Patient context for your team.
Leading AI, with local control.
Bring patient-context chat, email assistance, and follow-up support into one local workspace. flow8 connects your practice systems to leading AI models through approved providers, with local data minimization and pseudonymization, staff review, and a full workflow audit trail.
Less searching. More context for the work.
Today’s friction
Before staff can answer a patient’s email, prepare an appointment, or follow up on an open request, they need context. The relevant information sits in the practice management system and the inbox. Repeated record lookups and manual response preparation make everyday work harder to coordinate.
Leading AI models can assist with this preparation, but sending raw patient information to external AI services exposes sensitive data outside the practice’s local controls. flow8 prepares only the permitted context, protects it locally before a model call, checks the response, and records the workflow through staff review and approval.
Who it is for
- Medical and dental teams, practice assistants, and front-desk staff who prepare patient work and answer requests
- Practice managers who own response times, workload and consistent processes
- Data protection, IT and medical leads who must approve how patient data is processed and by which model
Connect available PMS interfaces or agreed exports and your existing mailbox. Configure the authorized context, local protection steps, approved model providers, and review responsibilities around your practice’s workflows.
Leading AI models. Local privacy controls.
Use approved models for language-heavy work while flow8 manages context selection, local protection, validation, and staff review. Choose the provider and processing setup that meet your deployment requirements; a local model remains an option where your policies require it.
The model that fits the task
- Support for summaries and drafts. Use leading models to summarize relevant context, organize requests, and draft replies. Evaluate quality for the practice’s languages and tasks.
- Managed cloud inference. Access model capacity through an approved service without operating that model’s inference hardware in the practice. Local privacy processing remains within your environment.
- Governed model updates. Evaluate and approve a new model before changing the workflow configuration. Keep the local privacy, validation, and review steps in place.
- Your choice of approved provider. Configure the model and provider per workflow. The provider connection follows the agreed processing, retention, and logging requirements.
- Fit quality, speed, and cost to the task. Select models for classification, summaries, or drafting under defined token and cost budgets.
Local protection around every call
- Local de-identification and pseudonymization. A local privacy model and deterministic rules remove or tokenize names, addresses, insurance numbers and contact details before the cloud call.
- Only the context the task needs. Data contracts select the relevant facts — for example an age band instead of a birth date — while keeping medication, units and chronology intact where they matter.
- Identity mappings stay local. Restore identifiers only in your environment, after response validation, and only for authorized staff.
- Stop on detected protection failures. Block model calls when required input checks fail. Stop the affected workflow and route detected masking or remapping failures to staff.
- Approved processing settings. Confirm the provider, actual processing locations and routing, retention, and logging before use. Configure connectivity and access to match those requirements.
Patient-context chat, email, and follow-ups for your team
The AI prepares answers, drafts, and worklists. Authorized staff assess the context, review the result, and decide what happens next.
Patient-context chat for staff
Authorized staff select a patient, inspect the permitted context, and ask questions about the relevant record. Conversations stay separate when switching patients.
- AI prepares
- Contextual answers and summaries with source references and flags for detected missing or conflicting information
- Staff decide
- Check the sources, assess the answer, and retain responsibility for clinical decisions
Email assistance
Summarize incoming patient and referrer emails and add permitted PMS context when the patient can be matched clearly. Route uncertain matches to staff.
- AI prepares
- Request summary, classification, suggested routing, and an editable reply draft with source references
- Staff decide
- Confirm the patient match, edit the reply, and send it through the existing email system
Follow-up support
Surface missing documents, open requests, and follow-up needs in a staff worklist. Patient events, dates, and practice rules identify items to review.
