Learn
The handbooks and the rubric.
What candidates and assessors are told, and the criteria every case analysis and portfolio is scored against. Weightings are initial calibration parameters.
GOVBRM Academy candidate handbook
Status: draft v0.1, GOVBRM original, not yet validated.
What this handbook is
The plain account of what you sign up for when you book a place on a practitioner cohort or, from 2027, the master portfolio assessment. It is written to the certification architecture published by GOVBRM Academy and follows it exactly. If a sales page and this handbook disagree, this handbook is right and we will fix the page.
What the credentials are, and are not
GOVBRM Academy credentials are continuing professional development credentials issued by an independent publication. They are not academic qualifications, they are not regulated qualifications, and they are not accredited by any third party or professional body. Nobody has yet independently validated the method they teach. Please describe them accurately; misrepresenting one is grounds for revocation.
| Tier | Credential | What passing proves |
|---|---|---|
| Foundation (free) | AI Demand Foundation, AI Value Associate, Responsible AI Practitioner | You read the course and passed a ten-question check. Nothing about competence. |
| Practitioner | Certified AI Demand Strategist, Certified AI Value Orchestrator | You applied the GOVBRM method to a case of your own to a published standard, and an assessor agreed. |
| Master (2027) | Master AI Value Partner | You ran a real item from request to realised value using the whole method, and two independent assessors agreed. |
Before you book a practitioner track
- Do the free micro-courses first. They are not required, but the masterclasses assume you know the vocabulary: the need sentence, value as a range, the weakest dimension, the six routes, the autonomy ladder, month nine.
- Have a case in mind. You will write your case analysis on a real request from your own organisation, or a realistic one if a real one cannot be shared. You do not need permission to use a real case if you can anonymise it so that neither the organisation nor any individual is identifiable. If you cannot, use a realistic case and say so. Realistic cases are scored on exactly the same rubric.
- Check the dates. Two live masterclasses on consecutive days, 14:00 to 18:00 UK time, online. You need to attend both in full. The cohort is twelve people at most.
- Check the price. 495 (GBP) a seat, or included with a founding membership. Substack and Stripe charge in your local currency.
What happens on a practitioner track
| When | What |
|---|---|
| Two weeks before | Joining instructions, the rubric for your track, the toolkit canvases you will use, the participant workbook with the case pack, and this handbook. |
| Masterclass one | Four hours live, 14:00 to 18:00 UK time. Teaching, and activities in pairs and trios on the case with feedback out loud. |
| That evening | Thirty to forty-five minutes: finish the canvas fields started in the room and read the case documents for masterclass two. |
| Masterclass two | Four hours live the next afternoon. Teaching, the scenario group activity in threes or fours (formative, not scored), and how to write your case analysis. |
| Within seven days after masterclass two | Sit the knowledge check online. |
| Within fourteen days after masterclass two | Submit your case analysis, worked through the canvases for your track. |
| Within fourteen days of submission | Your score sheet and written feedback. |
| Within a week of a pass | Your PDF certificate, your badge and your registry entry. |
The three things you are assessed on
Attendance
Both masterclasses, in full. Missing more than thirty minutes of one counts as missing it. If you miss one for a good reason (illness, caring, an emergency at work), tell us and you can make it up in the next cohort of the same track at no charge. Your case analysis and knowledge check wait until you have.
Your individual case analysis
- Two pages at most, about 1,000 words. Anything after page two is not read.
- Attach the completed canvases for your track as an appendix. Canvases are evidence, not decoration; they are not scored for tidiness.
- It must be your own work. You may use an AI assistant to help you think and draft, in the same way you might use one at work; you must say that you did, and the reasoning and the decisions must be yours. We may ask you about your case at any point, and a candidate who cannot explain their own analysis has not written it.
- It is scored by an assessor against the published rubric. You get the rubric before you start. Every criterion has four levels and a written description of each. There are no hidden criteria.
- Borderline results and every not-yet-passed result are second-scored by a moderating assessor before you hear anything. A sample of passes is second-scored too, so that the standard is the same for everyone.
The knowledge check
Twenty questions, online, after the second masterclass. Fourteen correct passes. It tests whether you can apply the reasoning, not whether you memorised the slides. If you do not pass, you can sit it again once, after seven days, at no charge.
