ClearDox is Following an AI-Agent Strategy when Building Automated Operations Workflows
By Irina Reitgruber and originally published on CTRMCenter.
Automation of complex workflows is becoming essential for commodity companies. It reduces manual effort, decreases error rates, accelerates processes, and allows staff to focus on high-value activities. In today’s fast-moving market, this efficiency as well as ability to identify and processes risk insights are a significant competitive advantage.
Many companies are trying to build automation in-house, assuming their workflows are too bespoke for standard solutions. This “do-it-yourself” approach is typically slower, more expensive, harder to maintain, and less adaptable to technological advancements comparing to standardized vendor solutions. However, modern vendors can address the need for workflow flexibility through AI-agent frameworks, which can accommodate non-standard processes without requiring custom software development.
AI-agent architectures are rapidly emerging as the future of AI-enabled enterprise applications. At the core of this concept are AI agents – specialized tools trained to perform specific tasks – which can be orchestrated into workflows that require high degree of automation but cannot be automated by using conventional tools due to their complexity. Recent discussions with Katie Carter, Vice President, Product and Marc Lefebvre, CTO and Co-Founder from ClearDox provide detailed insights into the application of this strategy.
In ClearDox’s commodities intelligence platform, the model is built around two types of agents:
- Chat Agents
Interactive agents powered by an LLM and equipped with an automation toolset. They can download files, up- and download data to or from the ClearDox platform, draft emails, suggest mappings, approve or correct records, and perform other user-initiated tasks.
- Process Agents
Non-interactive agents that automate system processes. Examples include linking documents based on business rules, running prompts automatically, classifying insights, and pushing alerts to users based on events or thresholds.
Together, these agents support ClearDox Intelligent Applications by automating actions triggered by predefined prompts and allowing users to execute system actions directly from the application context accelerating issue resolution and information sharing.
Configuration of agents is performed through the UI, so no programming skills are required. Users can interact with Chat Agents through natural-language dialogue or via preconfigured action buttons that streamline common tasks.
To illustrate the agent’s application Katie provided examples of agent supporting data normalization and vessel operations workflows.
Mapping data fields is traditionally one of the most time-consuming tasks when normalizing data for building interfaces or setting up reconciliation processes. When an AI agent detects unmapped fields or issues preventing normalization, it provides a detailed explanation and recommends corrective actions such as creating new mappings. After user approval, the agent executes the changes automatically. “Human in the loop” is the essential part of concept of AI based automation.
In the example of vessel operations, the agent is configured to ingest documents such as bills of lading, vessel ETAs, or freight recaps as they appear along the life cycle of the trade, analyze the documents, link them to the appropriate trades and deliver concise summaries including identification of possible discrepancies or risks.
Looking ahead, Marc noted that ClearDox sees AI agents as the foundational technology for future AI applications in the commodity and energy sectors, given their flexibility and broad applicability. ClearDox also plans to introduce a third agent type – Application Agents – purpose-built agents embedded within specific applications to deliver structured, targeted automation for particular use cases.