- AI prepares
- A worklist with the triggering rule, relevant context, and a suggested next step for staff review
- Staff decide
- Set priorities, decide whether action is needed, and approve the next step
Five steps, one controlled workflow
flow8 retrieves permitted context, protects it locally, and sends only the prepared input to an approved model. The response returns for local validation and staff review, with a full workflow audit trail.
approved model through the provider and processing configuration selected for the workflow.-
01Local
Connect systems Staff questions, PMS context, email, and follow-up events enter the workflow; patient access is checked. authorized access
-
02Local
Prepare & protect context Select only the relevant facts, then de-identify and pseudonymize them locally. data contract
-
03Approved AI
☁️Approved AI model A model from the deployment's allowlist summarizes, classifies or drafts. limited context
-
04Local
Validate & restore Check structure, sources, and permitted actions; restore identifiers locally for authorized users. schema + sources
-
05Staff
Staff review & act Staff assess answers, edit drafts, and approve the next action. Clinical decisions remain with professionals. staff approval
Workflow orchestration & audit
Access scopes · Local privacy controls · Approved model calls · Staff approvals · Full workflow audit trail
Stays in your environment
- Raw PMS records and email attachments
- Local mapping between pseudonymized tokens and patients
- Patient search index and embeddings
- Conversation history and review decisions
- Audit evidence for every run
Crosses to the model
- Task-specific, minimized context
- Pseudonymized references instead of names and identifiers
- Facts needed for the task; pseudonymized context can still contain personal data
flow8 coordinates the sequence from the staff request, email, or configured follow-up event. It reads permitted data through supported PMS interfaces or agreed exports, resolves patient context, and checks access. The workflow selects the relevant fields and runs local de-identification and pseudonymization before the model call.
The approved model receives only that prepared context. flow8 checks the returned structure, source references, and permitted actions, then restores identifiers locally for authorized staff. Responses and follow-up actions remain subject to staff review and approval. Data access, model calls, validation, and recorded decisions form the workflow audit trail.
What makes cloud AI usable with patient context
These controls are configured per workflow and documented, so the practice can explain what the system does and who is responsible.
Local privacy controls
A local privacy model plus deterministic rules for known identifiers. Accuracy is evaluated per language before cloud calls are enabled for it.
Configurable data contracts
Define which fields a workflow may use, how sensitive fields are handled, what structure outputs must follow and which actions require approval.
Approved models & processing
Maintain the approved models and providers, actual processing locations and routing, retention, logging, and budgets. Staff use the configured connection through the local protection workflow.
Output validation
Check response structure, source references, and allowed actions. Flag or reject detected mismatches, unknown tokens, or invalid references before staff review.
Human oversight
Staff review answers, drafts, and follow-up suggestions before action. Approvals are linked to the final content and recorded in the workflow.
Full workflow audit trail
Trace authorized data access, context selection, model calls, local validation, and staff decisions. Record the model, rules, policy versions, and approvals used for each run.
Practical value. Clear responsibilities.
What the practice gains
- Less context gathering. Bring relevant patient information into the workspace for staff questions and response preparation.
- Less repetitive administration. Prepare summaries, classifications, and editable drafts using the relevant patient context.
- Clear follow-up work. Combine missing documents and open requests in worklists using the practice’s configured rules.
- Traceable decisions. Link sources, review steps, and recorded approvals through the full workflow audit trail.
Practice controls and responsibilities
- Clinical decisions stay with professionals. Staff assess AI-assisted answers and operational follow-up suggestions. Diagnosis and treatment decisions remain with qualified healthcare professionals.
- Pseudonymization, not anonymization. Tokens can be reversed locally, and combinations of clinical facts may still identify a person. Prepared context can remain personal data and retains its protection requirements.
- Processing requirements are agreed per practice. Configure integrations, actual processing locations, retention, and logging against the practice’s data-protection and professional-secrecy obligations.
The flow8 platform, configured for patient context
flow8 combines local context retrieval, privacy controls, workflow orchestration, approvals, and audit evidence. PMS access, provider connections, and review steps are configured around each practice.
Built for healthcare — reusable wherever context is sensitive
The same pattern — prepare locally, use an approved model on limited context, review before action — applies to any regulated team.
Part of the same governed platform
The healthcare solution uses the same governed building blocks as the wider flow8 platform.
Put patient context to work for your team.
Connect your practice systems, protect patient context locally, and support chat, email, and follow-up workflows with approved AI models. Explore the integrations, data controls, and review steps that fit your team.
Discuss your workflows →