How intake works

From a photograph or a note to a set of facts on the register.

Intake is the path from a meeting to a structured client record. Every channel ends in the same pipeline.

Channels

• Console upload: photo, PDF, Word document or typed notes from the Intake tab. Works well on a phone.
• Telegram: send to the practice bot; a note becomes an intake and starts a draft automatically.
• WhatsApp: send to the practice number; drafting is started by hand afterwards.

The pipeline

1. Queued. The file is stored in the practice vault and a job is queued.
2. Extracting. A vision model running on the practice server reads the file and fills a structured schema: personal details, spouse or partner, dependants, income and other income, expenses, assets, liabilities, existing cover, retirement contributions, objectives, risk profile, and anything contradictory or missing.
3. Second pass. A verification pass compares the extraction with the source. Disagreements and unsupported values are flagged on the review form; safe fills are applied only where the source supports them.
4. Derived figures. Totals, loan terms, projected asset values, expense calendars and a tax position are calculated on the server. See Review form.
5. Pending review. The intake waits for a person. Commit and Draft stay disabled until every flag has been checked.
6. Committed. The client is created or updated on the register with consent off, and linked to a household where a spouse is present.

Identity stays on the server

Extraction runs locally, so the file and the identity fields never leave the practice server. When a draft is started later, identity is stripped and free text is scrubbed before anything is sent to the drafting model. See the Security page.