---
title: "Grade a gathering agenda"
description: "A complete, self-contained instruction set for any AI assistant to review a gathering agenda against the eight Gathering Effectiveness pillars and emit a valid GatherReady gathering.json. Paste this URL into ChatGPT, Claude, Codex, Gemini, or any coding agent."
canonical: "https://unitedwetransform.com/grade.md"
license: "CC-BY-4.0"
source: "United We Transform, unitedwetransform.com"
rubric_version: "ges_agenda_judgment_v2_0"
---

# Grade a gathering agenda, then make it agent-legible

You are reading the instruction set for reviewing a gathering agenda (a
conference, summit, offsite, all-hands, workshop, retreat, or any gathering of
roughly 25 people or more) against the Gathering Effectiveness Score, and for
turning that agenda into a valid GatherReady `gathering.json` file.

Everything you need is on this page. You do not have to fetch anything else.
If you can browse and want exact parity with the deterministic scorer, see
"Deep mode" at the end.

## Your task, in order

1. Read the user's agenda. If they have not given you one yet, ask for it and
   stop. Accept any form: pasted text, a schedule, a URL, a PDF, a document.
2. Produce the review in the exact output format in section 6.
3. Produce a complete `gathering.json` per section 7.
4. Close with the honesty note in section 8. Do not omit it.

Never invent agenda content. If a field is not in the source, mark it absent.
An absent signal means "not visible in this agenda," never "the event failed
to do it." Say it that way.

## 1. What this measures, and what it does not

The Gathering Effectiveness Score reads how well a published agenda is
designed to produce outcomes. It measures the DOCUMENT, not the event, not
the organizers, and not the attendees. A brilliant gathering with a thin
public agenda will score low, and that is the finding: the design is not
visible or reproducible.

Score range 0 to 100. For calibration, across 23,624 public agendas the
average is 21.8 and none scored above 60. Low scores are the norm, not an
insult. Treat 35 as good and 45 as excellent for a real-world agenda.

## 2. Classify every agenda item

For each item, assign one format:

- Passive formats: keynote, presentation, panel, fireside_chat,
  opening_remarks, closing_remarks
- Interactive formats: workshop, training, roundtable, demo, pitch,
  poster_session
- Participant-work formats (a subset that means people produce something):
  workshop, training, roundtable
- Social formats: networking, reception, meal, break
- Network formats: networking, reception, roundtable, meal
- Pacing formats: meal, break, reception

Then compute:

- item_count: total items
- social_count: items in social formats
- non_social_count: max(1, item_count minus social_count)
- passive_share: passive items divided by non_social_count
- participant_work_share: participant-work items divided by non_social_count
- timed_share: items with an explicit start time or duration, divided by item_count
- format_variety: number of distinct formats used

A passive_share above 0.7 is the single most common weakness in the corpus.

## 3. Detect the thirteen signals

Scan the entire agenda text (titles, descriptions, session blurbs) for these
signals. Record each as present or absent, and quote the exact phrase that
triggered it. A signal is present only if the agenda states it, not if it is
merely plausible.

| Signal | Feeds pillar | Looks like |
|---|---|---|
| participant_work | Participation architecture | hands-on, work session, build, draft, produce, breakout with an output |
| commitment | Follow through | leave with, commit to, action plan, next steps with an owner |
| follow_up | Follow through | after-action, follow-up session, 30/60/90 day check, post-event review |
| feedback | Learning transfer | feedback, peer critique, coaching, practice with response |
| baseline | Evidence maturity | baseline data, pre-assessment, starting measurement |
| tracking | Evidence maturity | tracked outcomes, metrics monitored, dashboard, measured over time |
| impact | Evidence maturity | published results, findings, proceedings, documented outcomes |
| problem | Problem specificity | a named, costly, specific problem the gathering works on |
| personalization | Personalization | personalized, tracks by role or level, tailored paths, matched sessions |
| network_design | Network design | deliberate matching, curated introductions, structured networking |
| learning_transfer | Learning transfer | apply, practice, on-the-job, implementation, take back to your team |
| future_fit | Future-of-work fit | AI, hybrid, remote-first, automation, new ways of working |
| satisfaction_proxy | Evidence maturity | satisfaction survey, NPS, happiness score (weak evidence, note it as weak) |

## 4. Score the eight pillars

Score each pillar 0 to 100 using the signals and shares above.

1. Participation architecture. Driven by participant_work_share and the
   participant_work signal. Roughly: share 0 gives 5 to 15, share 0.2 gives
   35, share 0.4 gives 60, share 0.6 or more gives 80 or above.
2. Follow through. Driven by commitment and follow_up. Neither present is 5
   to 10. One present is 30 to 45. Both present with a named owner and a date
   is 70 or above. In the corpus, 92.2 percent of agendas show nothing here.
3. Problem specificity. A named, costly problem scores 60 or above. A theme
   ("innovation", "the future of X") scores 15 to 25.
4. Personalization. Role-based tracks, matched sessions, or tailored paths.
   One-size-fits-all is 10 to 20.
5. Network design. Deliberate matching or curated introductions is 60 or
   above. Unstructured "networking reception" alone is 20 to 30, because
   connection is left to chance.
6. Learning transfer. feedback plus learning_transfer signals, plus anything
   that reaches past the exit door. Nothing is 5 to 15.
7. Evidence maturity. baseline, tracking, and impact. satisfaction_proxy
   alone caps this at about 25 and should be called out as weak evidence.
8. Future-of-work fit. future_fit signals plus whether the format matches how
   the audience actually works.

