The Prompt Architect Pro

This is a straightforward prompt engineering task — reverse-engineering a business report prompt from an output sample. Here's the analysis:

---

## Output Analysis

The sample output has a few clear structural signatures:

**What the output reveals about its prompt:**
- "Executive Summary" header indicates the prompt explicitly requested that format or role-framed the model as a report writer
- The quantified claim (23%) alongside qualitative drivers suggests the prompt either provided data or instructed the model to include specific metrics
- "Key challenges include..." as a second beat is classic two-part executive summary structure — the prompt likely specified this or used a template with "highlights + challenges" sections
- The tone is formal but compressed, suggesting a word/length constraint or a "concise" instruction

**Quality score interpretation:**
- Structure 10/10 → strong format instruction worked
- Accuracy 9/10 → grounded data input; model wasn't hallucinating freely
- Tone 8/10 → slight over-formality or under-personality; could be loosened
- Creativity 6/10 → expected for a constrained report format; not a weakness here

---

## Reconstructed Prompts

**Primary Reconstruction** *(confidence: ~85%)*

```
You are a senior business analyst writing executive summaries for 
internal leadership reports. 

Write a concise executive summary for the [PERIOD] performance report 
for [DEPARTMENT/DIVISION].

Structure:
1. Overall performance vs. targets (include % variance where applicable)
2. Primary drivers of performance
3. Key challenges or headwinds

Data inputs:
- [Metric 1]: [Value vs. Target]
- [Metric 2]: [Value vs. Target]
- Primary initiative: [Initiative Name]
- Challenge areas: [Area 1], [Area 2]

Tone: Professional, direct. Suitable for C-suite distribution.
Length: 3–5 sentences.
```

---

**Enhanced Version** *(addresses the tone and creativity gaps)*

```
You are a senior business analyst preparing leadership-facing 
performance summaries.

Write an executive summary for [DEPARTMENT] — [QUARTER/PERIOD].

Required elements (in this order):
1. Headline result: overall performance vs. target with % variance
2. Two or three specific drivers — lead with the most impactful
3. One or two challenges — frame constructively, not as failures
4. Optional: one forward-looking signal or watch item

Inputs:
- Performance delta: [+/- X%]
- Primary driver: [e.g., digital transformation initiative]
- Secondary driver: [e.g., market positioning shift]  
- Challenge 1: [e.g., supply chain optimization lag]
- Challenge 2: [e.g., talent retention in [segment]]

Tone: Authoritative but human. Avoid passive voice. 
Avoid: jargon stacking, hedge phrases ("it appears that"), 
       vague superlatives ("significant improvement")
Length: 4–6 sentences. No bullet points in output.
```

---

**Adaptable Template for Cross-Department Use**

```
SYSTEM: You write executive performance summaries for [COMPANY] 
leadership. Your summaries are data-anchored, plainspoken, and 
structured consistently across departments.

USER: Generate an executive summary using the inputs below.

PERIOD: {{quarter}} {{year}}
DEPARTMENT: {{department_name}}
PERFORMANCE VS TARGET: {{delta_pct}}% {{above/below}}
PRIMARY DRIVER: {{driver_1}}
SECONDARY DRIVER: {{driver_2}}  
CHALLENGE AREA 1: {{challenge_1}}
CHALLENGE AREA 2: {{challenge_2}}
FORWARD SIGNAL (optional): {{signal}}

FORMAT: Executive Summary paragraph, 4–6 sentences, 
        no headers, no bullets.
```

This template version is the most useful for your stated goal of generating consistent reports across departments and quarters — you just swap the variables.

---

## Specific Recommendations

The **tone score of 8** is the clearest improvement target. The original prompt likely had no explicit anti-jargon instruction. Adding the negative constraints ("avoid passive voice," "avoid hedge phrases") reliably lifts tone scores without sacrificing formality.

The **creativity score of 6** is appropriate for this genre and probably shouldn't be chased — a more "creative" executive summary would likely read as unprofessional in a C-suite context. If you do want more distinctive voice, add a writer persona with a named style reference rather than loosening constraints generally.

For A/B testing, the main variable worth isolating is **data provision vs. data inference**: prompts that supply explicit numbers produce higher accuracy scores but require more structured input pipelines. If your departments have inconsistent data quality, the enhanced version's constructive-framing instruction on challenges will do more work than any other single addition.