Patient intake that lands in the chart, not on a clipboard

Patients complete intake on their phone before they arrive. The workflow reads the forms and insurance card photos, maps each answer to the right EHR field, and stages it for a staff member to verify, so nobody retypes a paper packet at check-in.

01the problem and who owns it

Intake is still a clipboard and a scanner in many practices. Front desk staff retype names, addresses, insurance IDs, and medication lists while the waiting room fills, and typos in a member ID become claim denials weeks later.

Patient access owns check-in, but the downstream cost lands on billing, which works the denials, and on clinicians, who start visits with an incomplete history. The goal is a clean, verified record before the patient walks in.

02what the AI does, step by step

1. Send intake when the visit is bookedThe scheduling event triggers a secure link by text or email, with forms chosen by visit type and whether the patient is new or returning. Returning patients confirm existing data instead of re-entering it.
2. Capture cards and documentsPatients photograph their insurance card and ID. Extraction reads payer name, member ID, group number, and subscriber details, each with a confidence score.
3. Structure free-text historyAnswers about medications, allergies, and prior conditions are structured into lists for staff review. The model does not interpret clinical meaning or fill gaps; anything unclear stays as the patient wrote it.
4. Match the patient recordThe workflow searches the EHR for an existing patient by name, date of birth, and phone. Possible duplicates are flagged for a person rather than merged automatically.
5. Stage data in the EHRVerified-ready fields are written through the EHR's API or interface engine to a pending state, using FHIR Patient and Coverage resources where supported or HL7 v2 ADT messages where not.
6. Verify at check-inStaff see a short list of changed and low-confidence fields, confirm them against the card or with the patient, and release the update. Signed consent forms are filed to the chart as documents.

03systems it connects to

• EHR or practice management system. Epic, athenahealth, eClinicalWorks, NextGen, or a specialty system, via FHIR APIs or an HL7 interface.
• Intake front end. A HIPAA-eligible form tool or a custom patient-facing page.
• Document extraction. OCR and a model for card and form reading, running under a BAA.
• Messaging. A HIPAA-eligible SMS and email provider for intake links.

04human checkpoints

• Front desk verification. Nothing changes the legal record until a staff member confirms it at or before check-in.
• Duplicate merges. Possible duplicate charts are resolved by health information management staff, never by the workflow.
• Clinical history review. Medication and allergy lists are reconciled by a nurse or provider during the visit, as they are today.

05what to measure

• Intake completed before arrival. Share of visits with forms submitted ahead of time, by visit type.
• Check-in duration. Time from arrival to roomed, measured before and after.
• Registration-related denials. Claims denied for demographic or eligibility errors.
• Field correction rate. How often staff change an extracted value, by field.

06risks and guardrails

• PHI in every step. Intake is PHI end to end. Every vendor in the path, including the model provider, needs a business associate agreement, and access should be limited to the roles that verify intake. Keep audit logs of who viewed and released each record.
• Wrong-patient errors. A bad match writes one patient's data into another chart. Matching must stay conservative, with humans resolving anything uncertain.
• Accessibility. Not every patient can complete forms on a phone. Keep a staffed path for patients who need help, language support, or paper.

07build vs buy

Patient intake products such as Phreesia, and the intake modules many EHR vendors sell, work well when your EHR is one they integrate with deeply. Check that first.

A custom build fits when your EHR has limited vendor integrations, when intake differs sharply by specialty or location, or when you want intake data to also feed eligibility checks and scheduling rules you already run.

08related playbooks

Browse every healthcare clinic operations playbook or the full library.