Snowflake's AI Streamlines Finance Analysis

Finance teams often drown in the repetitive grunt work of closing the books, a reality Snowflake’s FP&A department faced head-on. Hours were lost each month assembling spreadsheets, aligning data, and chasing budget versus actual analysis before any meaningful strategic work could begin. This manual grind, a common pain point for FP&A professionals hired to explain business performance, was ripe for disruption. The traditional process meant delays, visibility gaps, and a constant battle against formatting issues and organizational changes. Actuals from systems like Workday often had a two-hour refresh lag, further bottlenecking timely insights. This was the exact problem Snowflake’s internal finance team set out to solve using Snowflake Cortex Code, or CoCo.

StartupHub data

Companies working on this

Profiles of the companies named in this story, with funding and a one-liner from our database.

Workday

Workday is an enterprise resource planning software company focusing on AI for its next chapter.

Snowflake

$12.4B

A cloud-based data platform enabling data warehousing, data lakes, data engineering, and data sharing.

From Static Workbook to Live Dashboard

The appeal of CoCo lay in its native integration within Snowflake. This meant the AI coding agent already had context on the company’s data models, mappings, and forecast tables. Instead of onboarding data into a separate environment, analysts could describe their desired variance view in plain English, with CoCo translating it into action against existing Snowflake objects. This drastically shortened the path from concept to a functional tool. A clean hierarchy was established in Snowflake, enabling full drill-down capabilities from consolidated P&Ls to individual cost centers. CoCo then accelerated the build of a budget vs. actual dashboard in Streamlit, entirely replacing the static Excel workbook. The pipeline now runs continuously, pulling actuals directly into Snowflake without manual intervention, eliminating the lag and enabling real-time variance review during close.

This shift means FP&A teams no longer wait for workbooks to be assembled and distributed. They can open the dashboard during close, review variances as they emerge, and investigate discrepancies before formal review cycles even start. The prototype was ready within two weeks, and a fully functional application went live within a month, a speed previously requiring dedicated BI support and months of development. Stakeholder feedback could be incorporated the same afternoon, fundamentally changing the economics of getting solutions into the hands of users.

More Than a Dashboard: A Smarter Workflow

The new application offers dynamic views, allowing users to toggle across forecast versions and time periods. Adding new business units or cost centers is now seamless, with the dashboard automatically reflecting changes through the underlying data and hierarchy. The workflow is now better aligned with how FP&A professionals actually work, enabling them to drill down from consolidated P&Ls to specific cost centers to pinpoint variance drivers.

The transformation extends to AI-generated commentary. Selected expense categories now include AI-drafted explanations that analysts can review, refine, and publish. This turns a highly time-consuming part of the close process into a review task. CoCo is not just accelerating workflow creation; it's fundamentally changing task execution by providing a first draft for pressure-testing.

A collaborative layer allows cross-functional teams to post questions directly on line items, tag owners, and maintain a traceable history of issue investigation and resolution within the workflow. This keeps discussions attached to the relevant data, unlike scattered email chains. For leadership, a single presentation view consolidates the P&L and can export directly to Excel and PowerPoint, further reducing manual work before executive reviews.

Trust and Governance

Crucially, this automated workflow operates within Snowflake’s governance framework. Access is managed via role-based permissions, ensuring that sensitive financial data remains secure. Formatting, calculations, and hierarchies are standardized and locked, minimizing the risk of errors. Every action is timestamped and traceable, providing a level of auditability impossible with emailed workbooks. The intuitive interface also reduces the onboarding burden for new team members.

The most significant shift is empowering finance teams to build their own solutions. Analysts no longer need to wait in technical backlogs to test new views or workflows. They can prototype directly against production data and iterate rapidly with stakeholders. This democratizes development while maintaining enterprise-grade governance, shifting the bottleneck from technical assembly back to business judgment.

The Future of FP&A AI Workflow

This is just the beginning. The next phase involves expanding AI-generated commentary across the entire P&L, aiming to deliver a complete first draft of variance explanations at the start of every close. Forecast data will be integrated for forecast-vs-forecast reviews, allowing for dynamic story updates as numbers evolve. The vision includes agentic investigation flows that can drill down from a variance to its root cause, grounding explanations in detailed journal entries and business context.

Ultimately, this move to an automated budget vs. actual layer doesn't replace financial judgment; it protects and enhances it. By standardizing definitions, ensuring timely refreshes, and surfacing variances in familiar formats, FP&A teams can reclaim their core mission: explaining variances, aligning the organization, and driving critical business decisions, free from the manual tax of endless spreadsheet assembly.