferenceProject option, on top of a private deployment |
|---|---|---|---|---|---|
| Storage | Infrastructure Akui uses to deliver the service. The privacy policy names OVHcloud, in the European Union, for databases, files and embeddings. | Dedicated (single-tenant) environment. Provider and location are agreed in the project; neither OVHcloud nor a specific region is assumed. | In the private cloud account (VPC) managed by the customer. Provider and region are those of that account. | The customer’s physical or virtualised infrastructure. The platform uses object storage with an S3-compatible interface; the specific solution is defined in the project. | It does not change where platform data is stored: that depends on the chosen option. |
| Where models run | At the configured model providers (for example OpenAI, Azure OpenAI, Google, Anthropic or Mistral), depending on the component. | External providers or OpenAI-compatible endpoints, as agreed. Inference is not local by default. | Whatever is configured: external providers or OpenAI-compatible endpoints, inside or outside the customer’s cloud. | Whatever is defined. Installing the platform on-premise does not mean models run locally. | Open-source models deployed on your own infrastructure or other European providers, with suitable inference hardware. |
| External services and data outflows | The prompts and context needed are sent to the chosen model provider. Connectors reach the services the organisation authorises. | Depends on the models, connectors and tools chosen. What data leaves, and to where, must be documented. | Installing the platform in the customer’s cloud does not prevent calls to external services. Models, connectors, logs and auxiliary services such as transcription or voice are reviewed. | It is not assumed to prevent external calls: a provider-hosted model, a connector or an auxiliary service sends data out. Restricting this requires defining and testing the permitted destinations. | Auxiliary services, such as transcription or speech synthesis, may still use external providers. Logs, backups and destinations are reviewed. |
| Operation, maintenance and backups | Akui operates the service. Support is by email, without an SLA unless specifically agreed. The scope of backups is confirmed in the offer. | The split of operation, maintenance, updates and backups is defined in the proposal. | The customer manages the account and its infrastructure. The split of installation, updates, maintenance and backups is defined in the project. | The customer operates the infrastructure. The split of installation, updates, maintenance and backups is defined in the project. | Whoever hosts the models operates the hardware and its maintenance; the split is defined in the project. |
| Requirements and conditions | Oriented, as a reference and not a rigid boundary, to organisations of 1 to 200 users. Limits and conditions follow the commercial offer. | Aimed at Enterprise scenarios (reference: 50 to 3,000+ users), depending on the infrastructure, integrations and data governance required. | A cloud account and a Linux environment with Docker Compose, as the platform is containerised. Account access and permissions are agreed in the project. | Linux servers with Docker Compose and S3-compatible object storage. Additional hardware if local inference is required. | Models, inference hardware, compatible tools and data destinations are defined. Custom models may have different billing treatment. |
| Evidence for the customer | Public privacy policy and the applicable data processing agreement. This website does not publish certifications or audit reports. | Location, sub-processors and responsibilities are documented in the project. Compliance and certifications require evidence for the specific deployment. | Location and controls depend on the customer’s account. Deployment documentation is agreed in the project. | Installation and operation documentation is agreed in the project. Compliance and certifications require evidence for the specific deployment. | The project agrees what documentation on models, services and data destinations the customer receives. |
An organisation wants to consult restricted documents from its own cloud. Before defining the solution, permissions, models, hardware, external services and operations are reviewed. The proposal separates licence, implementation, infrastructure, consumption and support, with explicit responsibilities.
With managed SaaS (SMB Cloud), Akui operates the service on shared infrastructure. Enterprise options (dedicated infrastructure, customer cloud, on-premise or hybrid) are defined per project, including the split of operation, maintenance and backups.
Not by itself. Provider-hosted models, connectors and auxiliary services send data outside the infrastructure. Restricting this requires defining and testing the permitted destinations in the project.
The privacy policy mentions OVHcloud as the infrastructure of the managed service, in the European Union. For dedicated infrastructure, customer cloud or on-premise, the location is agreed in the project and is not assumed.
Choosing compatible models, having suitable inference hardware, and reviewing logs, backups and auxiliary services. It is a project option, not a standard feature.
