Sagan: Governed AI Moderation for HCP Interviews: an independent product by Kaushal Khodifad

Sagan: Governed AI Moderation for HCP Interviews is an independent product by Kaushal Khodifad in healthcare & life sciences. Watch a full AI-moderated oncologist interview in about a minute. Code decides what to ask and checks every question before it is sent, adverse events are caught against EPHMRA criteria and put on a one-business-day clock for human review, and every finding cites a transcript turn that is verified to exist.

Watch a full AI-moderated oncologist interview in about a minute. Code decides what to ask and checks every question before it is sent, adverse events are caught against EPHMRA criteria and put on a one-business-day clock for human review, and every finding cites a transcript turn that is verified to exist.

  • Live AI moderator, code-gated
  • EPHMRA adverse-event engine
  • Hash-chained interview record
Open product

Independent work by Kaushal Khodifad on public and synthetic data. Where a real company is named, it anchors the domain only: this is not client work, and it is not affiliated with or endorsed by any company named.

Sagan is a governed AI moderator for structured HCP interviews in pharma market research. A deterministic engine runs the discussion guide and decides every next question; a language model phrases part of it, and a fifteen-rule gate blocks double questions, leading options, praise, clinical advice, product claims, safety judgements and brand names the respondent has not said before anything is sent. The respondent is told at the first turn that they are speaking with an AI, and is never interrupted: adverse events, special reporting situations and product complaints are detected against EPHMRA 2025 criteria, including a group of patients, and followed up case by case at the end. Each one is stored with the detector version, starts a one-business-day clock in the interview country's calendar, waits for a human decision, and exports as an AER-shaped form with a per-country reconciliation that includes zero rows; a human also reads every transcript. Synthesis codes every answer against a written codebook with two coders, a validator drops any quote that is not word for word, and themes count only where the coders agree or a human adjudicated. Transcripts, consents and decisions are hash-chained. Respondents are synthetic, scripted or LLM-simulated, and labelled as such everywhere. ZoomRx ships a product also named Sagan; this independent product is not affiliated with or endorsed by ZoomRx.

What it does

  • A deterministic engine runs the discussion guide and decides every next question; a language model phrases part of the interview, and a fifteen-rule gate checks every question before it is sent.
  • Adverse events, special reporting situations and product complaints are detected against EPHMRA 2025 criteria, stored with the detector version and put on a one-business-day clock for a human decision; a reviewer also reads every transcript.
  • Follow-up happens case by case at the end of the interview, never mid-guide; each case exports as an AER-shaped form, reconciled per country with zero rows.
  • Synthesis codes every answer with two coders, drops any quote that is not word for word, and counts a theme only where the coders agree or a human adjudicated.

What is live and what is simulated

  • Code: the interview engine, the gate, the identifier screen, adverse-event detection, the clock, the rules coder and themes.
  • Live model, in a reserved segment of up to 10 questions per interview: question phrasing, the second coder, a safety second look and cross-interview questions. Recorded output is labelled with its date and model.
  • Simulated: the respondents (scripted or LLM-simulated), verification, PV forwarding, attestations and reviewer identity. Nothing leaves the site.
  • People: reportability decisions, transcript review and coding adjudication.

About the name

ZoomRx ships an AI market-research platform called Sagan (launched June 2025) and markets it today as Sagan Agents (launched 20 May 2026). This Sagan is an independent product designed and built by Kaushal Khodifad. It is not affiliated with or endorsed by ZoomRx, uses none of its code or data, and nothing here describes how ZoomRx's product works.

All products