SupplyTact

SupplyTact, as described by its author, is an agentic supply-chain planning system designed for pharmaceutical operations. The project was built during a hackathon and centers around ReplenAgent, a specialized AI agent powered by GPT-5.6 Luna, that investigates risks, runs deterministic simulations, and shows the evidence behind every recommendation.

The author states that SupplyTact is not a general-purpose chatbot but an agent embedded within a structured planning workflow. It integrates with a deterministic engine to ensure numerical accuracy while allowing ReplenAgent to explore scenarios, trace causality, and propose recoveries—without modifying baseline data directly.

Key features include:

• Risk analysis at portfolio and SKU levels
• Deterministic monthly planning engine
• Reversible scenario modeling
• 3D inventory runway visualization
• Evidence-based recommendation system

The author emphasizes that the system is built on real-world supply-chain experience, not abstract AI experimentation. It uses GPT-5.6 Luna for reasoning, tool selection, and explanation, but delegates numerical authority to validated tools.

The single most important open question: Is there sufficient evidence of real-world traction or customer validation to suggest that this system has commercial viability beyond a hackathon prototype?

The description states:

• SupplyTact is an agentic supply-and-inventory planning system.
• It includes ReplenAgent, a specialized agent powered by GPT-5.6 Luna.
• ReplenAgent works within a deterministic planning engine, not inventing inventory calculations or providing unsupported recommendations.
• The system allows planners to ask questions like:
  • “Which product is facing the most serious supply risk, and why?”
  • “Why does Bactrel 1 g stock out in March? Show me the evidence.”
• It supports:
  • Risk ranking
  • Scenario creation and comparison
  • Evidence tracing
  • Deterministic simulation
  • Reversible actions

Inferred from the description: The system is designed for pharmaceutical supply chain planning, particularly around inventory, demand forecasting, safety stock, expiry, quality release, and customer allocation.

Not evidenced:

• Whether SupplyTact has been tested in live environments.
• Whether it supports real-time data integration or API connections.
• Whether ReplenAgent can be extended beyond the described use case.

The author states:

• The system was built from a personal, professional need, not a generic tech idea.
• It is positioned as an AI agent that shows its work—not a chatbot giving advice.
• The goal was to create an agent that works inside the planning process, inspecting structured data and calling analytical tools.

The evolution of the product:

• Started with a panoramic visualization tool called SupplyScape.
• Evolved into Planorama, then SupplyTact.
• The name change reflects a shift from visual tool to planning agent as the core value.

Inferred:

• The positioning is rooted in operational workflow integration, not just AI novelty.
• It aims to reduce time spent on investigation and increase transparency in decision-making.

Not evidenced:

• No claims about market fit or competitive differentiation beyond the author’s personal experience.
• No evidence of prior versions or feedback loops from users outside the author.

The description states:

• The system targets supply-chain professionals, especially those working in pharmaceutical operations.
• It addresses a problem faced by planners who still rely heavily on Excel-based planning.
• The demo uses a fictional pharmaceutical product (Bactrel 1 g) to simulate real-world challenges.

Inferred:

• The ICP likely includes supply chain planners, logistics managers, and procurement specialists in regulated industries like pharma or medical devices.
• It may appeal to organizations with complex inventory and demand planning needs, especially where compliance and traceability are critical.

Not evidenced:

• No stated customer segments beyond the author’s own experience.
• No evidence of actual customers or pilot programs.
• No indication of whether the system supports other industries (e.g., manufacturing, retail).

The description states:

• The project is a hackathon submission, not a commercial product.
• There is no mention of pricing models, licensing, or monetization strategies.

Inferred:

• If commercialized, it might follow a SaaS model for enterprise users in regulated industries.
• It could be priced per user, SKU, or scenario.

Not evidenced:

• No business model, pricing structure, or revenue assumptions.
• No evidence of any sales process or go-to-market strategy.

The description states:

• Built using GPT-5.6 Luna, Codex, and ChatGPT for design and implementation.
• Uses technologies such as:
  • React, TypeScript, Node.js
  • OpenAI APIs (Sol and Luna)
  • SQLite-wasm, Playwright, Vercel, Three.js
  • CSV/XLSX import workflows
  • Browser-local persistence

Inferred:

• The system is a browser-based application, likely using local storage for data.
• It integrates AI tools into a structured planning workflow.
• It supports scenario versioning and reversibility.

Not evidenced:

• No evidence of scalability or cloud infrastructure beyond Vercel hosting.
• No evidence of API integrations, real-time data feeds, or multi-user support.
• No mention of security or compliance features (e.g., HIPAA, GDPR).

The description states:

• This is a hackathon project submitted to the OpenAI 2026 hackathon.
• The author spent significant time in ChatGPT before coding.
• It was built using Codex as an engineering collaborator, not a one-shot prompt.

Inferred:

• The system has maturity in concept and design, but no evidence of deployment or user feedback.
• It is likely at the prototype stage.

Not evidenced:

• No customer data, usage metrics, or adoption rates.
• No evidence of product-market fit or real-world testing.
• No indication of future roadmap or development plans beyond the hackathon.

The description states:

• The author did not begin with a technology-first approach but with a real-world problem.
• It is not positioned as a general-purpose AI assistant for supply chain.
• It contrasts itself with chatbots attached to spreadsheets, emphasizing structured planning and deterministic simulation.

Inferred:

• It competes with traditional Excel-based planning tools or basic ERP systems.
• It may compete with AI-powered planning platforms that lack transparency or deterministic logic.
• It is likely a niche solution for highly regulated, complex supply chains.

Not evidenced:

• No mention of competitors or market analysis.
• No evidence of competitive advantages beyond the described architecture.
• No indication of pricing, distribution, or positioning relative to existing tools.

The description states:

• The system is not a commercial product, but a hackathon submission.
• It uses GPT-5.6 Luna, which may not be suitable for production-level reliability.
• The author emphasizes that the AI agent does not perform inventory arithmetic—this is a strength, but also a limitation.

Red flags:

• No real-world validation or customer feedback.
• Limited scalability and infrastructure (browser-based, local storage).
• Dependence on GPT-5.6 Luna, which may not be available in production environments.
• No evidence of integration with enterprise systems or APIs.
• Highly specialized for pharmaceutical use case, limiting broader applicability.

Not evidenced:

• No risk assessment or mitigation strategies.
• No indication of how the system would scale to larger portfolios or more complex planning needs.

1. What are your plans for transitioning from a hackathon prototype to a commercial product?
2. How do you intend to integrate SupplyTact with existing ERP, WMS, or planning systems?
3. Have you tested the system with actual supply-chain professionals beyond yourself?
4. What is the expected performance and latency of ReplenAgent in real-world scenarios?
5. How does the deterministic engine handle edge cases or data inconsistencies?
6. Are there any plans to support multi-user collaboration or role-based access control?
7. What are the long-term plans for AI model availability, especially if GPT-5.6 Luna is not available?