PE Portfolio Company Gives Teams a Natural-Language Data Analyst
A PE portfolio company gave finance and operations a natural language data analyst across five systems, turning a four day wait for answers into 14 seconds and documenting $3.6M in run rate savings in 90 days.
The Challenge
The PE sponsor required an 8 to 10 percent cut in SG&A within 180 days while holding revenue. Data lived in five isolated systems (Postgres, a Snowflake finance mart, Salesforce, a WMS, and freight invoices), so ad hoc questions took four to seven days, savings opportunities were missed, dashboard adoption was low, and a small BI team created a bottleneck.
What They Built
Eliza built a natural language agent that lets teams ask questions in plain English through Slack or a web interface. Read only connectors to Postgres and Snowflake enforce SSO and role based access, and an automatic semantic layer maps business terms to the right tables and fields. SQL is generated behind guardrails (cost limits, PII redaction, sample before full run), and answers come back as tables, charts, narratives, and one click follow ups, with Slack and email subscriptions for recurring questions.
Read only database roles with row level security kept access safe, a lightweight approval workflow covered sensitive queries, and 90 minute enablement sessions with embedded prompt guidance drove adoption. First value came in three to four weeks, focused on freight, inventory, and pricing.
AI Role
AI turns plain-English questions into governed, read-only SQL, runs it against the warehouse, and returns tables, charts, and narratives with full source lineage, keeping humans in control of decisions.
Infrastructure
Postgres, Snowflake finance mart, Salesforce, WMS, freight invoice system; read only roles with row level security and SSO/RBAC
Integration Points
Read only connectors to Postgres and Snowflake; Slack and web interface; Slack and email subscriptions for recurring queries
Impact
$3.6M in run rate savings in 90 days, verified by SQL lineage and finance sign off (freight $1.2M, inventory $1.1M, pricing $900K, analytics $400K)
Median time to answer a business question fell from about four days to 14 seconds
120+ monthly active users at a 92 percent useful rating on first attempt answers
Implementation Complexity
Best Fit For
Finance and operations leaders at PE backed companies who need self serve, auditable answers fast, especially where data is spread across warehouses, CRMs, and operational systems.
