Riftveil Essentials
Reviewing AI outputs before you act on them
A half-day workshop on the method. You review a real AI output on five levels, set its stakes, name who must verify what, and leave with a complete report on an output from your own work.
- Format
- In-person workshop
- Duration
- Half-day, 3 h 30, two breaks
- Group
- Up to 12 participants, at tables of 4 to 6
- Led by
- A Human Frontier certified trainer
- Prerequisites
- You work with AI-generated outputs, your own or your colleagues’. No technical background.
- Before the workshop
- One real AI output from your work, redacted (optional). Without Human Frontier Essentials: the diagnostic and chapter 1, about 30 minutes (recommended).
- Certificate
- Riftveil Essentials, numbered and verifiable online
- Leads to
- With Human Frontier Essentials: Advanced
What you will learn
- Review an AI output on five levels of analysis
- Focus your scrutiny on what is really at stake
- Identify the unverified assumptions, and who is accountable for each
- Produce a report that keeps the final decision human
Who it is for
Professionals who act on AI-generated analyses, drafts and recommendations: managers, analysts, legal and compliance, finance, procurement, medical and regulatory affairs, consultants. Human Frontier Essentials is the natural first step, but not a requirement: the opening includes a short recap of the Human Frontier model.
Learning objectives
By the end of the workshop, each participant can:
- State the risk the method addresses, the test criterion, and the difference between decision-checking and fact-checking.
- Name the five levels, having seen at least one finding at each of Levels 0 to 3 on the same output.
- Place at least three of their own decision types on the triage grid (reversibility, impact scope, financial commitment on the scale of the organisation) and state the escalation rule.
- Produce alone, in writing, at least one unstated assumption and three questions on an output they had not seen before.
- Complete a report on one output: a triage line, at least two findings with their impact, two assumptions each with a named verifier, and three questions.
- Name two or three recurring patterns and one output type to which they will apply Riftveil Lite.
Programme, 3 h 30
| Step | Time | What happens |
|---|---|---|
| Opening and where Riftveil sits | 20 min | The risk; a recap of Human Frontier; decision-checking, not fact-checking |
| The five levels, live | 45 min | The trainer and the room review one projected output, level by level |
| Triage by stakes | 20 min | Each table places its own decision types on the grid |
| Break | 10 min | |
| Review it yourself, then compare | 50 min | Alone and in silence, then at the table, then comparison with a reference reading |
| Break | 10 min | |
| The report on your own output | 40 min | On the output you brought, or a library exercise; alone or with a chosen partner; sharing is voluntary |
| Closing: Riftveil Lite and one commitment | 15 min | Recurring patterns on the flipchart, one commitment, closing question |
Three moments are never cut: “review it yourself”, the report on your own output, and the closing. Durations are indicative.
Rules for the output you bring
Remove or replace any client name, confidential detail, personal data or figure that would identify a real deal or person. Nobody is required to show their output to the room. The trainer never keeps a copy, photographs it or asks to see the AI conversation behind it, and nothing about it is kept after the session. If you followed Human Frontier Essentials, bring the output behind the decision you marked at the end of that workshop: the two workshops then work on the same decision.
How it is taught
Every exercise runs in the same order: alone and in silence, then at the table, then with the whole room. Nobody is asked to share their own result, their own decisions or their own documents. A quiet corner is available for anyone who prefers not to discuss their own work in public.
What you receive
- A triage table for your own decision types
- One completed report on your own output or a library exercise
- Riftveil Lite: five questions on one page, for use after the workshop
- The Riftveil Essentials certificate, verifiable online
How learning is checked
- In the room, for the session as a whole and never by name: at least one unstated assumption and three questions for every participant, every report complete.
- The certificate is issued once you have attended the full workshop and completed your report.
- Follow-up on day 1, day 7 and day 30, by email or an optional group call.
Confidentiality
- Your report and your workbook belong to you.
- Your organisation receives attendance and an aggregated, anonymous summary, never an output, a report or an individual result.
- The public verification page shows only the holder’s name, the certificate and its date.
What this programme is not
Not tool training and not fact-checking: Riftveil checks the decision an output leads to, not every sentence in it. The workshop teaches the method, not the software. The certificate records a completed programme. It does not certify a competence and does not make an organisation compliant with any regulation. It can form part of the record an organisation keeps of the measures it takes to develop AI literacy.