As businesses navigate the complexities of modern operations, the demand for intelligent tools that don’t just answer questions but actively solve problems has never been higher. Charlie, Fortude’s AI-powered virtual assistant, drives productivity by simplifying tasks and accelerating processes. Built on the Retrieval-Augmented Generation (RAG) framework, Charlie excels at delivering precise and context-aware responses by blending retrieval and generation models.

This blog explores how we are extending Charlie’s core strengths to incorporate agentic capabilities, transforming it from a powerful knowledge assistant into a proactive, solution engine. We will detail four groundbreaking use cases that leverage this new architecture to drive efficiency and foresight in enterprise environments.

Understanding agentic AI 

Agentic AI represents the evolution of artificial intelligence towards autonomous agents that can perceive their environment, make decisions, and take actions to achieve specific goals. Unlike traditional AI tools that are purely reactive, agentic systems use decision-making frameworks to choose the best actions and can adapt to changing environments, often learning from experience using techniques like reinforcement learning.

Furthermore, in complex enterprise settings, these agents often operate in multi-agent systems, communicating and cooperating to achieve common goals. This capability is proving vital across various industries, including finance, healthcare, and supply chain management.

Charlie: From knowledge assistant to proactive agent 

Charlie is a central piece of Fortude’s Automation product portfolio, which also includes the dedicated HR knowledge assistant, Ellie. Charlie already uses a skills Manager to execute tasks like creating tickets and automating routine actions, making it a proactive tool for efficiency, notably supporting functions like HR self-service and currently piloting an internal HR use case by integrating the employee handbook.

Expanding Charlie’s capabilities with CharlieX

While Charlie specializes in knowledge retrieval and task automation, CharlieX complements it by focusing on enterprise intelligence. It connects directly to ERP and CRM systems, allowing anyone, regardless of technical skill, to ask business questions in natural language, uncover hidden patterns, and detect anomalies before they become business issues. By supporting advanced querying and rapid context parsing, CharlieX works to democratize access to Business Intelligence (BI) across the enterprise, regardless of technical skill. Its primary benefits include drastically reducing insight generation time and automating the detection of patterns and anomalies within data.

These capabilities are being extended in our roadmap, with the aim of making Charlie more agentic in the future. Let’s take a quick look at the new features that will support this goal.

1. Signal-based inventory levelling agent 

Designed for supply chain managers, planners, and retail operations teams, this capability tackles the common problem of imbalanced inventory across multiple stores or warehouses.

The challenge: Retailers frequently experience stock-outs in some locations, leading to lost sales, while simultaneously accumulating excess stock elsewhere, which inflates carrying costs and ties up working capital. Manual redistribution decisions are often slow, reactive, and lack a proactive mechanism to predict demand variations.

The agentic solution: This AI-powered inventory levelling agent uses machine learning to forecast demand based on historical sales and demand signals (like promotions and seasonality). It then assesses inventory risks and proactively recommends balancing actions across the network.

Key outcomes: The agent delivers actionable guidance through:

• Stock out risk assessments: Highlighting stores likely to run out soon.
• Multi-store optimization: Identifying which stores should donate stock and which should receive replenishment.
• Levelling recommendations: Providing near real-time, data-driven redistribution or replenishment steps.

2. Model Context Protocol (MCP) for Infor ERP

This capability facilitates secure and contextual interaction between AI agents and the Infor M3 ERP system.

The challenge: Infor M3 ERP currently lacks a unified, AI-accessible interface layer. This means that every new automation or AI use case requires redundant API and logic design, and there is no standardized protocol to define intent, context, and response formats.

The agentic solution: We introduced an MCP server for Infor. This reusable, structured interface allows AI agents (like Charlie) to interact with Infor APIs using context-aware prompts. The MCP defines the exposed Infor APIs as tools for AI agents and establishes a secure authentication and permission model.

Key outcomes: This solution provides a centralized control and governance layer, enabling:

• A reusable protocol layer: A single MCP definition can support multiple automation use cases with consistency.
• AI-ready ERP access: Enables natural language agents to securely interact with ERP data.
• Faster time-to-automation: Reducing the need for bespoke integrations for every new use case.

3. Demand forecasting and inventory optimization for fashion 

This agent is built to help retail planners, inventory managers, and supply chain analysts anticipate demand changes by leveraging external market intelligence.

The challenge: Retail and fashion businesses struggle with stockouts or overstocking due to inaccurate forecasts. Traditional planning processes are often manual and siloed, failing to incorporate external factors such as promotions, weather, or customer sentiment that heavily influence demand.

The agentic solution: The Signal-Based Forecasting Agent integrates internal business data (including inventory, sales tickets, and purchase orders) with a wide array of external market signals. These signals are categorized into short-term (e.g., weather, holidays), medium-term (e.g., emerging styles, campaigns), and long-term (e.g., historical sales trends) influencers.

Charlie uses AI/ML models to combine this internal data with weighted external signals. The agent then forecasts SKU level demand and suggests future purchase order (PO) quantities and required in-house dates across various time windows.

Key outcomes:  This automation:

• Reduces stockouts and excess inventory by anticipating real demand.
• Incorporates external signals to dramatically improve forecast accuracy.
• Automates the creation of data-backed PO recommendations, freeing planners from manual spreadsheet analysis.

4. M3 release impact analysis 

This use case is designed for ERP consultants and release managers supporting Infor M3 clients, helping them identify which feature releases will impact their business and plan accordingly.

The challenge: Infor M3 publishes multiple release updates annually, requiring consultants to manually read detailed release notes, identify affected modules, and cross-check changes against each client’s custom configurations and source code. This process is error-prone, highly repetitive, and risks missing critical compliance tasks.

The agentic solution: Charlie’s multi-agent architecture completely automates the release analysis process. When a new M3 release report is published, a coordinated team of agents goes into action:

• A Release Report Analyzer Agent parses the document.
• The Client Resource Analyzer and Source Code Agent check the client’s configurations and customizations.
• An Impact Analysis Agent compares the changes with the client-specific setup.
• A Jira Integration Agent then auto-generates tickets detailing the required changes, complete with full context and module references. These tickets are automatically assigned to the correct consultant.

Key outcomes: This agentic workflow provides consultants with a complete, pre-assessed set of tasks before they even open the release report. It saves 80–90% of the time spent on manual analysis, prevents missed critical changes, and builds client trust through consistent, intelligent foresight