FreeSeek
Talk to us
MED · Healthcare Services

Agents that carry the paperwork of care so clinicians can treat

Clinicians spend a large share of each shift on documentation, phone calls and payer forms. FreeSeek builds agents that take on this administrative layer inside your EHR and messaging channels. Diagnosis and treatment decisions stay with your clinical staff.

Systems we connect to
EpicOracle Health (Cerner)HL7 FHIRHISPACSLISWhatsApp Business API
Application scenarios Case studies
Industry context

Where healthcare operations stand today

Hospitals and clinic networks have digitized most records, yet the work around those records is still largely manual. Nurses phone discharged patients from spreadsheets. Physicians write discharge summaries after hours. Front-desk staff re-key referral details from faxes and PDFs into the HIS. Each step depends on people copying information between systems that rarely share a common data model, even where HL7 FHIR interfaces exist.

Payer requirements add a second layer. Prior authorization requests, coding queries and claim denials consume billing teams and pull clinicians into peer-to-peer calls. Meanwhile, chronic disease populations grow and value-based contracts reward outcomes measured months after a visit. Organizations need continuous contact with patients between visits, but they cannot hire enough nurses and coordinators to provide it by phone alone.

Clinician documentation load

Physicians spend hours per day in the EHR, much of it after clinic. Documentation burden is a leading driver of burnout and attrition.

Gaps after discharge

Readmission penalties and value-based contracts make the 30 days after discharge a financial and clinical risk. Most follow-up still depends on manual phone lists.

Payer friction

Prior authorization and claim denials delay care and add rework for billing teams. Each payer keeps its own rules, forms and portals.

Staffing shortages

Nurses, coders and front-desk staff are hard to recruit and retain. Administrative volume keeps rising while headcount stays flat.

Agent run · example

Watch an agent work a real task

Every step is a tool call against your own systems, logged and reviewable. The result goes to a named person for sign-off.

Agent run · examplerun_c2a57e
TaskCheck in with patients discharged after heart failure admission in the last 7 days.
  1. fhir.query()
    Found 38 patients with a heart failure discharge in the last 7 days and loaded their care plans.
  2. whatsapp.send()
    Sent the day-5 check-in. 31 replied; 1 reported a 2.4 kg weight gain in 3 days with new ankle swelling.
  3. ehr.task_create()
    Matched the cardiology red-flag rule and opened an urgent task with the readings and the patient's own words.
Result · Escalated 1 patient to the heart failure nurse for a same-day call. The nurse decides on next steps.
Application scenarios

Four workflows, analysed step by step.

Each scenario shows how the work is done today, what the agent takes over, where a person signs off, and what we measure.

MED-1Post-discharge and chronic-disease follow-upMED-2Discharge summary and referral letter draftingMED-3Patient intake and appointment routingMED-4Prior authorization and claims coding support
MED-1

Post-discharge and chronic-disease follow-up

An agent keeps scheduled contact with discharged and chronic patients, tracks adherence and escalates red flags to a nurse.

Integration depth Moderate
Typical pilot8–10 weeks
How it works today

Care coordinators work from exported discharge lists and call patients one by one. Many calls go unanswered, notes land in free text, and patients with diabetes or heart failure may go weeks without contact. Medication gaps and worsening symptoms surface only when the patient returns to the emergency department. Coverage depends on how many calls a coordinator can fit into a shift.

Systems used
EpicHL7 FHIRWhatsApp Business APIWeComTwilio
What the agent does
  1. Builds a follow-up plan per patient from the discharge order set, care plan and medication list in the EHR.
  2. Sends scheduled check-ins in the patient's language over SMS, WhatsApp or WeCom using protocol questions approved by the care team.
  3. Records answers on medication intake, home readings and symptoms as structured data and writes a summary back to the chart.
  4. Applies the clinician-defined red-flag list and routes any match to the on-call nurse queue at once, with the patient told to seek urgent care where the protocol says so.
  5. Reports adherence trends and unreachable patients to the coordinator each morning.
Human checkpoint
A nurse reviews every escalation and decides on clinical action. Clinicians approve the question sets, red-flag rules and escalation thresholds before go-live and at each revision.
What we measure
  • 30-day readmission rate
  • follow-up contact rate
  • time from red flag to nurse contact
  • medication adherence rate
MED-2

Discharge summary and referral letter drafting

An agent assembles draft discharge summaries and referral letters from the chart for the responsible clinician to edit and sign.

