May 2026 · Design
Methodology in action — at smaller scale
Stage 3 of a series on the design choices behind CoNoggin. A charity director with a three-person team, four weeks, and a sentence describing what she wants.
The question
Can a nine-parameter pass produce a credible intervention when the team is three people, the timeframe is four weeks, and the author would rather describe her problem in plain English than fill in a structured form?
Inputs
Scenario
A UK registered charity. The Director of Development wants her three-person fundraising team to write proposals that arrive submission-ready, so she can stop editing every draft and the team can apply for more grants. Four weeks. No formal L&D apparatus, no training budget, no spare time.
The room
What's already true:
- UK registered charity, ~30–50 staff, mid-size
- Fundraising team: 3 people. Mixed seniority — Hannah (senior fundraiser, ~4 yrs), Michael (grants officer, ~1 yr), Sarah (new starter, 3 months in)
- ~30 grant applications/year; director averages 6–8 hrs/week editing drafts
- Strategic posture: growth-stage; expanding programmes; capacity-constrained on grants throughput
- Cultural priors: mission-driven, lean, generalist, time-poor, no formal L&D function
The goal — in her own words
The director types into the goal-author surface what's actually on her mind:
I want my team to learn to write better proposals so we can apply for more grants and I do less editing each time.
The system proposes — using the org context it already has
The system reads her sentence against the org config (3-person fundraising team rostered, recent application history, the director-as-final-editor pattern), produces a structured draft she can review and adjust, and shows its reasoning as it does. She edits a phrasing or two, adds something, and commits.
The change (and how we'll see it)
The fundraising team writes proposals that arrive submission-ready with light editing only. We'll see it in: drafts coming back with fewer rounds of revision; team applying for more grants because writing isn't a bottleneck; my editing time per proposal dropping from ~2 hours to ~30 minutes.
For whom
The fundraising team — Hannah (senior fundraiser, ~4 yrs), Michael (grants officer, ~1 yr), Sarah (new starter, 3 months in).
The obstacle
Time, on me. The team don't get consistent feedback because I'm in too many places. They need a way to learn the pattern without me having to mark up every draft.
The nine-parameter pass
| Question | Read | Confidence |
|---|---|---|
| What kind of problem is this? (Cynefin) | Complicated. Proposal writing has known good patterns. The director herself holds the expertise; AI doesn't have to discover it. Analyse-then-prescribe is appropriate. | High |
| What's the change? | Submission-ready drafts; director's editing time drops. | High |
| What's the behaviour shift? | Team self-edits and peer-reviews drafts before they reach the director. | High |
| Where's the audience starting from? | Three distinct experience levels — not bimodal, just different points along a proficiency ladder. | High |
| Why now? | Director's editing burden is unsustainable; team capacity constrains grant-application throughput. | High |
| How will we know? | Lead: per-draft “first-pass-ready” flag from director (binary). Lagging: editing time per proposal; total grants applied. | High |
| By when? | 4 weeks. | High |
| What's the lead measure? | First-pass-ready rate — drafts the director can sign off with light edits only. | High |
| What's the obstacle? | Director's time is the bottleneck. Intervention has to use her time efficiently, not multiply it. | High |
Outputs
Methodology pick
The load-bearing methodology pick:
Sense-making
Cynefin
Classifies proposal-writing as Complicated — known good patterns exist; the director holds the expertise. Permits expert-led teaching, no probe-sense-respond posture needed.
Stance · expert-led teaching
Discipline
Action Mapping
Strip everything to what does a submission-ready proposal look like?Move directly into real-work practice — no theory courses on grant-writing.
Discipline · keeps it on-the-job
Evaluation
Reverse Kirkpatrick
Designed backwards from L4 (director's editing time per proposal ≤45 min). The lead measure is the per-draft sign-off classification.
Measurement architecture · backwards from L4
Also in the reasoning, shaping the design more lightly: 70-20-10 (the work IS real grant applications), apprenticeship-pattern coaching (one-off direct teaching from the director), Self-Determination Theory (autonomy is preserved — the team owns the drafts). The AI carries these alongside, helping shape the intervention.
Activity composition (four steps over four weeks)
Goal · Fundraising team · 4 weeks
Submission-ready proposals with light editing only
- Video + checklist — What good looks like
- Real grant application as live work
- 1:1 coaching call with director
- Peer review trio
The director's time investment:
- ~1 hour upfront — recording the video and writing the 1-page checklist
- ~1.5 hours in week 2 — three 30-min coaching calls
- ~30 min/week ongoing — reviewing finals (instead of editing drafts)
~2.5 hours of concentrated time upfront, then she's largely out of the loop.The peer review trio replaces what would have been ongoing director feedback. That's the structural move that makes the intervention work at this scale.
Measurement
| # | Measure | Success |
|---|---|---|
| 1 | Watched + checklist downloaded | All 3 within 3 days |
| 2 | Each team member produces ≥1 grant application via this process | By week 4, all 3 have at least one submission |
| 3 | Coaching calls held | 3/3 in week 2 |
| 4 | Peer review sessions held | 3 sessions across weeks 2–4 |
| — | First-pass-ready rate (lead measure) | By week 4, ≥2 of 3 drafts come back as light-edits (≤30 min) |
Intervention judgment (Reverse Kirkpatrick)
- L4 — Results. Director's editing time per proposal ≤45 min averaged across the 5 proposals after week 4. Grant application throughput up ≥30% in the following quarter. (Lagging.)
- L3 — Behaviour. Team uses the checklist before submitting; visible in revision history.
- L2 — Learning. Team can articulate what makes a strong proposal (informal week-4 conversation).
- L1 — Reaction. Team reports the rhythm is sustainable; director reports her editing burden has dropped.
Successful if L4 holds. At this scale, the L4 outcome is observable directly and quickly — the chain from lead measure to lagging measure is short enough that the intervention mostly judges itself.
What worked
- The plain-English goal carried. The director typed one sentence; with adequate org context, the AI did the heavy lifting and proposed a structured goal she could review and commit.
- Her time was respected. ~2.5 hours upfront, then largely out. The peer review trio replaced what would have been ongoing feedback.
- The L4 outcome was observable directly. Her own editing time. No long causal chain to track.
What didn't, or wobbled
- The director needs to actually record the video and write the checklist. The intervention is light, but it asks her for ~1 hour of careful upfront work. If she puts it off, the rest collapses. The system should treat the upfront step as the gate, not an optional starter.
- Three people is small enough that one absence breaks the peer review trio. If Sarah goes on leave in week 3, the rotation cracks. The framework should flag cohort-size dependencies as a confidence note.
Conclusions (for now)
The output is plausible. The methodology worked. If the director wanted something different — a tighter cap on her own time, the team using an existing framework rather than one derived from her input — she can push back and CoNoggin will edit accordingly.