Why I built Vega?

Business teams across finance, marketing, and ops are stuck copying data across tools, waiting on analysts, and wrestling with inconsistent workflows. Vega eliminates that friction by combining familiar tools with AI.

Take a financial model update for example:

• Financial analysts often start by downloading data from their ERP.
• This data is then manually copied and pasted into an existing Excel model.
• Formulas need to be checked and updated to reflect the new data.
• Finally, the updated results are shared with stakeholders.
• This entire process is manual, error-prone, and incredibly time-consuming, often requiring multiple iterations as new data arrives or errors are discovered during review.

Or take an AR reconciliation for example:

• The process typically begins by exporting source datasets from an ERP, often into Excel.
• Concurrently, target datasets are exported from payment platforms like Stripe, also usually into Excel.
• The reconciliation itself is a manual effort, involving looking up individual invoices and meticulously matching them to remittance data.
• Findings are then presented, typically in Excel, and any necessary adjustments require manual journal entries back into the ERP.

Either repeat manually ad inifinitum, or invest a ton of money and resources into specialized implementation of a new BI, Analytics, or ERP platform that promises to organize your data and automate reporting.

The promise of AI is different

Snowflake

Excel

Python

ERP

AI should just work with your existing tools. It should execute complex workflows seamlessly across systems, just as a skilled colleague would.

By operating inside the tools you already use — spreadsheets, Python, and Snowflake — Vega helps any business team act on their data without new dashboards, migrations, or retraining.

The coolest part about Vega is how it works

Vega implements multi-step tool calling to generate Python or Javascript (for excel), it then sends this code to the respective execution envs, executes it, analyses the output, and continues until it's task is finished. To seamlessly interoperate between env's, Vega implements an Asset store that allows it to pass and persist dataframes with ease.

Technical Architecture

# Data processing
import pandas as pd
from vega import store

# Get data from store
data = store.get('financial_data')
df = pd.DataFrame(data)

# Process and transform
df['growth'] = df['revenue'].pct_change()

# Save results back to store
store.set('processed_data', df.to_dict())

// Excel automation
import { store } from 'vega';

async function updateExcel() {
  // Get processed data
  const data = await store.get('processed_data');
  
  // Update Excel worksheet
  await Excel.run(async (context) => {
    const sheet = context.workbook.worksheets
      .getActiveWorksheet();
    
    // Add data to sheet
    const range = sheet.getRange("A2:C10");
    range.values = formatForExcel(data);
    
    await context.sync();
  });
}

Hybrid Excel/Python Notebooks

Connect a spreadsheeet/database and ask away, Vega will tirelessly clean, scavenge, and visualize your data to answer your question. Unless it's a trivial question that requires a simple query, Vega will generate a Notebook that consists of cells in either markdown, python, or javascript. Notebooks are distinct from chats and can be deterministically run at a future point in time. This enables you to use Vega to explore data and build a pipeline, then to re-run it later without having to actually chat with Vega and regenerate the code, just rerun the notebook!

Customize your workflows by uploading your own data sources, adjusting logic, and rerunning notebooks with updated inputs. Vega adapts to your unique business needs.

Customizable and Self-Improving Knowledge Engine

Steer Vega with your own prompts to implement behavioral rules, memory for workflows, or even implement a task management system. Vega implements a powerful knowledge engine that observes usage and generates memories to optimize user experience, you can edit these memories for complete control over how Vega interacts.

Pricing & Early Adopters

Contact chris@patterns.app to get access to Vega. We're seeking forward-thinking partners to refine Vega and demonstrate its transformative potential in financial operations.

• We are looking to work with consulting firms to share implementation responsibilities and collaboratively develop best-in-class customer support models. Your expertise in diverse client environments and ability to create reusable automation logic and domain-specific prompts will be invaluable.
• We are looking to work with finance teams within corporations (and other teams driving reporting and automation inside their business functions) of all sizes who are feeling the pain of manual processes and are eager to pilot AI-driven solutions for tangible efficiency gains.
• We are also interested in engaging with independent financial analysts and smaller advisory practices who can leverage Vega to scale their services, offer more sophisticated data-driven insights to their clients, and develop domain-specific prompts.

If you work in or support a team that builds reports, monitors performance, or makes decisions based on messy data — Vega is your AI assistant. Get in touch to test it.

chris@patterns.app