Integration depth Deep integration
Typical pilot10–12 weeks
How it works today

Residents and attending physicians compile discharge summaries by scrolling through progress notes, lab results and medication changes, often hours after the patient has left. Referral letters are rewritten from scratch for each specialist. Summaries arrive late at primary care, omit pending results, or contain medication lists that conflict with the final orders.

Systems used
EpicOracle Health (Cerner)LISPACSHL7 FHIR
What the agent does
  1. Pulls the admission note, progress notes, procedures, final medication reconciliation and pending results from the EHR and LIS.
  2. Drafts the summary in the hospital's template and links every statement to the source note or result.
  3. Flags conflicts such as a discontinued drug that still appears in the home medication list, or a result still pending at discharge.
  4. Prepares the referral letter in the receiving specialty's format with the relevant imaging reports from PACS attached.
  5. Places the draft in the clinician's EHR inbox, unsigned, with a change log of every edit the clinician makes.
Human checkpoint
The responsible physician reads, edits and signs every document. Nothing is sent to a patient or external provider without that signature.
What we measure
  • time from discharge to signed summary
  • clinician editing time per document
  • source-linked statement rate
  • medication discrepancy rate
MED-3

Patient intake and appointment routing

An agent collects intake information, prepares the triage queue and books the right clinic and slot under staff-defined rules.

Integration depth Moderate
Typical pilot6–8 weeks
How it works today

Referrals arrive by fax, email, portal and phone. Front-desk staff transcribe demographics and insurance, chase missing documents, and guess which clinic or clinician fits. Mismatched bookings waste specialist slots, and urgent referrals can sit in a shared inbox for days. Triage nurses receive incomplete information and have to call patients back before they can prioritize.

Systems used
HISEpic CadenceHL7 FHIRe-fax gatewayWhatsApp Business API
What the agent does
  1. Reads incoming referrals and portal messages, extracts demographics, insurance, reason for referral and attached results.
  2. Contacts the patient through the portal or messaging channel to collect missing intake forms and consent.
  3. Checks referral completeness against each clinic's acceptance criteria and drafts a priority suggestion with the supporting facts for the triage nurse.
  4. Proposes clinic, clinician and slot by matching specialty rules, payer network and patient preference, then books once triage confirms.
  5. Sends reminders and preparation instructions and refills cancelled slots from the waitlist.
Human checkpoint
A triage nurse sets the clinical priority of every referral. Staff confirm bookings for urgent cases and any referral the agent cannot match to a rule.
What we measure
  • referral-to-appointment time
  • incomplete referral rate
  • no-show rate
  • specialist slot utilization
MED-4

Prior authorization and claims coding support

An agent prepares prior authorization packets and suggests codes with chart evidence for coders and billing staff to approve.

Integration depth Deep integration
Typical pilot8–12 weeks
How it works today

Billing staff look up each payer's authorization rules, copy clinical notes into portal forms and track responses in spreadsheets. Coders read full encounters to assign ICD-10-CM and CPT codes and query physicians by message when documentation is thin. Denials often trace back to missing documentation that existed in the chart but never reached the payer.

Systems used
Epic ResoluteX12 278/837HL7 FHIRpayer portalsICD-10-CM / CPT code sets
What the agent does
  1. Checks scheduled procedures and medications against the payer's current authorization requirements.
  2. Assembles the request packet with the relevant notes, imaging and lab evidence mapped to the payer's criteria.
  3. Suggests ICD-10-CM and CPT codes for completed encounters, each tied to the supporting sentence in the note.
  4. Drafts compliant physician queries where documentation does not support specificity, without leading the physician toward a code.
  5. Tracks submissions and denials and drafts appeal letters with the evidence the payer cited as missing.
Human checkpoint
Certified coders accept or change every code before a claim is released. Billing staff submit each authorization, and physicians answer every documentation query themselves.
What we measure
  • authorization turnaround time
  • initial denial rate
  • coder throughput
  • query response time
Case studies

How engagements in this field have played out.