If you do not pass first time
- Case analysis: you get written feedback that names the criteria not met and what a resubmission needs to show. You may resubmit once, within sixty days of the feedback, at no charge. A resubmission is scored fresh against the same rubric by the same assessor where possible, and moderated.
- Knowledge check: one further attempt after seven days, as above.
- If after the resubmission and the second attempt you have still not passed, the result is "not yet passed". You may join a later cohort at the standard price, or ask about a discount; no promise is made about a discount. Nothing about a not-yet-passed result is published anywhere.
Appeals
You can appeal a result on the grounds that the rubric was not applied, that the process was not followed, or that there was a conflict of interest. You cannot appeal simply because you disagree with an assessor's judgement. The two-stage process is in GOVBRM Academy moderation, appeals and resubmission policy, which you receive with your feedback. During the beta the final stage is decided by the Director of Studies, who may also have been your assessor; we say so plainly because there is not yet anyone else, and it is a reason to appoint a Method Council.
Your credential
- The certificate carries your name as you give it, the credential, the guided learning hours, a credential ID and a verification link. Example credential ID: GBRM-DS-2026-1234. The year is the syllabus year, not an expiry.
- It does not expire. Every three years you will be invited to make a short continuing professional development declaration; it keeps your registry entry marked current. If you do not, the entry says "Revalidation due" and nothing else changes.
- The public registry at govbrm.com/verify lists your credential ID, the credential, the issue date and the status. Your name appears only if you opt in. You can change your mind at any time.
- Describe it as what it is: a GOVBRM Academy practitioner credential. Do not call it a degree, a diploma, a licence, a regulated or accredited qualification. Doing so is grounds for revocation.
Master AI Value Partner (from 2027)
You need both practitioner credentials. You submit one real item taken from request to realised value, with the evidence from each gate it passed. Realistic cases are not accepted at this tier. Two assessors score it independently against a published rubric; then you defend it for forty-five minutes online, where practical. One resubmission within ninety days at no charge. 150 GBP (GBP), or included with a founding membership. The full candidate guidance for the master tier is published before bookings open.
Your data
Accessibility and adjustments
Tell us before the cohort what you need: extra time on the knowledge check, a different submission format, captions, a break pattern, an alternative to the oral defence at master tier. Reasonable adjustments do not change the standard; they change how you show you have met it. The site is built to WCAG 2.2 AA but has not yet been independently audited.
Where to ask
Questions before booking go to the GOVBRM LinkedIn page. Once you are on a cohort you get a direct contact in your joining instructions.
Director of Studies, GOVBRM Academy
GOVBRM Academy assessor handbook
Status: draft v0.1, GOVBRM original, not yet validated.. Assessor appointment terms require legal review.
Who this is for
Anyone who scores a practitioner case analysis or a master portfolio for GOVBRM Academy, or who moderates another assessor's scoring. During the beta that is the Director of Studies and any invited moderating assessor. Read GOVBRM Academy certification architecture first; this handbook does not repeat the model, it tells you how to apply it.
Your job in one paragraph
You decide whether a piece of work meets a published standard. You do not decide whether you like it, whether you would have done it differently, or whether the candidate seems capable. The rubric is the standard. If the rubric is wrong, say so in the change log and score against it anyway; it changes for the next cohort, not this one.
Before the cohort
- Confirm you have signed the assessor declaration in GOVBRM Academy assessor conflict of interest and independence policy for this cohort.
- Read the current rubric for the track. Note its version and date; you will cite it on every score sheet.
- Check the candidate list against your conflict of interest register. Declare any conflict before the first masterclass.
During the masterclasses
- Take attendance at the start, after the break and at the end. Record partial attendance with times; a candidate who misses more than thirty minutes of a masterclass has not attended it in full.
- The scenario group activity is formative. Give feedback out loud and generously. Nothing said in the group activity counts towards the score.
- Do not pre-score anyone's case in your head. The written analysis is the evidence.
Scoring a practitioner case analysis
What arrives
- A written analysis of two pages at most (about 1,000 words). Anything beyond two pages is not read; tell the candidate you stopped at the end of page two.
- The completed toolkit canvases for the track, as an appendix. Canvases are read for evidence; they are not scored for neatness.
