Case Study - Agile without Scrum

Taming GenAI

Helped stakeholders move from GenAI hype to reality by defining inputs, logic, and outputs for trustworthy strategic insights.

Context and Background

A finance consulting firm wanted to build an internal, GenAI-powered decision-support tool for enterprise strategy teams. The envisioned product would ingest diverse, unstructured inputs—such as vision and strategy documents, financial statements, and industry outlook reports—and generate a prioritized list of strategic initiatives.

These initiatives would then be mapped on a Strategic Alignment vs Business Impact matrix, with drill-down views estimating cost, effort, timelines, and business impact.

Key constraints shaped the problem:

  • All data had to remain fully internal, with no exposure to external systems
  • Input documents had no fixed format, structure, or volume
  • Output expectations were ambitious, but business logic and parameters were undefined

At the outset, the client perceived GenAI as a near-magical capability that could derive precise strategic insights from loosely defined prompts and documents.

Problem Statement

The core challenge was not technical feasibility, but misaligned expectations.

  • Input data lacked structure, consistency, and defined extraction logic
  • Output parameters (impact, cost, alignment) were not clearly defined
  • There were no equations, heuristics, or business rules to guide analysis
  • The client underestimated the effort required to translate strategy into machine-consumable logic

This created a high risk of:

  • Hallucinated outputs
  • Inconsistent results
  • Low stakeholder confidence in the tool
  • Failure to achieve decision-grade reliability

Product Insight and Direction

The key product insight was that GenAI systems require structure to produce trustworthy outcomes, especially in enterprise decision-making contexts.

Rather than treating GenAI as an end-to-end solution, we reframed the product as:

  • A guided intelligence system
  • Combining GenAI’s reasoning abilities with deterministic business logic
  • Built incrementally to balance speed, accuracy, and explainability

This required shifting the client’s mindset—from “prompt-based magic” to human-in-the-loop intelligence design.

Execution and Implementation

Execution began with education and alignment. I led multiple sessions with client stakeholders to explain, in simple terms, how GenAI processes data, where it excels, and where guardrails are essential.

Key execution steps included:

  • Defining a semi-structured input framework that allowed flexibility while ensuring extractable signals
  • Identifying and documenting key parameters to be extracted from strategy, financial, and industry documents
  • Collaborating with stakeholders to define output metrics and verbal logic to estimate impact, effort, and alignment
  • Breaking delivery into incremental milestones, allowing the client to supply inputs progressively
  • Designing the system to keep GenAI focused on interpretation and synthesis, while calculations and scoring followed explicit rules

This approach minimized hallucinations, improved explainability, and allowed continuous validation with stakeholders.

Stakeholder Management and Change Adoption

Initial resistance was high, as the client had not anticipated the depth of effort required on their side. The breakthrough came by:

  • Demonstrating early prototypes that showed the difference between vague inputs and structured ones
  • Transparently highlighting risks of proceeding without clarity
  • Positioning the additional effort as an investment in reliability, not overhead

The client eventually extended project timelines and committed resources to defining inputs and logic, enabling true co-creation.

Outcomes and Impact

  • Delivered a reliable, internally hosted GenAI product aligned with enterprise data policies
  • Enabled consistent generation and prioritization of strategic initiatives
  • Reduced hallucinations through structured inputs and explicit logic
  • Improved client confidence in AI-driven strategic recommendations
  • Established a repeatable framework for future GenAI initiatives

Key Learnings

  • GenAI products succeed when creativity is paired with structure
  • Educating stakeholders is a core product management responsibility in emerging technologies
  • Incremental delivery reduces risk and builds trust in AI systems
  • Enterprise GenAI requires explainability as much as intelligence