Sources and history
Inspect the source behind a fact, when it applied and when Plooms learned it. Earlier facts remain available when newer knowledge supersedes them.
Explore the history examplePlooms / Security & deploymentSecurity in the infrastructure.
Security in the infrastructure.
Not just the paper policy.
Choose where company knowledge lives, what separates it from other customers and who operates each part of the system. Evaluate those boundaries alongside the policies and handling terms.
Talk to the Plooms teamChoose the boundary your business needs
What changes with each deployment.
Scroll across to compare all three deployment options.
| Your decision | Shared hosting | Fully managed enterprise | Your Azure |
|---|---|---|---|
| Application & stored knowledge | Shared managed host, logically isolated per company. | Dedicated stack in an isolated Azure subscription operated by Spej. | Application, documents and knowledge graph in your Azure tenant. |
| Operating responsibility | Managed for the team. | Spej operates the application infrastructure. | Plooms operates through a revocable, resource-group-scoped Azure Lighthouse grant. |
| Model processing | Separately managed open-weight model serving. | Separately managed serving. Serverless or dedicated capacity is selected for the workload. | Relevant context is sent to separately managed model serving. |
| Commercial structure | Monthly seats and prepaid usage. No setup fee or dedicated-infrastructure minimum. | Setup, seats and usage, plus infrastructure at cost with a fixed markup. | Setup, seats and usage. Azure infrastructure is billed directly to you. |
| Next step | Small-business and nonprofit signups are not open yet. | Discuss and request a managed enterprise workspace. | Discuss and request a customer-cloud workspace. |
Application isolation and model inference are different boundaries. The serving arrangement, provider and handling terms are part of the deployment discussion. Explore enterprise options or check shared-hosting availability.
Stored knowledge and model processing are separate.
APPLICATION & STORED KNOWLEDGE
Your Plooms workspace
In your cloud or Plooms hosting
Your documentsApproved connections
Relevant contextAnswer returned
SEPARATE MODEL PROCESSINGManaged model serving
Managed for you by Plooms
Dedicated deployment example. The application and knowledge stack can run in your Azure tenant or Plooms hosting. Relevant context is sent to separately managed inference to produce an answer.
Chosen sources, explicit access
Per-user connectors let people choose what to ingest. Personal items are private unless shared. In an organizational workspace, administrators can access workspace data under the organization’s policies.
Provider handling commitments
The current privacy policy names separate hosting and model-serving providers and states no retention or training by the serving provider. Review those terms alongside the data flows and enabled tools.
Model choice and evaluation
Open weights support technical evaluation and deployment choices. Plooms routes jobs with budgets, fallbacks and cost attribution. Correctness and security still require validation and controls.
Three kinds of evidence01 / Knowledge 02 / Workflows 03 / Security
Know what you can inspect.
Approvals and execution
Playbooks support human approval gates and audited execution. Approved agents use scoped tools, with trace IDs connecting runs and graph writes.
Read about PlaybooksSecurity event records
The privacy policy describes up to 90 days of detailed security event records. Confirm event coverage, retention and export requirements for your deployment.
Read the privacy policyThese records answer different questions. Source history alone does not establish a complete action or security audit.
See the sources and history behind the answer.
Revisit a policy before and after a change. The example shows why keeping earlier knowledge matters when a reviewer asks what applied at the time.
Source visibility helps a person check the answer. It does not guarantee correctness or explain every part of a model’s reasoning.
Company context, with a history.
“Who approves a new supplier?”
CURRENT ANSWER
Operations reviews the supplier. Finance also approves payment terms.
Supplier policy · 12 August
Earlier policy retained as historyCurrent policy
For business and technical teams
Before you connect company data.
Privacy and data handlingModel operationsWhere is our data stored and processed?
Application hosting and stored knowledge are configured in your cloud or Plooms hosting. Model processing runs on separately managed model serving, using serverless or dedicated capacity for the workload. The proposed architecture should identify the provider, region and flow of relevant context between them.
What do open-weight models let us inspect?
Open-weight models make the weights available for technical evaluation and deployment. A closed model API does not provide that weight access, though its product can still offer access controls and audit logs. With Plooms, discuss the model, version and update process for your workload. Weight access does not reveal every cause behind an answer or remove the need for validation.
Does an isolated application mean dedicated GPUs?
No. Application and data isolation are separate from model serving. Managed serverless inference can reduce idle-capacity costs; dedicated serving can suit requirements for capacity, latency or stronger compute isolation. We select the serving arrangement for the workload and explain where relevant context is processed.
What can our team inspect?
The workspace supports company context with source provenance and history. Review the specific access controls, event coverage, retention and any export requirements with us for your configuration.
How do connected sources work?
Plooms lets users bring connected knowledge into their workspace. Scope, sharing and administrative policies should be reviewed as part of deployment, including how access changes and deletion are handled.
Is Plooms automatically compliant for our industry?
Compliance depends on your obligations, the configuration and how the system is used. We work through data handling, identity, retention, logging and contractual requirements with your stakeholders before you commit.
What changes when external tools are used?
Web search and connected services can introduce additional data flows. Review which tools are enabled, what information they receive and the applicable handling terms.
How should we evaluate cost?
Start with your workflows, expected demand and capacity requirements. Review setup, seat subscriptions, infrastructure and model usage together, including the agreed operational responsibilities.
Make it real for your team
Discuss the boundary your business needs.
Bring your data requirements and use case to an introductory conversation. We’ll arrange the right next step with your team.
Talk to the Plooms teamStart with a conversation. Explore the right setup.