- The candidate's declaration that the work is their own and that the case is real or realistic as stated.
How to score
- Read the whole submission once without a pen.
- Score each criterion in the rubric at one of the four levels. Write one sentence of evidence for each level you award, quoting or pointing to the submission. A score without evidence is not a score.
- Add the weighted points. Apply the pass threshold and the essential criteria rule in the rubric.
- Decide the outcome: pass, borderline, or not yet passed. Borderline is defined in the rubric as within the tolerance band above or below the threshold, or any pass with an essential criterion at level one.
- Write the feedback. Format below.
- Send the score sheet and feedback to the moderating assessor if the result is borderline or not yet passed, or if the submission is in the moderation sample. Do not tell the candidate anything until moderation is complete.
Score sheet fields
| Field | Content |
|---|---|
| Candidate reference | Cohort code and seat number, not the name |
| Track and rubric version | For example DS rubric v0.1 dated (date) |
| Case type | Real or realistic, as declared |
| Per criterion | Level awarded, points, one sentence of evidence |
| Total points | Out of 100 |
| Essential criteria | Met or not met, each |
| Outcome | Pass, borderline, not yet passed |
| Moderation | Required (reason) or sampled or not required |
| Assessor | Role and date |
Feedback format
Feedback is written to the candidate and is theirs to keep. Use this shape, four short paragraphs at most:
- The outcome in one sentence.
- What met the standard, naming the criteria.
- What did not, naming the criteria and saying exactly what a resubmission would need to show. Point to the canvas or the section of the analysis.
- What to do next: the resubmission window, the knowledge check, or the certificate timeline.
Write in the second person. No hedging, no praise sandwich, no jargon. A candidate should be able to act on it without asking you a question.
Scoring the knowledge check
The knowledge check is scored automatically. Your job is to check the question bank before each cohort for any item the masterclass did not cover, and to remove it. Twenty questions, fourteen to pass. One further attempt after seven days. If a candidate disputes a question, record the dispute and refer it to the Director of Studies; a question found to be faulty is voided for every candidate in that cohort and the pass mark is scaled to the nearest whole number that keeps the same proportion.
Scoring a master portfolio (from 2027)
- Two assessors score independently. Do not discuss the portfolio with the other assessor until both score sheets are lodged with the Director of Studies.
- Read the portfolio in full, including the evidence at each gate. Ask yourself at each gate: is there a record that this gate was actually passed on a real item, by real people, with real consequences, or is there a description of what should have happened.
- Score against the master rubric. Realistic cases are not accepted at this tier; if the item is not real, stop and refer to the Director of Studies.
- Prepare five to eight questions for the oral defence, each aimed at a criterion where the evidence is thinnest.
- Oral defence: forty-five minutes, online, recorded with the candidate's consent (audio only is acceptable). Ten minutes for the candidate to present, thirty for questions, five to close. Both assessors attend. Score the defence criteria after the call, independently, before speaking to each other.
- Lodge both score sheets. Where both pass, the result is pass. Where both are not yet passed, the result is not yet passed. Where they differ, or either is borderline, the moderation process applies.
Moderation duties
If you are the moderating assessor:
- Score the submission blind: do not read the first assessor's score sheet until yours is lodged.
- Compare. Where the two outcomes match, record "moderated, agreed" and the result stands. Where they differ, follow GOVBRM Academy moderation, appeals and resubmission policy.
- Look for patterns across the sample, not only individual results: is the first assessor consistently generous or harsh on one criterion. Report patterns to the Director of Studies as a note on the cohort, not as a change to any individual score.
What you must never do
- Score a submission from anyone with whom you have a conflict of interest that has not been declared and cleared.
- Give a candidate their outcome before moderation is complete where moderation is required.
- Change a score after moderation without recording why.
- Coach a candidate on their own case analysis between the masterclasses beyond the formative feedback everyone receives in the group activity.
- Describe any credential as accredited, academic or regulated. None is.
- Use candidate submissions as marketing material, exemplars or content without written consent.
Record keeping
Time budget per cohort (planning assumption, Hypothesis)
| Task | Per candidate | Cohort of twelve |
|---|---|---|
| Reading and scoring a case analysis | 40 minutes | 8 hours |
| Feedback | 15 minutes | 3 hours |
| Moderation (sample of three plus borderlines) | 40 minutes each | 2 to 4 hours |
| Knowledge check administration | 5 minutes | 1 hour |
Not yet demonstrated. Adjust after the first cohort and record the actuals.
