Last updated: 21 July 2026
AI does the paperwork; it does not do your thinking. This page lists every place AI touches your portfolio — and the one place it deliberately holds back. It is written in scenarios, not model names, and we keep it current: when the product's AI behaviour changes, this page changes with it.
Forward an email, upload a certificate, share a photo, talk into a voice note or record a whole lecture, and AI reads it to draft the portfolio entry for you: a title, the date, an estimate of CPD points, a first-person summary of what the activity was, and a suggested categorisation against your profession's framework. It also flags anything that looks like personal information so you can deal with it before the entry is saved, and spots when two dumps look like the same event so you can combine them.
After a lecture or course you can write, paste or dictate your raw notes ("Write a note"). AI distils them into up to ten gems — the facts and insights worth remembering — and, only where you expressed an intention ("TODO…", "ask the department…"), marks a gem as an action: the to-do kind of gem, shown under "To do". Anything you emphasised in your notes — bold, highlights, headings — is treated as what mattered most to you, and your wording is kept. You can edit, re-sort or delete every item, and your original notes are kept word-for-word on the entry, one tap away. The same extraction runs on other evidence too, from the evidence itself.
When you review an entry, your own words get one notes box — prefilled verbatim with whatever you supplied (the note you typed on a forwarded email, your voice-note transcript, your pasted notes), never rewritten. Pressing Next runs one AI pass over your facts and notes: a two-sentence details line, your reflection answers drafted from your words, the gems, and the categorisation — all editable before you approve. Categories are filed automatically and can be changed any time from the activity. Gems themselves are opt-in per entry: unless you press "Generate gems" at review (or later on the activity), the extraction is discarded and nothing is kept or resurfaced.
Your gems come back to you for revision: a Gems page, an optional daily "morning gem" phone notification, and optional weekly and monthly summary emails. These only ever contain your own approved content, and every one of them can be switched off in settings — or silenced per-gem by marking it done.
Every reflection in your portfolio started with you. AI never invents one.
If your own words contain reflection — a comment you typed above a forwarded email, musings in a voice note — AI shapes those words into your profession's reflection answers, and fills only the answers your words actually support. If your evidence is purely factual (a certificate, an agenda, "there was an MDT on Tuesday"), the reflection boxes stay empty and the app invites you to talk your reflection through in your own words first. The "shape into reflections" button then tidies your ramble into well-formed answers — tidying, never replacing.
One deliberate exception: recorded lectures, where AI may propose reflection drafts from the lecture content itself — see "Recording a lecture" below.
Any text box with the ✦ sparkle can polish what you wrote into flowing prose — keeping your meaning and your specifics, never adding facts you didn't give it. One tap undoes it.
Voice notes and the in-app microphone are transcribed by AI into text you can read and edit before anything else happens with it. The transcript is yours to correct — the audio is handled under the retention rules on the privacy page.
Lecture mode records a whole talk in ~10-minute parts. AI transcribes each part (deleting its audio the moment it is read), stitches the full transcript, and drafts your entry from it — treating the minutes you marked with the emphasise button, and any notes you typed during the talk, as the strongest signals of what mattered to you. The transcript is kept as your entry's own notes.
Lectures are the one place AI may draft reflections you didn't dictate: when you typed nothing and marked nothing, it can propose reflections drawn from what the lecturer actually said — clearly labelled as drafts from the lecture content, for you to rewrite or discard. It still never invents opinions, feelings or commitments for you, and if the recording can't support a genuine learning point, the reflections are left blank for you.
When several dumps turn out to be the same event — an email, a certificate and a voice note about one course — AI drafts the combined entry: one summary and one woven answer per reflection question, drawing only on what the sources contain. Anything you wrote yourself takes precedence over AI drafts.
At appraisal time, AI can draft your appraisal-period report and answer appraisal-form questions in your voice — grounded only in the activities and reflections you have actually evidenced. If it isn't in your portfolio, it doesn't go in the report.
AI processing changes nothing about retention: what we store, for how long, and what gets scrubbed after analysis is set out in the privacy policy — see "What happens to your uploads". Questions: hello@cpddump.com.