Finsights AI
Consolidated QuickBooks reporting with AI commentary for a finance platform serving 80+ advisory and accounting customers.
Headline outcome
80+
Customers served
The challenge
Where they started.
A finance reporting business with no technical team, serving 80+ customers with custom Power BI reports that were built by hand, per customer, every month. Every customer's books lived in its own QuickBooks file. The numbers came out through Excel exports. There was no ingestion pipeline and no shared data model, so each new customer meant another set of manual reports. Reporting was the product, and it could not scale past a few dozen customers.
How we work with Accounting & Advisory FirmsThe approach
What we actually built.
01
Unified ingestion layer
Fivetran for QuickBooks + bank feeds, one connection per company file. Custom Python in Airflow for the edges. Every customer's data isolated by design.
02
Full-stack analytics application
Onboarding flow, a multi-tenant data model shared across customers, and an in-app chat service. Embedded Sigma dashboards, so each customer sees only its own numbers.
03
dbt transformation
One codebase and one chart-of-accounts mapping for every customer. Row-level security enforced in the warehouse. Tests on every critical metric.
04
LLM-powered insights
AI-generated daily and weekly commentary on each customer's numbers, shipped as part of the product.
Results
The numbers that shifted.
80+
Customers onboarded
<1s
Query response at p95
90%
Manual reporting eliminated
Same problem?
Tell us where your setup looks like this one. We'll say what we'd do first.