Director of Studies, GOVBRM Academy
GOVBRM Academy assessment rubric and marking scheme
Status: draft v0.1, GOVBRM original, not yet validated.. All weightings and thresholds are initial calibration parameters, set by judgement before any cohort has run, and are to be reviewed after each of the first three cohorts.
How the scheme works
Every assessed submission is scored against criteria. Each criterion is scored at one of four levels. Each criterion carries a weight in points; the level multiplies it. Totals are out of 100. The pass threshold and the essential criteria rule decide the outcome.
Levels
| Level | Name | Multiplier | Meaning |
|---|---|---|---|
| 1 | Absent or wrong | 0 | The criterion is not addressed, or the work does the opposite of what the method asks. |
| 2 | Partial | 0.5 | The criterion is attempted but the method is applied loosely, inconsistently or without evidence. |
| 3 | Met | 0.8 | The criterion is met as the method describes, with evidence, in a way a colleague could follow. |
| 4 | Met with judgement | 1.0 | Met, and the work shows the candidate choosing, adapting or declining a step for a reason the method would recognise. |
Points for a criterion equal weight multiplied by the level multiplier. Level 3 on everything gives 80 points.
Outcomes
| Outcome | Rule |
|---|---|
| Pass | Total at or above the threshold and every essential criterion at level 3 or above. |
| Borderline | Total within five points either side of the threshold, or total at or above the threshold with any essential criterion at level 2. Always moderated. Moderation decides pass or not yet passed. |
| Not yet passed | Total more than five points below the threshold, or any essential criterion at level 1. Always moderated. |
Thresholds are initial calibration parameters:
| Assessment | Threshold (points out of 100) |
|---|---|
| Practitioner case analysis, either track | 60 |
| Master portfolio, written | 65 |
| Master oral defence | 65 |
Rationale for 60 (Hypothesis, not yet demonstrated): a candidate at level 3 on every criterion scores 80; a candidate at level 2 on everything scores 40. The threshold sits where a candidate has met most criteria properly and attempted the rest. Review after three cohorts against the distribution of scores and the moderators' notes.
Rubric A: Certified AI Demand Strategist case analysis
The case: a real or realistic AI request arriving at the front door, taken through Discover, Assess and Prioritise to a routing decision. Canvases required: 02, 03, 04, 05, 06, 07, 08, plus the route.
| Code | Criterion | Weight | Essential | Level 1 | Level 2 | Level 3 | Level 4 |
|---|---|---|---|---|---|---|---|
| DS1 | The need sentence closes with no tool in it, and the trigger and counterfactual are stated | 15 | Yes | The request is restated as the need, or a tool is named in the need | A need is stated but the trigger or counterfactual is missing or vague | Need, trigger and counterfactual are each stated and distinguishable from the request | As level 3, and the analysis shows where the request and the need diverged and what that changed |
| DS2 | Value is stated as a range with the partner's unit, a lower bound they would act on, and conditions for the upper half | 15 | Yes | A single number, or no value, or a unit the partner would not recognise | A range without conditions, or conditions that cannot be checked | A range in the partner's unit with checkable conditions | As level 3, and the lower bound is tied to a decision the partner would actually take |
| DS3 | Readiness is scored on evidence seen, dimension by dimension, and the verdict follows the weakest dimension | 12 | Yes | Readiness asserted without evidence, or the verdict ignores the weakest dimension | Some dimensions evidenced; the verdict is an average rather than the weakest | Every dimension scored on stated evidence and the verdict follows the weakest | As level 3, and the analysis names what evidence would move the weakest dimension |
| DS4 | The risk flag is raised or not with a reason, and the register row has a cause, an owner and a condition | 12 | No | No risk treatment, or a risk list with no owner | A flag with a reason but the register row is incomplete | Flag with a reason; register row with cause, owner and condition; the four agent questions asked where an agent is proposed | As level 3, and the row routes to an assurance the organisation already has |
| DS5 | The value map and prioritisation use shared criteria the sponsor could repeat | 12 | No | Ranking by preference, or no ranking | Criteria stated but applied inconsistently | Shared criteria stated and applied; the sponsor could reproduce the rank | As level 3, and the analysis shows one rank the criteria produced that the candidate found uncomfortable, and kept |