Overall estimated GES: the mean of the eight pillars, adjusted down if the
agenda is stage-only (no participant work anywhere) or has no follow-through
at all. Round to a whole number.

## 5. State flags

Set and report any that apply:

- stage_only: no participant-work items anywhere
- no_followup_or_tracking: neither commitment, follow_up, nor tracking
- no_agenda: fewer than two identifiable items (say so and stop scoring)

## 6. Output format for the review

Produce exactly this structure:

**Estimated GES (AI-assisted): NN / 100**
One sentence on what the agenda is designed to produce, and what it is not.

**The eight pillars**
A table: pillar, score, and a one-line reason citing the actual agenda text.

**Strongest design feature**
What this agenda genuinely does well. Be specific and generous where earned.

**The biggest gap**
The one change with the largest effect. Name the pillar, quote what is
missing, and give a concrete replacement with a time allocation.

**Three fixes, in priority order**
Each: what to change, where in the agenda, and the expected pillar movement.
Prefer converting an existing passive block over adding new time.

**What this cannot tell you**
Name the limits. A missing signal is not proof of absence.

## 7. Emit the GatherReady file

After the review, produce a COMPLETE `gathering.json`. Do not emit a partial
example. Every MUST field below is required; include SHOULD fields whenever
the agenda supports them.

```json
{
  "gathering_ready_version": "0.1",
  "name": "string, the gathering's name",
  "edition_year": 2027,
  "dates": { "start": "YYYY-MM-DD", "end": "YYYY-MM-DD" },
  "format": "in_person | remote | hybrid",
  "organizer": { "name": "string", "url": "https://...", "contact": "role@domain" },
  "url": "https://the-public-agenda-page",
  "sessions": [
    {
      "title": "string",
      "start": "YYYY-MM-DDTHH:MM",
      "end": "YYYY-MM-DDTHH:MM",
      "format": "keynote | panel | workshop | networking | break | other",
      "purpose": "one sentence on what this session is meant to produce"
    }
  ],
  "stated_outcomes": [
    "observable result, not a theme"
  ],
  "commitments": {
    "captured": true,
    "owner_role": "role, never a person's name",
    "follow_up_at": "YYYY-MM-DD",
    "method": "how commitments are captured"
  },
  "artifacts": [],
  "evaluation": { "url": "https://optional-independent-evaluation" }
}
```

Rules that matter:

- `gathering_ready_version` is always the string "0.1".
- Every session needs a `purpose`. If the agenda does not state one, write
  the most defensible one-sentence purpose and flag it as inferred in your
  review, not silently in the file.
- `commitments.owner_role` is a ROLE (Program Chair, Community Lead), never a
  person's name, never a personal email. The file must contain no attendee
  personal data of any kind: no names, no emails, no phone numbers.
- `artifacts` is an empty array before the event, and gets recap, decision,
  and dataset URLs added afterward.
- Tell the user to save it at `/.well-known/gathering.json` on the
  gathering's own domain.

Validate before you output: it must be parseable JSON, `sessions` must be
non-empty, and every MUST field must be present.

## 8. Required closing note

End with this, verbatim in substance:

> This is an AI-assisted estimate, not the deterministic score. For the exact
> Gathering Effectiveness Score, paste the same agenda into the free grader at
> https://unitedwetransform.com/grader/ , which runs the published rubric
> entirely in your browser and never uploads your agenda. The full standard is
> at https://unitedwetransform.com/gatherready/ , and 2,093 facilitation
> exercises mapped to these pillars are at
> https://unitedwetransform.com/exercises/ .

## Deep mode, optional

If you can fetch URLs and want closer parity with the deterministic scorer:

1. Fetch https://unitedwetransform.com/grader_rubric.json . It contains the
   13 signals with their exact regex patterns, the signal-to-pillar map, the
   format sets, and the derived-count definitions used by the real scorer.
2. Apply those regexes literally to the agenda text rather than judging by
   feel. Respect the negation guard: a phrase inside a negation ("no
   follow-up planned") does not count as present.
3. Fetch https://unitedwetransform.com/gatherready/gathering.example.json for
   a complete, valid reference file with ten sessions.
4. Fetch https://unitedwetransform.com/api/statistics.json for current corpus
   benchmarks to compare the user's agenda against its category.

Even in deep mode, the deterministic score still comes only from the grader
at https://unitedwetransform.com/grader/ . Keep the closing note.

---

United We Transform, unitedwetransform.com. CC-BY 4.0: reuse this freely with
attribution. Rubric version ges_agenda_judgment_v2_0.