Data Enrichment

Your database is full of "thin" records: just an email, or just a company name. Data Enrichment Workflows use AI to flesh out these records into full profiles.

The Strategy

Instead of paying for expensive enrichment APIs (which often have stale data), use an AI Agent equipped with a Web Browser tool. It can visit the company's actual website, read their current pricing page, and find their exact value proposition.

Implementation

1. The Trigger: db.row_created

Listen for new rows in your primary table (e.g. Leads or Companies).

cencori.on('db.companies.insert', async (row) => {
  await cencori.workflows.trigger('enrich-company', {
    companyId: row.id,
    domain: row.domain
  });
});

2. The Browser Agent

We define a workflow that spawns a Browser Agent.

const enrichment = await cencori.steps.run('browse-site', async () => {
    // 1. Visit the site
    const page = await browser.goto(`https://${domain}`);
    
    // 2. Scrape key pages
    const pricing = await page.goto('/pricing').msg;
    const about = await page.goto('/about').msg;
    
    // 3. Synthesize
    return { pricing, about };
});

3. The Extraction Step (Structured Output)

Now we feed the raw HTML/Text into an LLM and force it to output a specific JSON schema.

const schema = z.object({
  industry: z.string(),
  pricing_model: z.enum(['SaaS', 'Usage', 'Enterprise']),
  competitors: z.array(z.string()),
  summary: z.string()
});
 
const structuredData = await cencori.ai.generateObject({
  model: "gpt-4o",
  schema: schema,
  prompt: `Analyze this content: ${enrichment}`
});

4. The Action: Update Database

Finally, write the clean data back to your database.

await db.updateTable('companies')
  .set(structuredData)
  .where('id', '=', companyId)
  .execute();

Advanced: "Watcher" Workflows

You can also run this periodically.

• Trigger: schedule.monthly
• Action: Visit the pricing page again.
• Logic: Did the price change?
• Notification: Alert the sales team: "Competitor X just raised their prices!"