Industries · Healthcare
AI for healthcare teams: less admin, with patient data kept where it belongs
Clinics and care providers lose hours every day to scheduling, paperwork and the same patient questions, yet patient data can’t be handled like any other data. This page covers the healthcare work we take on, how patient data is handled in each case, and what we need to agree with you before starting.
How patient data is handled in practice
These rules apply to every healthcare project, before anything else is decided.
Records stay on your systems
Patient records stay in your practice-management or records system. An AI feature reads only the fields its task needs, through access scoped to that task, and nothing is copied into a separate database unless that has been agreed in writing.
The model sees as little as possible
Where a task allows it, names and other identifiers are removed before text reaches an AI model. We use AI services whose business terms exclude training on your data, or models hosted inside your own cloud when data must not leave it.
Every access is logged and limited
Staff get access by role. Every AI request is logged: who asked, when, and which record was involved. How long logs and conversations are kept is set with you, not left at a vendor’s default.
Clinical questions go to clinicians
Patient-facing assistants handle administrative questions only. Anything about symptoms, treatment or results is handed to your staff, and urgent messages trigger the emergency instructions you define.
Healthcare work we take on
Administrative and operational work, where AI saves staff time without making clinical decisions.
Appointment booking and reminders
AI chatbots →UsesYour scheduling system, appointment types, the patient’s contact details.
GuardrailNo triage: symptoms and urgent requests go straight to your staff.
Answers to routine patient questions
AI chatbots →UsesPractice information and preparation instructions that you have approved.
GuardrailNever gives medical advice; clinical questions are handed to a person.
Intake forms and referral letters
Workflow automation →UsesThe incoming document, turned into structured fields.
GuardrailStaff check every extracted field before anything is saved to a patient record.
Staff knowledge assistant
AI integration →UsesYour policies, procedures and internal guidelines. No patient data.
GuardrailEvery answer cites the document it came from, so staff can check it in one click.
What the AI sees, and what it doesn’t
Set per feature, written into the project documentation, and enforced by access rules rather than by trust.
| Topic | Can see, when the task needs it | Never sees |
|---|---|---|
| Booking assistant | Open slots, appointment types, the patient’s name and contact details | Clinical notes, diagnoses, test results |
| Patient questions | Your approved practice information | Any patient record |
| Intake processing | The document being processed, with identifiers removed where the task allows | Other patients’ records |
| Staff knowledge assistant | Policies, procedures and guidelines | Patient data of any kind |
Questions we ask before starting
Your answers decide the design. Some of them may mean the right first project involves no patient data at all.
- What data will the system touch, and could the task be done with less of it, or none?
- Which laws and contractual obligations apply to that data where you operate? For example HIPAA in the US, UK GDPR in the UK, PIPEDA and provincial laws in Canada, or the Privacy Act in Australia.
- Where must the data stay: your own cloud, a specific region, or on your premises?
- Which AI providers and hosting options has your organisation already approved?
- Who signs off on privacy and security, and what do they need to see from us?
- What must a patient-facing assistant never say, and where should urgent messages go?
Terms, agreed in writing
- Data-processing and compliance terms are agreed in writing for each project, before any patient data is involved.
- If your organisation needs a specific agreement, audit or certification from its suppliers, we tell you plainly whether we can provide it before you commit to anything.
- A project can start with work that involves no patient data, such as the staff knowledge assistant, while those terms are agreed.
What it costs, roughly
A staff knowledge assistant or patient-questions assistant without access to records usually fits Starter ($500–$1,500) or Growth ($2,000–$8,000). Work that touches patient records is typically Growth or larger, because privacy review, access controls and logging are part of the build.
Compare all packages and termsStart with the admin that costs your staff the most time
We reply within 24 hours on weekdays, and after a free call you get a written plan, timeline and fixed quote.