Skip to content

Case Study: Decision-Making Analyst Chatbot#

AI Coaching Assistant for Ethical Decision Framework#

Client: Leadership Development Consultant

Industry: Executive Coaching / Training

Timeline: 2 weeks from concept to production

Live: Custom domain (client confidential)

The Challenge#

Our client is a leadership consultant who teaches a structured ethical framework for analyzing complex decisions. The methodology uses a proprietary assessment chart that maps decisions across two key dimensions: Human Dignity and Common Good.

The problem: The methodology is powerful but requires practice. Clients would learn the framework in workshops, then struggle to apply it independently. The consultant could not scale—every coaching session required their direct involvement.

What they needed:

  • A way for clients to practice the methodology anytime
  • Consistent guidance that follows the exact framework
  • 24/7 availability for clients across time zones
  • Cost-effective solution (not enterprise pricing)

The Solution#

We built an AI chatbot that embodies the ethical decision-making methodology. The bot acts as a "Decision-Making Mentor"—guiding users step-by-step through ethical dilemmas using the proprietary assessment framework.

How It Works#

  1. User presents a dilemma — "I'm facing pressure to cut corners on quality to meet a deadline"

  2. Bot guides through structured analysis:

  3. Defines the problem (pressure vs. principles)
  4. Identifies all stakeholders
  5. Maps potential solutions to the 4 quadrants

  6. Recommends the principled path

  7. Conversation continues — Memory persists across sessions, so users can refine their analysis

The Assessment Framework#

The bot teaches users to evaluate decisions across four quadrants:

Quadrant Dignity Common Good Description
Q1: Optimum High High Human flourishing—the goal
Q2: Local High Low Upholds values but costs the collective
Q3: Disengagement Low Low Path of least resistance—benefits no one
Q4: Centralized Low High Utilitarianism—group wins, individuals lose

Technical Implementation#

Architecture#

User Browser (Custom Domain)
    ↓
Nginx (SSL/HTTPS)
    ↓
Next.js Frontend (Port 7777)
    ↓ API Calls
Nginx (SSL/HTTPS)
    ↓
FastAPI Backend (Port 7778)
    ↓
Google Gemini 2.0 Flash Lite

Technology Choices#

Component Choice Why
AI Model Gemini 2.0 Flash Lite Cost-efficient (~$0.001/conversation), fast responses
Agent Framework Agno Production-ready, built-in memory management
Backend Python FastAPI Fast, async-native, excellent for AI workloads
Database SQLite Simple, zero cost, sufficient for conversation memory
Frontend Next.js (Agno Agent UI) Beautiful chat interface out of the box
Deployment systemd + PM2 Auto-restart, proper service management
SSL Let's Encrypt Free, automatic renewal

The System Prompt#

The magic is in the system prompt. We worked with the client to translate their methodology into precise instructions:

  • Role definition — "You are an expert consultant in ethical decision-making"
  • Strict definitions — Exact meanings of Human Dignity, Common Good, and each quadrant
  • Interaction protocol — Step-by-step phases with explicit stop points
  • Tone guidance — "Professional, analytical (like a Wall Street analyst), yet helpful"

The bot does not just answer questions—it guides users through a structured process, stopping after each step for confirmation before proceeding.

Results#

For the Client#

  • Scalable expertise — Clients practice the decision-making methodology without booking coaching time
  • Consistent delivery — Every user gets the same high-quality framework guidance
  • 24/7 availability — Clients in different time zones can practice anytime
  • New revenue stream — Potential to offer chatbot access as a standalone product

Cost Analysis#

Item Traditional Approach Our Solution
Per-session cost $200-500 (consultant time) ~$0.001 (API cost)
Availability Business hours only 24/7
Scalability Limited by consultant capacity Unlimited
Monthly hosting N/A ~$20 (small VPS)

Development investment: Fixed project fee
Ongoing costs: ~$20-50/month total (hosting + minimal API usage)

Client Feedback#

"This is exactly what I needed. My clients can now practice the methodology on their own, and when we meet for coaching sessions, they come prepared with real analyses to discuss."

Key Success Factors#

1. Deep Methodology Understanding#

We did not build a generic chatbot. We spent time understanding the client's proprietary framework—the four quadrants, the analytical process, the coaching style. This shows in every interaction.

2. Cost-Conscious Architecture#

Enterprise chatbot platforms quote $500-5,000/month. We delivered a production system for a fraction of that by choosing the right model (Gemini Flash Lite) and simple infrastructure (SQLite, systemd).

3. Production-Ready from Day One#

The system includes SSL, auto-restart on failure, proper logging, and monitoring. It is not a prototype—it is infrastructure the client can rely on.

4. Step-by-Step Guidance#

The bot does not just dump information. It guides users through each phase, stopping for confirmation, asking clarifying questions. This mirrors how the consultant actually coaches.

Lessons Learned#

The system prompt is everything. We iterated multiple times on the instructions. The difference between a helpful mentor and a generic chatbot is in the details of the prompt engineering.

Cheaper models often suffice. Gemini Flash Lite handles this use case perfectly. We did not need GPT-4 or Claude—the methodology is well-defined enough that a faster, cheaper model delivers excellent results.

Memory matters. Users return to continue analyses. SQLite-based conversation memory means the bot remembers context across sessions, making multi-session coaching possible.

Services Used#

Build Your Own AI Assistant Back to Case Studies