| DS6 | The portfolio heatmap places the request against the whole portfolio and yields decisions for leadership | 8 | No | No portfolio view | A heatmap with no decisions drawn from it | Heatmap with up to three decisions for leadership, each with a reason | As level 3, and a decision that declines or defers something |
| DS7 | The request ends in exactly one of the six routes, with reasons the requester could respect | 12 | Yes | No route, or a route that does not exist in the method | A route chosen but the reasons restate the score | One route with reasons that follow from the canvases and are written for the requester | As level 3, and an honest decline or defer is handled as a service, not a rejection |
| DS8 | Lane chosen in proportion to blast radius, and stated | 6 | No | No lane | Lane named but not justified | Lane named with the blast radius reasoning | As level 3, and the lane changed during the case for a stated reason |
| DS9 | Writing: two pages, plain, a colleague could act on it | 8 | No | Over length, or unreadable | Readable but the reader has to work | Clear, within length, decisions visible | As level 3, and the analysis could go to the sponsor unchanged |
| Total | 100 |
Rubric B: Certified AI Value Orchestrator case analysis
The case: a real or realistic AI proposal at or past the Commit gate, taken through Design, Adopt and Realise to a Value Realisation Review, planned or actual. Canvases required: 09, 11, 14, 15, 16, 17, 18, 19.
| Code | Criterion | Weight | Essential | Level 1 | Level 2 | Level 3 | Level 4 |
|---|---|---|---|---|---|---|---|
| VO1 | Benefits are typed and ranged, with honest alternatives and a pre-mortem | 15 | Yes | Untyped or single-point benefits; no alternatives | Typed or ranged but not both; alternatives are straw men | Typed and ranged benefits; at least one alternative that could have won; a pre-mortem with a named failure | As level 3, and the pre-mortem changed something in the proposal |
| VO2 | The autonomy ladder: the lowest rung that solves the task is chosen and defended, and the four agent questions are answered | 15 | Yes | Autonomy level not considered, or the supplier's description accepted as the rung | A rung chosen without showing lower rungs were considered | Lowest rung that solves the task, with the rungs below it shown not to, and the four agent questions answered | As level 3, and the analysis translates the supplier's words into what the system actually does |
| VO3 | Ownership after launch: business owner and product owner named, and the review booked at commit | 12 | Yes | No owner | Owner named but no review date, or a review with no owner | Both owners named with what each answers for; review booked with a date and a reason for the date | As level 3, and the review date departs from month nine for a reason tied to the value curve |
| VO4 | Stakeholder impact, including the objection that is right | 10 | No | No stakeholder view | Stakeholders listed without impacts | Impacts stated per stakeholder and at least one objection accepted as right with a response | As level 3, and the accepted objection changed the design |
| VO5 | Adoption measured from the work, in terms a supervisor could observe | 12 | Yes | Deployment treated as use | Adoption measures that are usage counts only | Measures drawn from the work itself that a supervisor could observe, with a baseline | As level 3, and a measure that would show non-adoption early |
| VO6 | Workforce impact described task by task, with where released time goes | 10 | No | No workforce treatment, or only headcount | Impact described at role level | Task by task, with released time allocated and stated | As level 3, and a task where the honest answer is that the time is not released |
| VO7 | Benefits realisation: evidence ledger, measurement owners, baselines, corrective triggers | 12 | Yes | No ledger | Ledger without owners or baselines | Ledger with owner, baseline and a corrective trigger per benefit | As level 3, and a trigger that has already fired or would plausibly fire, with the corrective action |
| VO8 | The Value Realisation Review puts measured value beside the committed number | 8 | No | No review | Review planned without the committed number | Review plan or record with measured beside committed, and the three value conversations located | As level 3, and a candid statement of what the review cannot yet know |
| VO9 | Writing: two pages, plain, a colleague could act on it | 6 | No | Over length, or unreadable | Readable but the reader has to work | Clear, within length, decisions visible | As level 3, and the analysis could go to the board or executive unchanged |
| Total | 100 |
Rubric C: Master AI Value Partner portfolio (from 2027)
The item: one real AI request taken end to end, from the front door to value evidenced at the review. Realistic cases are not accepted. Evidence is required at each gate: Request, Shape, Rank, Commit, Build, Review. The written portfolio and the oral defence are scored separately; both must pass.
