AI
Artificial intelligence advancements continue to progress at a steady pace, changing the face of work. How well has your organization been keeping up with these changes? Have these breakthroughs represented a transformational shift in process and efficiency, or have they, despite well-meaning intentions and best efforts, only produced internal discussions or commitments around a nebulous solution that may take years to operationalize?
Core technology companies are all competing to provide the latest innovation. These efforts have brought to the forefront a new paradigm: the dawn of agents and agentic systems. Agents work on your behalf. This form of Synthetic LaborTM allows you to define the persona, the agent's role, and the available tools. For instance, a customer support agent could be described as handling support inquiries and being enabled with the ability to search through the available documentation and resolved support tickets.
KnowledgeLake has built an extensible platform for seamlessly integrating AI agents into your organization, allowing you to design and create a complex agentic system easily. To demonstrate, I will detail the configuration of an agentic workflow for invoice matching using the KnowledgeLake platform.
The workflow begins with an employee completing an order form for a recognized vendor. The form generates a purchase order, which is emailed to the specific vendor. From there, the work item is placed in a hold status. A vendor can submit the associated invoice through the KnowledgeLake Portal, or it can be imported through an email inbox. Once the invoice is received, a task is created for the receiving department. When the order is received, the goods receipt note is uploaded to complete the task. Next, the Intuitive AI activity extracts the line item information from the submitted documents. At this point, the Matching Agent reviews the data and makes a determination. If the agent is confident in the accuracy, it advances the work item along to initiate a payment. If it detects inaccuracies or other discrepancies, then a task is created with notes for human review.
The most compelling part of the orchestration is that the Matching Agent configuration is defined using natural language and can be modified or reformed until the desired result is achieved. As mentioned earlier, the description defines the persona, which is the role of the agent. In this case, the agent persona is described with the following: "You are an invoice matching specialist, meticulously reviewing and matching incoming invoices with corresponding purchase orders and receipts, ensuring accuracy and compliance with company policies. Perform three-way invoice matching by verifying the supplier's invoice details (invoice number, date, supplier details, item descriptions, quantities, unit prices, totals, and taxes) against the corresponding purchase order (item descriptions, quantities, unit prices) and goods receipt note (item descriptions, quantities). Indicate whether the documents match or if there is a mismatch and document the specific discrepancies between these three documents."
The ease and simplicity of this configuration process highlights how accessible and adaptable the KnowledgeLake platform is for various applications. This same methodology can be applied to solve other use cases such as loan processing, student enrollment, and underwriting, to name a few. With the KnowledgeLake platform, you will be able to take on challenges unbounded with the support of Synthetic Labor.
Tag(s):
AI