From Raw Text to Structured Data: How Lexi Processes Unstructured Information

In today's digital world, businesses are drowning in a sea of unstructured data. From customer support emails and social media comments to legal documents and market reports, this raw text contains a goldmine of insights. The problem? Extracting that value is traditionally a manual, time-consuming, and expensive process.

What if you could automatically understand, categorize, and act on this information in real-time?

Enter Lexi, your autonomous AI language expert. Lexi isn't just another language model; it's an agentic workflow platform designed to transform chaotic, unstructured text into clean, structured, and actionable data. This is more than just Natural Language Processing (NLP)—it's Language-as-a-Service, ready to be integrated into any application with a single API call.

The Billion-Dollar Problem of Unstructured Data

For most organizations, unstructured text is a major operational bottleneck. Consider the manual effort involved in:

• Sifting through customer feedback to identify recurring issues and sentiment trends.
• Reviewing legal contracts to extract key clauses, dates, and obligations.
• Manually routing support tickets to the correct department based on the email's content.
• Analyzing product reviews to inform your development roadmap.

Each of these tasks requires human intelligence to read, comprehend, and categorize information. This manual processing is not only slow and prone to error but also means valuable opportunities are missed while your team is buried in text.

How Lexi Turns Chaos into Clarity

Lexi operates as an autonomous digital worker, capable of executing complex, multi-step language tasks. Instead of just giving you a raw model output, Lexi's 'Business-as-Code' approach delivers a reliable, valuable outcome. Here’s how it works when processing unstructured information.

Let's take a common example: a customer support email.

Raw Text Input:

"Hi team, I'm having trouble with my recent order #INV-12345. The checkout page keeps throwing a 500 error when I try to apply my discount code 'SUMMER25'. I'm on Chrome on a Mac. It's pretty frustrating as I wanted to get this order in today. Thanks, Alex"

A standard language model might summarize this. Lexi, the autonomous agent, performs a complete analysis workflow.

Step 1: Ingestion and Goal Setting

First, you send the raw text to Lexi via our API, defining the structure you want as the output. You're not just asking "what does this say?"—you're telling Lexi exactly what information you need.

Step 2: Classification and Identification

Lexi reads and comprehends the text. It immediately identifies the core intent. It recognizes keywords like "trouble," "error," and "frustrating" to classify this as a technical_support issue with negative sentiment and high urgency.

Step 3: Entity Extraction

Next, Lexi combs through the text to pull out specific, predefined pieces of data (entities). It methodically finds:

• Order Number: INV-12345
• Error Code: 500
• User Platform: Chrome on a Mac
• Context: Discount code 'SUMMER25'

Step 4: Analysis and Structuring

Finally, Lexi assembles all this processed information—the classification, sentiment, urgency, and extracted entities—into a clean, machine-readable JSON object that conforms to the schema you requested. The raw, messy email has been transformed into perfectly structured data.

From Code to Actionable Data

Integrating this powerful capability is incredibly simple. With the lexi.do SDK, you can define your desired output schema and let the agent handle the rest.

import { lexi } from '@do/sdk';

const customerEmail = `
  Hi team,
  I'm having trouble with my recent order #INV-12345.
  The checkout page keeps throwing a 500 error when I try to apply my discount code 'SUMMER25'. 
  I'm on Chrome on a Mac. It's pretty frustrating as I wanted to get this order in today.
  Thanks,
  Alex
`;

// Lexi analyzes the text and returns structured data based on your schema
const analysis = await lexi.structureText({
  text: customerEmail,
  schema: {
    topic: "billing | technical_support | general_feedback",
    sentiment: "positive | neutral | negative",
    urgency: "low | medium | high",
    summary: "A short summary of the user's issue.",
    entities: {
      orderNumber: "The invoice or order number mentioned.",
      errorCode: "The specific error code or message.",
      platform: "The user's OS or browser."
    }
  }
});

console.log(JSON.stringify(analysis, null, 2));

/* Output:
{
  "topic": "technical_support",
  "sentiment": "negative",
  "urgency": "high",
  "summary": "User is unable to apply discount code 'SUMMER25' to order #INV-12345 due to a 500 error on the checkout page.",
  "entities": {
    "orderNumber": "INV-12345",
    "errorCode": "500",
    "platform": "Chrome on a Mac"
  }
}
*/

The Agentic Difference: Beyond Just Analysis

The true power of lexi.do lies in its agentic nature. The structured data output isn't the end of the line—it's the beginning of an automated workflow.

Because you now have a predictable JSON object, you can build systems that:

• Automatically route any technical_support ticket with high urgency to your on-call engineering team's Slack channel.
• Create a ticket in Jira or Asana, pre-populated with the summary and entities.
• Log the sentiment in a database to track customer satisfaction trends over time.

Lexi doesn't just process language; it enables end-to-end automation, turning a simple language task into a sophisticated business process.

Unlock the Value in Your Text Today

Stop letting the valuable insights in your unstructured data go to waste. With Lexi, you can deploy an autonomous AI language agent that works 24/7 to classify, extract, and structure information, fueling smarter decisions and more efficient operations.

Ready to turn your raw text into structured, actionable intelligence? Explore what Lexi can do for you at lexi.do.