.jpg)
At Spendcraft, we offer several key features like automated data normalization, self-service capabilities, and AI Agents. But to truly appreciate the value of these features, it’s important to understand why they matter and how they differentiate Spendcraft from traditional solutions.Let’s start by looking at how traditional spend classification tools typically work. Only by comparison can we fully appreciate the innovation Spendcraft brings to the table.
Here’s the typical process a traditional classification tool follows:
1. Data Extraction and Cleansing – Automated data extraction, but manual cleansing.
2. Normalization and Enhancement of Master Data – Manual normalization and enhancement of vendor and item data.
3. Training Dataset Preparation – Manually preparing the training dataset for AI/ML models like SVMs.
4. Dashboard and Report Creation – Automated template reports, but manually built reports for specific user needs.
Each of these steps requires manual intervention, either from the user or an implementation team. But why is this necessary? In most traditional systems, end users have no direct interaction with the AI model.
Think of the AI model as a worker bee and the end user as the beekeeper providing instructions. The problem is, there’s always a middleman—the implementer—who acts as a translator. The end user cannot directly communicate with the AI worker bee. They have to rely on the implementer, creating bottlenecks at every step depending on:
• How effective the AI worker bee is.
• The complexity of the inputs the AI requires, like training datasets and normalized data.
• The domain knowledge of the AI or the implementer.
At Spendcraft, we recognized this limitation and set out with a bold idea: eliminate the middleman and make the AI worker bee smarter. Our goal was to let users directly interact with the AI without needing an implementer. To achieve this, we identified and overcame the following challenges:
1. Data Cleansing, Normalization, and Enhancement
2. Custom Model Training for Each Customer or Taxonomy Change
3. Direct User Input
4. Continuous Learning and Improvement
We developed a Domain-Aware AI Agent that guides users, understands inputs, and learns from feedback. Here are the key features that solve these challenges:
• Automated Data Normalization
• Flexible Taxonomy
• Self-Service Classification
• Continuous Learning
Let’s dive into how each of these features works.
Data normalization isn’t just a buzzword—it’s a crucial step in spend classification. In Spend Craft, this process happens automatically the moment your data is ingested.
• Cleansing: As soon as data enters the system, advanced algorithms run a 17-step cleansing process. This removes whitespace, stop words, rare words, and performs stemming and lemmatization. Everything is cleaned and ready when users see it in the tool.
• Normalization: Spend Craft also automatically normalizes vendor and item data, including de-duplication of vendor records and enhancement with external web data (e.g., line of business, primary industry). The end user is then presented with enriched data for review and adjustments as needed.
The importance of flexible taxonomy is often underestimated. Businesses frequently need to classify their spend under different taxonomies or make adjustments to their existing structure.Traditional tools often discourage this, making reclassification a major effort.
Spendcraft offers:
• Taxonomy Templates: Users can choose from pre-built templates based on their business needs.
• Level 3 and 4 Flexibility: Start with a template and customize your taxonomy as needed.
• Continuous Modifications: You’re not locked into one taxonomy. You can reclassify data whenever you need or create a new taxonomy entirely from scratch.
Later this year, we’ll introduce further flexibility with Level 1 and 2 modifications, as well as a bring-your-own-taxonomy option.
In traditional classification tools, the AI model is often a black box, with end users relying on an implementer to interact with it. Spend Craft breaks this dependency.
With Spendcraft’s Domain-Aware AI Agent, end users can:
• Configure AI Agents to classify spend data themselves.
• Use a 4-5 step guided process to review a small set of records that represent the entire dataset.
• Once reviewed, the AI Agent can classify future datasets with no further setup—unless you change the taxonomy or data source.
Users can define multiple AI Agents to classify the same data under different taxonomies, offering unmatched flexibility.
Traditional models often require multiple cycles of back-and-forth between the user and implementer to correct inaccuracies. In contrast, Spend Craft’s self-service platform allows users to provide feedback directly to the AI Agent.
At every stage—from Data Normalization to Agent Configuration to Post-Classification—users can review records and make adjustments. This continuous learning loop ensures that the AI Agent improves over time based on user input.
Spendcraft includes built-in dashboards and reports, allowing users to:
• Instantly view the results of classified data.
• Provide feedback directly to the AI Agent from within the tool.
• Navigate between dashboards and classified records without leaving the platform.
This creates a seamless, closed-loop process for quicker insights and more efficient spend management.
In summary, Spendcraft solves the traditional challenges of spend classification with a smarter, more flexible approach. Our Domain-Aware AI Agent puts the power back into the hands of end users, offering a more efficient, scalable solution for spend classification.
With Spendcraft, you’re not waiting for the future of spend management—it’s already here.
Author :
Aditya Chavali
Share This Article:
Sign up For Newsletter
Stay tuned! Subscribe to our newsletter
