Analysis Guide

AISAY doesn’t just extract data — it can evaluate, assess, and provide intelligent commentary on your documents. This guide shows you how to use the web portal for document analysis tasks.

Prerequisite: Read the Quick Start Guide first.

What Is Analysis?

Analysis goes beyond extraction. Instead of pulling out facts that already exist in the document, you’re asking the AI to think about the content and give you an evaluation.

Extraction	Analysis
“What is the company’s revenue?”	“Is the company financially healthy based on this report?”
“List all board members”	“Evaluate the diversity of the board”
“Extract the product specifications”	“Does this product meet our requirements?”

How to Structure Analysis Prompts

Good analysis prompts have four components:

Role/Context — Tell the AI who it is and what it’s working with
Goal — What you want to learn or evaluate
What to look for — Specific signals, sections, or criteria
Output format — How to structure the response

Example Prompt Structure

You are an expert analyst looking at company's sustainability reports.

Goal: How does the organisation engage local communities?
Find: donations/grants, volunteering, community programs, foundations/CSR funds, NGO partnerships.
Write:
What they did + who benefited + where (location)
Any numbers (amounts, hours, beneficiaries)
If none: Not found in report.

Tip: You only need to set the role/context in the first field. The AI carries this context across all fields in the same extraction run.

Example 1: Analysing a Long Report

Scenario: You have 20 sustainability reports and want to quickly extract insights from each one.

Field 1 — Community Involvement (Text):

You are an expert analyst looking at company's sustainability reports.

Goal: How does the organisation engage local communities (donations, gifts, community services, foundations)?
Find: donations/grants, volunteering, community programs, foundations/CSR funds, NGO partnerships.
Write:
What they did + who benefited + where (location)
Any numbers (amounts, hours, beneficiaries)
If none: Not found in report.

Field 2 — Auditor Profile (Text):

Goal: Who audited or assured the sustainability report (or ESG data)?
Find: "Assurance statement", "Independent assurance", "External assurance", "Verification".
Write:
Provider name (firm)
Scope (what was assured) + level (limited/reasonable) if stated
If none: Not found in report.

Field 3 — People Management (Text):

Goal: People/workforce strategies described as sustainability/ESG.
Find: training/upskilling, DEI, wellbeing, H&S, engagement/retention, human rights/labour practices.
Write:
Strategy → what it is (1 line) + any KPI/target
If none: Not found in report.

Field 4 — Greenhouse Gas Emissions (Text):

Goal: Total GHG emissions reported (tCO₂e) and key breakdowns.
Find: total, Scope 1/2/3, reporting year, methodology/standard (if stated).
Write:
Total + unit + period
Scope breakdown (if available)
Notes (boundary/method) only if stated
If none: Not found in report.

Field 5 — Areas for Improvement (Text):

Goal: Gaps/weaknesses in the sustainability report (based on what's missing/unclear).
Look for: missing Scope 3, no targets/baselines, unclear methodology, weak metrics, no year-on-year trend, vague claims.
Write:
3–6 bullets: issue + why it matters (short)

Field 6 — Quick Wins (Text):

Goal: Practical improvements you can implement in 3-6 months, inspired by common best practices.
Write 5–8 quick wins. For each:
Action (1 line)
First step (1 line)
Simple KPI (1 line)
Avoid claiming "Company does X" unless the report explicitly says so.

Field 7 — Overall Rating (Text):

Goal: Rate the sustainability report quality (10 = best).
Write:
Rating: X/10
3 strengths (with evidence)
3 weaknesses (with evidence or "not disclosed")
3 fixes to raise score by 1–2 points
Use these criteria briefly: clarity, completeness, credibility, comparability, usefulness.

Tip: Upload all 20 reports at once. AISAY will process each one with the same fields, giving you a consistent Excel spreadsheet for comparison.

Example 2: Evaluating Vendor Submissions

Scenario: Companies submit product brochures for a grant claim. You need to check if the equipment meets your specifications.

Field 1 — Vendor Name (Text):

Return only the name of the vendor or manufacturer who makes this equipment. For example, Panasonic.

Field 2 — Specification Compliance (Text):

Compare the required product specifications below against the provided document specifications. Determine if ALL requirements are explicitly met based solely on information clearly stated in the document.

REQUIRED SPECIFICATIONS:
1. Bowl capacity: Minimum 20 liters
2. For Planetary Mixer: Minimum 3 speed options
3. For Spiral Mixer: Minimum 2 speed options
4. Speed range: 100-500 rpm
5. Included accessories: Bowl plus minimum 2 of 3 tools

ASSESSMENT RULES:
1. Each specification must be explicitly stated in the document
2. Do not make assumptions or infer details
3. Partial matches or ambiguous descriptions = Not met
4. Missing information = Not met

RESPONSE FORMAT:
Specification [number]: [Met/Not met]
- Reason: [Quote relevant text from document OR state "No explicit information provided"]

FINAL VERDICT:
- If ALL specifications met: SATISFIED
- If ANY specification not met: UNSATISFIED
- Include explanation for each unmet specification

This prompt demonstrates the four components clearly: task description, context (the specifications), instructions (assessment rules), and expected output format.

Example 3: Analysing Visual Content

AISAY can also analyse image-heavy documents like advertisements, posters, and creative materials.

Tip: Use the Local model for visual/image-heavy content — it’s better at analysing design elements.

Scenario: Evaluate a government campaign advertisement.

Field 1 — Campaign Objective (Text):

What is the primary goal of this advertisement or campaign? Identify the call-to-action and the behaviour change it's trying to drive.

Field 2 — Target Audience (Text):

Based on the visual design, language, and content, who is this advertisement designed for? Consider age group, demographics, and context.

Field 3 — Messaging Effectiveness (Text):

Evaluate whether the call-to-action is clear and compelling. Is the key message easy to understand within 5 seconds? Are there any confusing elements?

Field 4 — Suggestions for Improvement (Text):

Provide 3-5 constructive suggestions to improve this advertisement. Consider clarity, visual hierarchy, accessibility, and persuasiveness.

Tips for Better Analysis

Give the AI a role — “You are an expert analyst” or “You are a compliance reviewer” helps set the right tone and depth
Be specific about what to look for — List the exact signals, sections, or keywords
Specify the output format — Tell the AI exactly how to structure its response (bullets, ratings, tables)
Include “if not found” instructions — Tell the AI what to say when information is missing, to avoid hallucinated content
Use analysis alongside extraction — Combine factual extraction fields with evaluative analysis fields in the same run
Keep humans in the loop — For complex judgement calls, use AISAY to summarise and analyse multiple dimensions, then make the final decision yourself