GOST Participant Enablement — Golden Example
The Benchmark
Client: The original deployment (March 2026) for a mid-market B2B services company Deployment date: March 2026 Location: [client-repo]/drafts/GOST-Tactics-Planner-AI-Tool-Participant.docx (participant document) and [client-repo]/drafts/GOST-Tactics-Planner-AI-Tool-Consultant-Brief.docx (consultant brief) Source prompt (markdown): [client-repo]/drafts/ghost-participant-enablement-prompt-kit.md
What This Example Demonstrates
Prompt Design
The original deployment prompt demonstrates the structural pattern all deployments should follow:
- Opening instruction — tells the LLM what to do (walk me through it step by step, one section at a time, produce a completed planner at the end)
- Company context block — embeds the client's Goal and all Objectives so the participant doesn't need to look anything up
- SMART framework — defines what a good tactic looks like with a bad/good example pair
- Golden example — a fully worked tactic plan (HR department) showing tactic statement, end state, KPIs, and backward plan
- Step-by-step instructions — six numbered steps telling the LLM exactly what to ask and in what order
- Output format template — the exact structure of the completed planner with labeled placeholders for every field
Participant Document Design
The original deployment participant document demonstrates:
- Directions first — what this is, what you'll get, time estimate, what you need, instructions
- "Have more than one tactic?" note — explicitly tells participants to run the prompt once per tactic in a new chat
- Clear copy boundaries — START COPY and END COPY markers so participants know exactly what to paste
- No jargon in the instructions — written for someone who may never have used an AI tool before
Consultant Brief Design
The original deployment consultant brief demonstrates:
- Suggested email template — ready to customize and send
- What's embedded — so the consultant understands what the prompt contains without reading it
- Key operational notes — one tactic per chat, works with any LLM, they can iterate, dates are hardcoded
- What to expect back — the specific output participants will produce
- How to review — a checklist for evaluating submissions
What Was Tested
The original deployment prompt was tested against three simulated department leaders with intentionally different profiles:
| Persona | Department | Objective | Personality | What the test validated |
|---|---|---|---|---|
| Marcus Webb | Sales | Obj 1 (growth-oriented) | Practical, results-oriented, not a strategic thinker | The prompt produces actionable plans for concrete thinkers who talk in actions not frameworks |
| Priya Sharma | IT Operations | Obj 3 (operational efficiency) | Detail-oriented, systematic, tends to over-scope | The prompt successfully narrows over-ambitious scope to one-quarter achievability |
| Derek Coleman | Client Services | Obj 2 (revenue diversification) | Relationship-focused, vague about metrics | The prompt pushes metric-averse participants to get specific on numbers and targets |
All three tests produced complete, SMART-compliant 12-Week Tactics Planners with:
- Tactic statements tied to department strategy and company objective
- End state visions with 3+ measurable KPIs
- Backward-planned calendars with weekly actions, owners, hours, and success criteria
- Capacity analysis, dependencies, and progress trackers
Values to Extract for a New Client Deployment
When building for a new client, extract and replace these values from the golden example:
Must Replace (client-specific content)
| Value | Original Deployment Default | Replace With |
|---|---|---|
| Company Goal | [Client's 3-year goal statement] | Client's actual 3-year goal |
| Objective 1 | [Client's Objective 1] | Client's Obj 1 |
| Objective 2 | [Client's Objective 2] | Client's Obj 2 |
| Objective 3 | [Client's Objective 3] | Client's Obj 3 |
| Objective 4 | [Client's Objective 4] | Client's Obj 4 |
| Number of objectives | 4 | Client's actual count (may be 3, 5, 6, etc.) |
| Quarter dates | Q2 2026: April 1 – June 30 | Client's target quarter |
| Month names in backward plan | April / May / June | Months matching the target quarter |
| End state date | June 30, 2026 | Last day of the target quarter |
| Weekly date ranges in output template | Apr 6-10, Apr 13-17, etc. | Calculated from the target quarter start date |
May Replace (if client has a better example)
| Value | Original Deployment Default | When to Replace |
|---|---|---|
| Golden example (HR department tactic) | Standardize job families and compensation bands... | If the client has a real, completed tactic submission that's high quality, use it instead. The HR example is generic by design and works for any industry. |
| Bad/good tactic examples | "Improve customer service" / "Implement structured client check-in cadence..." | If the client's training included industry-specific examples, substitute them. |
Do Not Replace (structural elements)
| Value | Why It Stays |
|---|---|
| SMART framework definition | Universal — not client-specific |
| Step 1–6 structure and instructions | Tested and validated across multiple persona types |
| Output format template (section names, table columns) | Must match the 12-Week Tactics Planner template — only change if the client uses a different template |
| Push-back instructions ("challenge me if my answers are too vague") | Core to the prompt's effectiveness — removing this produces lower-quality output |
| "Be friendly and conversational" closing | Sets the LLM's tone — participants respond better to warmth |
What the Golden Example Does NOT Provide
- The client's goal or objectives — always use the actual client data
- The client's department strategies — the prompt asks participants to provide their own
- Submission logistics — deadlines, email addresses, and format preferences come from the advisor
- Client-specific terminology — if the client calls objectives "pillars" or tactics "action items," update the prompt language to match
- The client's GOST training content — this kit supplements training, it does not contain or replicate it