EP2: An Agentic Operating System for Procurement Leadership

The Vision: Agentic Procurement at Executive Speed

Walt’s mandate for EP2 is radical simplicity: executives should be able to run world‑class procurement with nothing more than a third‑grade vocabulary and their voice. No hunting through dashboards, no manual mapping, no stitching together data from scattered systems.

EP2’s agentic architecture is built on three core principles:

1. Context‑aware intelligence – Agents understand procurement data models, business rules, and financial impact out of the box.
2. Autonomous action within guardrails – Agents don’t just answer questions; they initiate workflows, trigger escalations, and route decisions safely.
3. Executive‑grade speed and scale – Insights move from detection to decision in seconds, across categories, suppliers, and global teams.

Instead of yet another analytics layer, EP2 functions as an operating system for procurement leadership, compressing the time from signal to action and institutionalizing best practices in software.

Phase 1: Revenue at Risk Agent

The first production agent focuses on the most urgent question in supply risk: “What revenue is at risk right now, and who is accountable?”

Key capabilities

• Severity assessment by financial impact – Automatically evaluates escalation severity based on revenue exposure, not just operational noise.
• Automated revenue‑at‑risk calculation – Quantifies potential impact in dollars, so leaders can prioritize with precision.
• Instant stakeholder mapping – Retrieves supplier owners, category managers, and accountable executives for each escalation.
• Threshold‑based filtering – Surfaces only those issues that exceed predefined financial or risk thresholds.
• Concise, actionable responses – Delivers short, decision‑ready outputs instead of raw data dumps.

As Walt notes: “The quality engineer doesn’t know the financial consequences. The platform does. And that difference matters—because you only have hours to act.”

Business impact

• Seconds instead of days to identify critical revenue risk.
• Immediate accountability with clear owners and escalation paths.
• Reduced revenue leakage through faster intervention on high‑impact issues.
• Executive‑ready insight without manual analysis, spreadsheets, or ad‑hoc fire drills.

Phase 2: Database Chat Agent

The second agent turns natural language into production‑grade SQL—safely, accurately, and at scale.

Core capabilities

• Intelligent query routing – Directs each question to the right database, schema, or view based on intent and metadata.
• Schema‑aware SQL generation – Uses deep awareness of tables, joins, and constraints to generate valid, performant SQL.
• RAG over 8,000+ validated pairs – Leverages a retrieval‑augmented library of question→SQL pairs to ground responses in proven patterns.
• SOP‑aware reasoning – Incorporates standard operating procedures, business rules, and definitions into query logic.
• Automatic error correction – Detects SQL errors, refines queries, and re‑executes without human intervention.
• Executive‑grade markdown outputs – Returns clean, summarized results formatted for leadership consumption.

To minimize hallucinations and latency, TrueHorizon built an automated agentic training loop:

1. AI generates candidate SQL from a natural‑language question.
2. The query is executed against live data.
3. Results are validated for completeness and correctness.
4. The agent refines the SQL recursively until it passes validation.
5. Successful question→SQL pairs are stored in a vector database for future retrieval.

The result is a continuously improving system that gets more accurate, faster, and more aligned with the business over time.

Phase 3: Company Report Agent

The third agent automates deep company research that previously required analysts, consultants, and weeks of effort.

What it delivers

• Comprehensive company profiles using GPT‑4o and Perplexity, including:
• Financial performance and key metrics
• Leadership and organizational structure
• Competitive landscape
• Business model and strategic positioning
• Automated chart generation via QuickChart for:
• Revenue and margin trends
• Market share views
• Comparative benchmarks
• Visual quality validation – Uses image analysis to ensure charts are legible, accurate, and presentation‑ready.
• Clean, executive‑ready markdown reports – Structured narratives with embedded visuals, ready to drop into board decks or C‑suite briefings.

What once took weeks of manual research and slide building now happens on demand, with consistent structure and quality.

Why This Matters

Procurement today manages $11+ trillion in global spend. Even a 1–3% improvement unlocked by AI agents translates into $110B–$330B in incremental annual value.

EP2 is designed to capture that upside by:

• Turning fragmented procurement data into real‑time, financially grounded decisions.
• Embedding agentic workflows that move from detection to action without human bottlenecks.
• Giving CPOs and executives instant visibility into risk, performance, and opportunity.

As Walt puts it: “If I can put this tool in the hands of even 400 or 500 CPOs, it’s game‑changing.”

EP2 is not a dashboard, not a chatbot, and not analytics theater. It is an agentic operating system for procurement leadership—forged from real battle scars, real data, and real outcomes, built to operate at executive speed.

Building the "Single Pane of Glass" for Modern Procurement

Pierpont Holdings partnered with TrueHorizon AI to transform procurement from a fragmented, reactive function into a real-time, agentic decision engine.

Led by Pierpont co-founder and seven-time Chief Procurement Officer Walt Charles III, the vision was bold and explicit:

"I wanted a single pane of glass for my world — one place where I can see my people, my risks, my spend, and where value is leaking, and then act on it immediately."

The result is EP2 (EP Squared) — an AI-enabled procurement analytics platform powered by three specialized AI agents that operate directly on Pierpont's existing Microsoft SQL infrastructure.

Together, these agents:

• Surface revenue at risk in seconds
• Allow executives to query millions of records in plain English

EP2: An Agentic Operating System for Procurement Leadership

The Vision: A Single Pane of Glass

Led by Pierpont co-founder and seven-time Chief Procurement Officer Walt Charles III, the vision for EP2 (EP Squared) was explicit and uncompromising:

"I wanted a single pane of glass for my world — one place where I can see my people, my risks, my spend, and where value is leaking, and then act on it immediately."

EP2 is an AI-enabled procurement analytics platform powered by three specialized AI agents that operate directly on Pierpont's existing Microsoft SQL infrastructure. Together, these agents:

• Surface revenue at risk in seconds
• Allow executives to query millions of records in plain English
• Automatically generate deep supplier intelligence and visual reports
• Eliminate weeks of manual analysis
• Enable procurement leaders to intervene before value is lost

The Problem: Procurement Without a Data Lens

Modern procurement leaders control trillions of dollars in spend, yet most operate with limited visibility and fragmented data. As Walt describes it:

"Most procurement leaders are flying blind. You have ticketing systems that tell you someone needs to buy something — but they don't tell you where risk is, who's accountable, or how much money is actually at stake."