C1: Written portfolio
| Code | Criterion | Weight | Essential | Level 1 | Level 2 | Level 3 | Level 4 |
|---|---|---|---|---|---|---|---|
| MP1 | The line is complete: every gate has a dated record of what was decided, by whom, and on what evidence | 20 | Yes | Gates missing or described rather than evidenced | Most gates evidenced; some reconstructed after the fact | Every gate evidenced with a contemporaneous record | As level 3, and a gate where the record shows the decision was hard |
| MP2 | Front door quality (the Strategist criteria DS1, DS2, DS3, DS7 applied to the real item) | 15 | Yes | Fails two or more of the four | Meets two of the four | Meets all four | Meets all four with judgement shown |
| MP3 | Commit and after quality (the Orchestrator criteria VO1, VO2, VO3, VO5, VO7 applied to the real item) | 15 | Yes | Fails two or more of the five | Meets three of the five | Meets all five | Meets all five with judgement shown |
| MP4 | Realised value: the review puts measured value beside the committed number, with the ledger evidence, and the gap is explained honestly | 20 | Yes | No review has happened, or the number is asserted | Review happened; evidence thin or gap unexplained | Review record with ledger evidence and an honest account of the gap either way | As level 3, and a corrective trigger fired and was acted on |
| MP5 | Governance orchestrated, not replaced: the item shows the existing controls it routed to (business case, assurance, privacy, transparency, service standards or equivalents) and what each returned | 10 | No | Governance bypassed or not mentioned | Controls named but not what they returned | Each control named with what it returned and how that changed the item | As level 3, and a control that said no, handled |
| MP6 | Reflection: what the candidate would do differently, and what the method got wrong | 10 | No | No reflection or only praise for the method | Reflection generic | Specific reflection on the item and at least one criticism of the method | As level 3, and a proposed change to a canvas or gate with a reason |
| MP7 | Evidence handling: anonymisation, consent, and nothing the candidate is not entitled to share | 10 | Yes | Identifiable or unauthorised material | Anonymisation incomplete | Clean, with a statement of what was withheld and why | As level 3, and the withholding is itself evidence of good judgement |
| Total | 100 |
C2: Oral defence (forty-five minutes)
| Code | Criterion | Weight | Essential | Level 1 | Level 2 | Level 3 | Level 4 |
|---|---|---|---|---|---|---|---|
| OD1 | Owns the work: answers about any gate without notes and knows why each decision was taken | 35 | Yes | Cannot explain decisions in the portfolio | Explains most; some decisions attributed to others without understanding | Explains every decision as their own or as one they understood and accepted | As level 3, and disagrees with a past decision of their own for a reason |
| OD2 | Reasons under challenge: when an assessor offers a counter-view, engages with it rather than defending | 30 | Yes | Defensive or evasive | Concedes everything or nothing | Weighs the counter-view and either changes position or holds it with a reason | As level 3, and offers a better counter-view than the assessor did |
| OD3 | Method literacy: locates the item in the six stages, six gates, lanes, ladder and review without prompting | 20 | No | Cannot | Partly | Fluent | Fluent and critical |
| OD4 | Candour about the gap between committed and realised | 15 | No | Gap hidden or spun | Gap acknowledged, causes vague | Gap stated with causes and what was learned | As level 3, and the learning has been applied elsewhere |
| Total | 100 |
Where an oral defence is not practical, the same criteria are scored on written answers to the assessors' questions, returned within fourteen days, and the assessors record why the defence was written.
Moderation tolerance
Two assessors are taken to agree when their totals differ by ten points or fewer and their outcomes match. Anything else goes to moderation. Initial calibration parameter.
Calibration and change log
- The Director of Studies reviews every weight and threshold after cohorts one, two and three, using the score distributions, the moderation notes and the resubmission rate, and records any change here with a date. Changes apply from the next cohort.
| Version | Date | Change |
|---|---|---|
| v0.1 | (date of approval) | First draft, all parameters set by judgement |
Director of Studies, GOVBRM Academy
