AI Requirements Analysis Tool

AI & Automation · Komarva Marketing
Project Overview
Manual requirements analysis was one of the slowest and least consistent parts of delivery. Analysts spent hours turning meeting notes into requirements, user stories, and risk registers. Quality varied between analysts, handover to delivery teams lagged, and risks and assumptions often surfaced late in the SDLC.
My Contributions
Identified the opportunity through process analysis, built and presented a proof of concept, then drove adoption across the organization. Defined governance standards for AI output, designed the end-to-end architecture, led prompt engineering to match BA best practice, and delivered a stakeholder-ready interface. Stakeholders included the CIO, Head of Delivery, Product Owners, SMEs, and the BI team.
60+ min
saved per set of meeting notes
Up to 40%
faster discovery-to-backlog readiness
~25%
fewer analyst rework cycles
2×
earlier risk and assumption identification
$80K–$120K
projected annual savings per team
How it works

From raw notes to backlog-ready requirements

1
Discovery notes
The analyst submits raw stakeholder meeting notes.
→
2
Governed prompts
A standardized prompt framework analyzes the notes.
→
3
JSON schema
A structured schema enforces one consistent output.
→
4
Analyst review
A human validates and approves every result.
→
5
Backlog-ready
Requirements, user stories, acceptance criteria, and risks.
Komarva Marketing
Senior IT Business Analyst
Jun 2025 – Present

How it works: an analyst submits discovery notes, a governed prompt framework analyzes them, and a structured JSON schema enforces one consistent output: functional and non-functional requirements, user stories, Given/When/Then acceptance criteria, and explicit risks and assumptions. An analyst reviews and validates every result, and analysis and approval are kept as separate steps.

Results: over 60 minutes saved per instance of note conversion, discovery-to-backlog readiness up to 40% faster, analyst rework cycles down about 25%, a 100% consistent output format, and risks and assumptions identified twice as early. Projected cost avoidance: $80K–$120K per delivery team per year.

Tools: Python, OpenAI API, Streamlit, Pandas, Excel reporting, and secure credential management.

Lesson learned: AI delivers the most value inside a governance framework. Clear prompts, enforced output standards, and a human-in-the-loop review kept stakeholder trust while unlocking real efficiency gains.

Impact

Improvement by metric

Output format consistency
100%
Faster delivery readiness
Up to 40%
Fewer analyst rework cycles
~25%
Percentage improvement compared with the manual requirements process.
Before and after

What changed for the team

Before
Hours spent turning each set of meeting notes into documentation
Format and quality varied from analyst to analyst
Risks and assumptions surfaced late in the SDLC
Slow handover to delivery teams
After
60+ minutes saved per set of meeting notes
One standardized output, 100% consistent
Risks flagged at discovery, twice as early
Backlog-ready up to 40% faster
Sample output

What the tool produces

Requirements output
User story
As a campaign manager, I want to see campaign performance by channel, so that I can reallocate budget quickly.
Acceptance criteria
Given campaign data has loaded for the selected period,
When I filter by channel,
Then I see spend, conversions, and cost per conversion for that channel.
Non-functional requirement
The dashboard loads within 3 seconds for up to 12 months of data.
Risk
Channel data comes from three sources with different refresh schedules.
Medium severity
Illustrative example only. It does not contain client data.
Tech stack

Built with

Python
OpenAI API
Streamlit
Pandas
JSON schema
Excel reporting
Secure credential management