Illustrative cases: composites of typical deployments with client details removed. Figures show the range our pilots are designed to reach.

Heart failure follow-up reaches patients the phone list missed

Client: Regional hospital group, 4 acute hospitals, cardiology service lineRegion: Southern Europe
84%patients contacted within 7 days, up from 46%
−21%30-day heart failure readmissions
−35%coordinator time spent on routine calls
Challenge

Heart failure readmissions within 30 days were above the national benchmark. Two coordinators covered roughly 300 discharges a month by phone and reached fewer than half within a week. Contacts were logged as free text, so cardiologists could not see trends in weight or symptoms between visits.

What we built
  • Built a WhatsApp and SMS follow-up agent with question sets and red-flag rules written by the cardiology nurses.
  • Wrote structured answers and daily summaries back to the EHR through FHIR Observation resources.
  • Routed red-flag answers to the heart failure nurse queue with a 2-hour response standard.
  • Kept all data in an EU-hosted environment under a GDPR data protection impact assessment.
Timeline: 10-week pilot in 1 hospital, rolled out to all 4 by month 6

Discharge summaries signed on the day of discharge

Client: Private tertiary hospital, 600 beds, internal medicine and surgeryRegion: GCC
−42%physician time per discharge summary
71%summaries signed on the day of discharge, up from 18%
−30%medication discrepancies found in audit
Challenge

Discharge summaries were signed a median of three days after discharge, and primary care physicians complained of missing pending results. Residents reported spending evenings on summaries. The hospital required that any tool keep patient data in-country and leave final wording to the physician.

What we built
  • Deployed a drafting agent inside the hospital's network, reading from the EHR and LIS through FHIR APIs.
  • Linked every sentence in the draft to its source note and highlighted pending results and medication conflicts.
  • Delivered drafts to the physician's EHR inbox unsigned and logged all edits for quality review.
Timeline: 12-week pilot on 2 wards, production in month 5
Risk and compliance

Patient data, regulation and clinical safety

Health data protection

Deployments follow HIPAA in the US, GDPR Article 9 for health data in Europe and China PIPL rules on sensitive personal information, including separate consent and data localization where required. We sign business associate or data processing agreements and keep data in the client's chosen region.

Kept on the administrative side

Software that diagnoses or recommends treatment can qualify as a medical device under FDA SaMD policy, EU MDR Rule 11 and NMPA rules. Our agents draft, collect, route and remind, while diagnosis, triage priority and treatment remain decisions made by licensed clinicians.

Clinician in the loop

Clinical content goes out only after a named clinician has approved it, either as a signed document or as a pre-approved protocol. Every draft shows its sources so reviewers can check it quickly.

Audit logs and red-flag escalation

Each agent action, data access and human approval is logged with user, time and source record. Red-flag rules are written by clinicians, escalate immediately to a staffed queue, and are tested before every change.

Questions we hear
Do your agents diagnose patients or recommend treatment?

No. They handle documentation drafts, intake, scheduling, follow-up and payer work. Any clinical judgment, including triage priority, is made by your licensed staff, which also keeps the agents outside medical-device scope.

Which EHR systems can you connect to?

We integrate with Epic, Oracle Health (Cerner) and regional HIS platforms through HL7 FHIR, HL7 v2 or vendor APIs. Where no API exists, we agree a read-only extract with your IT team.

What happens when a patient reports a dangerous symptom?

The agent matches the answer against red-flag rules your clinicians wrote and routes it at once to a staffed nurse queue. It tells the patient to seek urgent care or call emergency services when the protocol requires it.

Where is patient data stored and processed?

In the region you choose, including your own data center or a local cloud region. Model calls run on endpoints covered by the same data agreements, and patient data is never used to train shared models.

Have a workflow like these?

Send us a short description of how the work is done today. We reply within two business days.

Start a conversation
Other industries
MFGAdvanced ManufacturingMATNew Materials R&DENGEngineering DesignFINFintechMKTMarketing