Agent · 100% on-premise · our own reasoner

Patient Flow

30-day clinic load forecast with seasonality detection.

Request demo View the full agent network

What this agent does

Patient Flow is the agent that projects the clinic’s expected demand for the next 30 days. It combines time series models (Prophet/SARIMA) with the actual schedule to detect seasonality, trends and load peaks by day, hour and specialty. All analytics run locally on the server.

Why it matters

The manager planning shifts and staff vacations blindly under- or over-shoots. Patient Flow gives a statistically grounded projection of the next 30 days to decide how many hygienists are needed Tuesday morning, or whether orthodontics should be reinforced in September. The decision remains with the manager, but stops being pure intuition.

How it works

The agent reads the full appointment history and applies a time series model per specialty and per 30-minute slot. It detects weekly and yearly seasonality, general trend and atypical events (long weekends, school holidays). It recomputes every 24h and emits a heatmap of expected demand plus schedule recommendations.

Integration with the clinical workflow

Patient Flow passes signal to Inventory to anticipate consumption by specialty and to Reception to prioritize re-booking campaigns on slots detected as underused. The manager receives the dashboard weekly with executive summary.

Autonomous decisions it makes

Recompute the 30-day forecast on each scheduled 24h tick
Detect weekly and yearly seasonality per specialty
Flag underused slots and propose a re-booking campaign
Flag future load peaks for staff planning
Notify the manager when a projected day exceeds nominal capacity

Inputs and outputs

Receives

· Full appointment history (minimum 12 months recommended)
· Specialty catalog and average duration per service type
· Working calendar and local holidays
· Clinic nominal capacity (chairs × hours)

Produces

· 30-day demand heatmap per specialty
· Identification of underused gaps and over-capacity peaks
· Staff planning recommendations
· Weekly executive summary to the manager

Production metrics

30 days

Forecast horizon

Every 24h

Recompute

30 min slot

Granularity

Collaborates with

Tech stack

Model

Prophet / SARIMA per specialty

Execution

Local CPU (classical time-series models)

Latency

Nightly recompute, immediate query

Privacy

Model trained with data from the clinic only

Frequently asked questions

Do I need minimum history for it to work?+

We recommend at least 12 months for the model to learn annual seasonality. With 6 months it already gives useful forecast but with lower precision for seasonal peaks. Each new month refines the model.

Does it work if I open a new clinic?+

Yes, in conservative mode. Without own history, the agent uses benchmarks from similar clinics (specialty, size, location) until enough history is accumulated. It is clearly marked as benchmark, not own prediction.

Can I see the forecast on my tablet during consultation?+

Yes. The dashboard is accessible from any device connected to the local clinic server. It does not require internet connection.

Can the manager act on underused gaps?+

Yes. The recommendation connects with Reception's message templates for targeted re-booking campaigns aimed at the right cohort (see Patient Segmentation).